Medical image analysis method, medical image analysis device, and medical image analysis system

Abstract

To provide a medical image analysis method for providing information related to a medial image.SOLUTION: A medical image analysis method includes: acquiring a correction parameter calculated on the basis of a correlation between a first morphological value related to a target element acquired from a first medical image acquired under a first scan condition and a second morphological value related to the target element acquired from a second medical image acquired under a second scan condition; acquiring a target medical image acquired under the second scan condition; acquiring a target area related to the target element from the target medical image by executing the segmentation of the target medical image in a plurality of areas corresponding to a plurality of elements including the target element; and acquiring a target morphological value related to the target element on the basis of voxel data corresponding to the target area.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present application relates to a medical image analysis method, a medical image analysis device, and a medical image analysis system for analyzing medical images. [Background technology]

[0002] With the advancement of image segmentation technology, it has become possible to segment medical images and calculate diagnostic support indicators related to diseases, and the field of medical image analysis has recently been attracting attention.

[0003] In particular, medical image analysis technology is being used in many areas to provide diagnostic support indicators for dementia, and it is essential to more accurately calculate the morphological characteristics of specific regions from medical images in order to provide objective diagnostic support information for dementia.

[0004] However, conventional medical image analysis technologies have various limitations due to the presence of errors in the medical images themselves, such as the inclusion of artifacts in the medical images that are not suitable for image segmentation, and have limitations in that auxiliary diagnostic information for the same subject may vary depending on the scan conditions under which the medical image is acquired.Furthermore, conventional medical image analysis technologies have a problem in that they cannot provide completely "personalized" auxiliary diagnostic information for the subject because they use a method of performing image segmentation after standardizing the brain contained in the medical image to a standard brain model, and therefore have limitations in situations where high accuracy is required.

[0005] Therefore, there is a need for an image analysis method and device that can provide a user with determinable diagnostic information even when noise is included in the image, and there is a need for an image analysis method and device that can obtain auxiliary diagnostic information by taking into account the scanning conditions under which the medical image was taken, and obtain auxiliary diagnostic information personalized for the subject. Summary of the Invention [Problem to be solved by the invention]

[0006] An object of the present invention is to provide a medical image analysis method, a medical image analysis device, and a medical image analysis system that provide information related to medical images.

[0007] The problems to be solved by the present invention are not limited to those described above, and problems not mentioned should be clearly understood by those skilled in the art from the present specification and drawings. [Means for solving the problem]

[0008] The medical image analysis method disclosed in the present application may include: acquiring a correction parameter calculated based on a correlation between a first morphological value associated with a target element acquired from a first medical image acquired under first scanning conditions and a second morphological value associated with the target element acquired from a second medical image acquired under second scanning conditions; acquiring a target medical image acquired under the second scanning conditions; segmenting the target medical image into a plurality of regions corresponding to a plurality of elements including the target element to acquire a target region associated with the target element from the target medical image; acquiring a target morphological value associated with the target element based on voxel data corresponding to the target region; acquiring a corrected morphological value based on the target morphological value and the correction parameter; and outputting a morphological index based on the corrected morphological value.

[0009] The medical image analysis device disclosed in the present application may include an image acquisition unit that acquires a target medical image; and a controller that provides medical image analysis information based on the target medical image. The controller may be configured to acquire correction parameters calculated based on a correlation between a first morphological value associated with a target element acquired from a first medical image acquired under first scanning conditions and a second morphological value associated with the target element acquired from a second medical image acquired under second scanning conditions; acquire the target medical image acquired under the second scanning conditions; acquire a target region associated with the target element from the target medical image by segmenting the target medical image into a plurality of regions corresponding to a plurality of elements including the target element; acquire target morphological values ​​associated with the target element based on voxel data corresponding to the target region; acquire corrected morphological values ​​based on the target morphological values ​​and the correction parameters; and output morphological indices based on the corrected morphological values.

[0010] The medical image analysis method disclosed in the present application includes acquiring a target medical image, acquiring target scan conditions associated with the target medical image, acquiring target morphological values ​​associated with the target element based on voxel data of a target area corresponding to the target element included in the target medical image, determining a target correction parameter from one or more correction parameters based on the target scan conditions, and outputting a morphological index based on the determined target correction parameter and the target morphological value. Determining the target correction parameter may include, if the target scan condition corresponds to a first scan condition, determining the target correction parameter as a first correction parameter for correcting a first morphological value associated with the target element acquired under the first scan condition, and if the target scan condition corresponds to a second scan condition, determining the target correction parameter as a second correction parameter for correcting a second morphological value associated with the target element acquired under a second scan condition different from the first scan condition.

[0011] The medical image analysis device disclosed in the present application includes an image acquisition unit that acquires a target medical image; and a controller that provides medical image analysis information based on the target medical image. The controller is configured to acquire target scan conditions related to the target medical image, acquire target morphological values ​​associated with the target element based on voxel data of a target area corresponding to the target element included in the target medical image, determine a target correction parameter from among one or more correction parameters based on the target scan conditions, and output a morphological index based on the determined target correction parameter and the target morphological value. If the target scan conditions correspond to the first scan conditions, the controller may be configured to determine the target correction parameter as a first correction parameter for correcting a first morphological value associated with the target element acquired under the first scan conditions, and if the target scan conditions correspond to the second scan conditions, determine the target correction parameter as a second correction parameter for correcting a second morphological value associated with the target element acquired under second scan conditions different from the first scan conditions.

[0012] The means for solving the problem of the present invention are not limited to the above-mentioned means, and the means not mentioned should be clearly understood by those skilled in the art from this specification and the attached drawings. [Effects of the Invention]

[0013] According to an embodiment of the present application, a quality analysis is performed on a target medical image, and information on the analysis result is provided to a user, thereby increasing the reliability of the medical image analysis result.

[0014] According to an embodiment of the present application, morphological indices are calculated based on medical images of the object, and correction parameters are appropriately applied taking into account the scanning conditions under which the medical images are obtained and the position of the target element, thereby enabling more accurate morphological indices to be calculated.

[0015] According to the embodiment of the present application, it is possible to selectively provide the user with necessary index information from among various index information acquired through the analysis of the target medical image, thereby improving ease of use.

[0016] The effects of the present invention are not limited to those described above, and effects not mentioned herein should be clearly understood by those skilled in the art from this specification and the accompanying drawings. [Brief explanation of the drawings]

[0017] [Figure 1] 1 is a schematic diagram of a medical image analysis system according to an embodiment of the present application; [Figure 2] 2 is a schematic diagram of an image analysis device 2000, a learning device 2200, a correction parameter acquisition device 2400, and an output device 2600 for image analysis according to an embodiment of the present application. [Figure 3] 1 is a block diagram of an image analysis device according to an embodiment of the present application; [Figure 4] FIG. 1 is a block diagram of an output device according to an embodiment of the present application. [Figure 5] FIG. 2 illustrates an example of an image alignment operation of the image analysis device 2000. [Figure 6] FIG. 2 illustrates an example of an image alignment operation of the image analysis device 2000. [Figure 7] FIG. 1 illustrates a process flow for image segmentation according to one embodiment of the present application. [Figure 8] 22 is a flowchart illustrating a neural network model learning method of the learning device 2200 according to an embodiment of the present application. [Figure 9] FIG. 2 is an exemplary structural diagram of an image dataset according to an embodiment of the present application. [Figure 10] 2 is an example of an artificial neural network model that can be used by a learning device 2200 according to an embodiment of the present application. [Figure 11] 22 is another example of an artificial neural network model that can be used by the learning device 2200 according to an embodiment of the present application. [Figure 12] 2 is a flowchart illustrating an image segmentation method using a neural network model in an image analysis device 2000 according to an embodiment of the present application. [Figure 13] FIG. 2 is an exemplary structural diagram of a target image according to an embodiment of the present application; [Figure 14] 2 is an exemplary diagram of an image obtained based on the result values ​​obtained by the segmentation process of the image analysis device 2000 according to an embodiment of the present application. [Figure 15] FIG. 1 illustrates a process flow for image segmentation according to one embodiment of the present application. [Figure 16] FIG. 2 is an exemplary structural diagram of an image dataset according to an embodiment of the present application. [Figure 17] FIG. 1 illustrates a process flow for image segmentation according to one embodiment of the present application. [Figure 18] FIG. 1 is a diagram illustrating an image quality determination process according to an embodiment. [Figure 19] FIG. 1 is a diagram illustrating an image analysis device that performs an image quality assessment process. [Figure 20] FIG. 10 is a diagram for explaining an image quality determination module. [Figure 21] FIG. 10 is a diagram for explaining a first image quality determination process. [Figure 22] FIG. 2 is a diagram for explaining a first image quality determining unit. [Figure 23] FIG. 10 is a diagram for explaining a second image quality determination process. [Figure 24] FIG. 10 is a diagram for illustratively explaining that the second image quality assessment process uses an artificial neural network model to determine the location of artifacts in a medical image. [Figure 25] FIG. 1 is a diagram illustrating a neural network model for acquiring artifact information according to an embodiment. [Figure 26] FIG. 10 is a diagram for explaining the third image quality determination process. [Figure 27] FIG. 10 is a diagram for explaining an embodiment of a third image quality determination process. [Figure 28] FIG. 10 is a diagram for explaining the fourth image quality determination process. [Figure 29] FIG. 10 is a diagram illustrating an image in which an artifact occurs. [Figure 30] 10 is a diagram illustrating determining whether an artifact occurs overlapping with a region of interest in the fourth image quality determination process according to another embodiment. FIG. [Figure 31] FIG. 10 is a diagram illustrating an image quality related information output module according to an embodiment. [Figure 32] FIG. 10 is a diagram for explaining an example of a first error information screen. [Figure 33-34] FIG. 10 is a diagram for explaining an example of a second error information screen. [Figure 35-36] FIG. 10 is a diagram for explaining an example of a third error information screen. [Figure 37-38] FIG. 10 is a diagram for illustrating an example of a fourth error information screen. [Figure 39] 39 is a diagram for explaining an example of a selection window output by the image quality related information output module 3900. FIG. [Figure 40] 39 is a diagram for explaining an example of a selection window output by the image quality related information output module 3900. FIG. [Figure 41-43] FIG. 10 is a diagram for explaining the first to third quality determination processes. [Figure 44-45] FIG. 10 is a diagram for explaining a fourth quality determination process. [Figure 46-47] FIG. 10 is a diagram for explaining the fifth quality determination process. [Figure 48]2 is a flow chart illustrating one operation of an image analysis method implemented by an image analysis device 2000 according to an embodiment of the present application. [Figure 49] FIG. 20 is an exemplary diagram of an image alignment method implemented by an image analysis device 2000 according to one embodiment of the present application. [Figure 50] 1 is a flow diagram of an image analysis method according to an embodiment of the present application. [Figure 51] 1A-1C illustrate an example of a method for modifying a first boundary of a first internal region of a skull according to an embodiment of the present application. [Figure 52] 1 is a diagram showing an example of a second internal region obtained by a method for correcting a first boundary of a first internal region of a skull according to an embodiment of the present application; [Figure 53] 10A-10C illustrate another example of a method for modifying a first boundary of a first internal region of a skull according to an embodiment of the present application. [Figure 54-57] FIG. 10 is an exemplary diagram illustrating morphological indices and associated information output by an embodiment of the present application. [Figure 58] 2 is a flow chart illustrating one operation of a morphological value correction method implemented by the image analysis device 2000 according to one embodiment of the present application. [Figure 59] 1 is a flow diagram of an image analysis method according to an embodiment of the present application. [Figure 60] 2 is a flowchart illustrating one operation of a correction parameter acquisition method implemented by a correction parameter acquisition device 2400 according to an embodiment of the present application. [Figure 61] FIG. 2 is an exemplary diagram illustrating information contained in multiple image data sets according to one embodiment of the present application. [Figure 62] 10 is a graph illustrating an example of a correlation analysis for acquiring correction parameters of the correction parameter acquisition device 2400 according to an embodiment of the present application. [Figure 63] 10 is a graph illustrating another example of correlation analysis for acquiring correction parameters of the correction parameter acquisition device 2400 according to an embodiment of the present application. [Figure 64]FIG. 2 is an exemplary diagram illustrating information contained in multiple image data sets according to one embodiment of the present application. [Figure 65] FIG. 10 is an exemplary diagram of a correlation analysis between morphological values ​​according to the location where target elements are distributed within the skull according to an embodiment of the present application. [Figure 66] 1 is a flow diagram of an image analysis method according to an embodiment of the present application. [Figure 67] FIG. 2 is a schematic diagram of a user interface according to an embodiment of the present application. [Figure 68] FIG. 10 is a diagram for explaining a medical information output process according to one embodiment. [Figure 69] 1 is a diagram illustrating an image analysis device according to an embodiment; [Figure 70] FIG. 10 is a diagram for explaining a medical data acquisition module. [Figure 71] FIG. 10 is a diagram for explaining a medical information acquisition module. [Figure 72] FIG. 10 is a diagram for explaining a diagnostic auxiliary information acquisition module. [Figure 73] FIG. 10 is a diagram for explaining a diagnostic auxiliary information output module. [Figure 74] FIG. 10 is a diagram illustrating relevant indices for several target diseases. [Figure 75] FIG. 10 is a diagram for explaining a medical information output screen according to an embodiment. [Figure 76] FIG. 10 is a diagram for explaining a medical information output process according to another embodiment. [Figure 77] 10 is a diagram illustrating an image analysis device according to another embodiment. [Figure 78] FIG. 10 is a diagram for explaining a selection information acquisition module. [Figure 79] FIG. 10 is a diagram for explaining a selection information output module. [Figure 80-81] FIG. 10 is a diagram for explaining a selection information output screen. DETAILED DESCRIPTION OF THE INVENTION

[0018] In one embodiment of a medical image analysis method according to the present application, the method includes the steps of: acquiring a target medical image including at least one artifact; acquiring artifact information related to the artifact based on the target medical image, the artifact information including an artifact area indicating an area where the artifact is distributed; acquiring a plurality of regions corresponding to one or more anatomically divided elements based on the target medical image using a first neural network model, the first neural network model being trained to acquire the plurality of regions based on the medical image, the plurality of regions including a first region corresponding to a first element, a second region corresponding to a second element, and a third region corresponding to a third element; acquiring a degree of overlap between the artifact region and a region of interest, the region of interest including at least a portion of the first region; performing a first quality judgment step of judging the quality of the target medical image based on the degree of overlap; and outputting quality-related information of the target medical image based on the first quality judgment.

[0019] The first element may be a skull, the first region may include a region corresponding to the skull, and the region of interest may be a region of the skull or an internal region of the skull.

[0020] The second element may be an element related to the diagnosis of a disease of interest, and the region of interest may include at least a portion of the second element.

[0021] The artifact region may be obtained using a second neural network model trained to determine whether an artifact is present in the medical image, and the artifact region may be a region on a feature map obtained from the second neural network model and associated with a target artifact present in the target medical image, the feature map having a correlation with the target artifact that is equal to or greater than a reference value.

[0022] The artifact region may include a first region and a second region located within the first region, and the first region may be a region on the feature map whose correlation with the target artifact is equal to or greater than a first reference value, and the second region may be a region on the feature map whose correlation with the target artifact is equal to or greater than a second reference value that is greater than the first reference value.

[0023] The artifact region can be obtained using a second neural network model trained to segment regions corresponding to the artifacts in the medical image.

[0024] The step of determining the degree of overlap may include obtaining an outline of the artifact region and an outline of the region of interest on the target medical image, and determining whether the artifact and the region of interest overlap based on whether the outline of the artifact region and the outline of the region of interest overlap on the medical image.

[0025] The first quality assessment step may include determining that the quality of the target medical image is normal if the number of intersections where the outline of the artifact region and the outline of the region of interest overlap on the target medical image is less than two, and determining that the quality of the target medical image is abnormal if the number of intersections is two or more.

[0026] The step of determining the degree of overlap may include the steps of: obtaining a number of common pixels contained in the artifact region and the region of interest on the target medical image; and determining the degree of overlap between the artifact and the region of interest using the number of common pixels.

[0027] The first quality judgment step may include judging that the quality of the target medical image is abnormal if the number of common pixels exceeds a predetermined reference value, and judging that the quality of the target medical image is normal if the number of common pixels is equal to or less than a predetermined reference value.

[0028] The method may further include a second quality judgment step of performing a quality judgment of the target medical image based on the third region, and the second quality judgment step may include a step of obtaining a morphological index based on the third region and a step of determining whether the morphological index meets a quantitative criterion.

[0029] The morphological indicator may be obtained based on the first region and the third region.

[0030] The morphological index may be obtained based on a ratio of a volume of the first region to a volume of the third region.

[0031] The third region may be a region corresponding to the bottommost element of the target medical image.

[0032] The third region may be a region corresponding to an element related to the diagnosis of the disease of interest.

[0033] The step of outputting quality-related information of the medical image may include displaying error information generated based on the first quality judgment and a selection window associated with the error information, and the error information may include information about the artifacts and the plurality of regions.

[0034] The selection window may include a first object, and the medical image analysis method may include proceeding with a subsequent operation in response to a user's selection of the first object, the subsequent operation including analyzing the target medical image, correcting the target medical image, or re-capturing the target medical image.

[0035] According to another embodiment of the present application, a medical image analysis device includes an acquisition unit for acquiring a medical image, a processing unit for acquiring image quality information based on the medical image, and an output unit for outputting the image quality information. The processing unit acquires a target medical image containing an artifact through the acquisition unit, acquires artifact information related to the artifact based on the target medical image, the artifact information including an artifact area indicating an area where the artifact is distributed, acquires a plurality of regions corresponding to one or more anatomical elements based on the target medical image using a first neural network model, the first neural network model being trained to acquire the plurality of regions based on the medical image, the plurality of regions including a first region corresponding to a first element, a second region corresponding to a second element, and a third region corresponding to a third element, acquires an overlap degree between the artifact region and a region of interest, the region of interest including at least a portion of the first region, performs a first quality judgment of the target medical image based on the overlap degree, and the output unit outputs quality-related information of the target medical image based on the first quality judgment.

[0036] The first element may be a skull, the first region may include a region corresponding to the skull, and the region of interest may be a region of the skull or an internal region of the skull.

[0037] The second element may be an element related to the diagnosis of a disease of interest, and the region of interest may include at least a portion of the second element.

[0038] The artifact region may be obtained using a second neural network model trained to determine whether an artifact is present in the medical image, and the artifact region may be a region on a feature map obtained from the second neural network model and associated with a target artifact present in the target medical image, the feature map having a correlation with the target artifact equal to or greater than a reference value.

[0039] The artifact region may include a first region and a second region located within the first region, and the first region may be a region on the feature map whose correlation with the target artifact is equal to or greater than a first reference value, and the second region may be a region on the feature map whose correlation with the target artifact is equal to or greater than a second reference value that is greater than the first reference value.

[0040] The artifact regions may be obtained using a second neural network model trained to segment regions corresponding to the artifacts in the medical image.

[0041] The processing unit can acquire the outline of the artifact region and the outline of the region of interest on the target medical image, and determine whether the artifact and the region of interest overlap based on whether the outline of the artifact region and the outline of the region of interest overlap on the medical image.

[0042] The processing unit may determine that the quality of the target medical image is normal if the number of intersections where the outline of the artifact region and the outline of the region of interest overlap on the target medical image is less than two, and may determine that the quality of the target medical image is abnormal if the number of intersections is two or more.

[0043] The processing unit may acquire the number of common pixels included in the artifact region and the region of interest on the target medical image, and determine the degree of overlap between the artifact and the region of interest using the number of common pixels.

[0044] The processing unit may determine that the quality of the target medical image is abnormal if the number of common pixels exceeds a predetermined reference value, and may determine that the quality of the target medical image is normal if the number of common pixels is equal to or less than the predetermined reference value.

[0045] The processing unit may obtain a morphological index based on the third region, and perform a second quality judgment based on whether the morphological index meets a quantitative criterion.

[0046] The morphological indicator may be obtained based on the first region and the third region.

[0047] The morphological index may be obtained based on a ratio of a volume of the first region to a volume of the third region.

[0048] The third region may be a region corresponding to the bottommost element of the target medical image.

[0049] The third region may be a region corresponding to an element related to the diagnosis of the disease of interest.

[0050] According to the brain image analysis method disclosed in the present application, the method may include: acquiring a brain image including voxel data; segmenting the brain image into a plurality of regions including a skull region and a reference region to obtain a first boundary and the reference region defining a first internal region inside the skull region; modifying at least a portion of the first boundary associated with the first internal region based on the reference region to obtain a second internal region having a second boundary, wherein a portion of the second internal region is included in the first internal region and includes the target element; acquiring a first volume value associated with the target element based on voxel data corresponding to the target element; acquiring a second volume value associated with the second internal region based on voxel data included in the second boundary; and calculating a volume index of the target element based on the first volume value and the second volume value.

[0051] The brain image analysis method disclosed in the present application may further include aligning the brain image to obtain the second internal region.

[0052] According to the brain image analysis method disclosed in the present application, aligning the brain image may include obtaining a first feature area and a second feature area from the brain image, calculating a first feature point from the first feature area and a second feature point from the second feature area, obtaining an imaging direction of the brain image based on the first feature point and the second feature point, and aligning the brain image so that the imaging direction is parallel to a reference direction.

[0053] According to the brain image analysis method disclosed in the present application, the first feature region may be a region corresponding to the anterior commissure, the second feature region may be a region corresponding to the posterior commissure, the first feature point may be included in a boundary defining the region corresponding to the anterior commissure, and the second feature point may be included in a boundary defining the region corresponding to the posterior commissure.

[0054] According to the brain image analysis method disclosed in the present application, aligning the brain image may include obtaining data related to the direction of the brain image included in the brain image, obtaining a photographing direction of the brain image based on the data related to the direction of the brain image, and aligning the brain image so that the photographing direction corresponds to a reference direction.

[0055] According to the brain image analysis method disclosed in the present application, obtaining the second internal region may include obtaining a reference plane adjacent to the reference region from the brain image, and modifying a portion of the first boundary of the first internal region based on the reference plane to obtain the second internal region having the second boundary.

[0056] According to the brain image analysis method disclosed in the present application, the reference plane is a plane parallel to a transverse plane of the brain image, and acquiring the second internal region may include modifying a portion of the first boundary located below the reference plane based on the first internal region located on the upper side of the reference plane, and acquiring the second internal region having the second boundary based on the modified first boundary.

[0057] According to the brain image analysis method disclosed in the present application, the reference region may be a region corresponding to the cerebellum, and the reference plane may be adjacent to the inferior edge of a boundary defining the corresponding region of the cerebellum.

[0058] According to the brain image analysis method disclosed in the present application, the reference region may be a region corresponding to the cerebellum, and the reference plane may be adjacent to the inferior edge of a boundary defining the corresponding region of the cerebellum.

[0059] According to the brain image analysis method disclosed in the present application, obtaining the second internal area includes obtaining a first feature area and a second feature area from the brain image, calculating first feature points from the first feature area and calculating second feature points from the second feature area, obtaining a reference direction connecting the first feature points and the second feature points based on the first feature points and the second feature points, obtaining a reference plane parallel to the reference direction, and modifying a part of the first boundary of the first internal area based on the reference plane to obtain the second internal area having the second boundary.

[0060] According to the brain image analysis method disclosed in the present application, the reference region may be a region corresponding to the neck bones, and the reference plane may be adjacent to the region corresponding to the neck bones and perpendicular to the sagittal plane.

[0061] According to the brain image analysis method disclosed in the present application, obtaining the second internal region may include modifying a portion of the first boundary located below the reference plane based on the first internal region located above the reference plane, and obtaining the second internal region having the second boundary based on the already modified first boundary.

[0062] According to the brain image analysis method disclosed in the present application, the volume index may be defined as a ratio of the first volume value to the second volume value.

[0063] According to the brain image analysis method disclosed in the present application, calculating the volume index of the target element may further include: acquiring a first correction parameter for correcting the first volume value, where the first correction parameter is obtained based on a scanning condition under which the brain image is acquired or a position of the target element, and acquiring a first volume correction value including calculating the first volume correction value based on the first correction parameter and the first volume value; acquiring a second correction parameter for correcting the second volume value, where the second correction parameter is obtained based on a scanning condition under which the brain image is acquired or a position of the target element, and calculating the second volume correction value based on the second correction parameter and the second volume value; and calculating the volume index based on the first volume correction value and the second volume correction value.

[0064] According to the brain image analysis method disclosed in the present application, the scanning conditions under which the brain image is acquired may be related to at least one of the following setting parameters: a magnetic field strength associated with the resolution of the brain image of the brain image acquisition device; a manufacturer of the brain image acquisition device; and a configuration of the magnetic field generated by the brain image acquisition device.

[0065] According to the brain image analysis device disclosed in the present application, the device may include an image acquisition unit that acquires brain images, and a controller that provides brain image analysis information based on the brain images. The controller may be configured to acquire brain images including voxel data, segment the brain images into a plurality of regions including a skull region and a reference region, thereby acquiring a first boundary and the reference region that define a first internal region that is inside the skull region, modify at least a portion of the first boundary associated with the first internal region based on the reference region, to acquire a second internal region having a second boundary, a portion of the second internal region being included in the first internal region and including the target element, acquire a first volume value associated with the target element based on voxel data corresponding to the target element, acquire a second volume value associated with the second internal region based on voxel data included in the second boundary, and calculate a volume index of the target element based on the first volume value and the second volume value.

[0066] According to the brain image analysis apparatus disclosed in the present application, the controller may be configured to align the brain image to obtain the second internal region.

[0067] According to the brain image analysis device disclosed in the present application, the controller may be configured to acquire a first feature region and a second feature region from the brain image, calculate a first feature point from the first feature region, calculate a second feature point from the second feature region, acquire an imaging direction of the brain image based on the first feature point and the second feature point, and align the brain image so that the imaging direction is parallel to a reference direction, thereby aligning the brain image.

[0068] According to the brain image analysis device disclosed in the present application, the first feature region may be a region corresponding to the anterior commissure, the second feature region may be a region corresponding to the posterior commissure, the first feature point may be included in a boundary defining the region corresponding to the anterior commissure, and the second feature point may be included in a boundary defining the region corresponding to the posterior commissure.

[0069] According to the brain image analysis device disclosed in the present application, the controller may be configured to acquire data related to the direction of the brain image included in the brain image, acquire a photographing direction of the brain image based on the data related to the direction of the brain image, and align the brain image so that the photographing direction corresponds to a reference direction, thereby aligning the brain image.

[0070] According to the brain image analysis device disclosed in the present application, the controller may be configured to acquire the second internal region by acquiring a reference plane adjacent to the reference region from the brain image, and modifying a portion of the first boundary of the first internal region based on the reference plane to acquire the second internal region having the second boundary.

[0071] According to the brain image analysis device disclosed in the present application, the reference plane is a plane parallel to a transverse plane of the brain image, and the controller may be configured to obtain the second internal region by modifying a portion of the first boundary located below the reference plane based on the first internal region located on the upper side of the reference plane, and obtaining the second internal region having the second boundary based on the modified first boundary.

[0072] According to the brain image analysis device disclosed in the present application, the reference region may be a region corresponding to the cerebellum, and the reference plane may be adjacent to the inferior edge of a boundary defining the corresponding region of the cerebellum.

[0073] According to the brain image analysis device disclosed in the present application, the controller may be configured to acquire a first feature region and a second feature region from the brain image, calculate a first feature point from the first feature region, calculate a second feature point from the second feature region, acquire a reference direction connecting the first feature point and the second feature point based on the first feature point and the second feature point, acquire a reference plane parallel to the reference direction, and modify a part of the first boundary of the first internal region based on the reference plane to acquire the second internal region having the second boundary, thereby acquiring the second internal region.

[0074] According to the brain image analysis device disclosed in the present application, the reference region may be a region corresponding to the neck bones, and the reference plane may be adjacent to the region corresponding to the neck bones and perpendicular to the sagittal plane.

[0075] According to the brain image analysis device disclosed in the present application, the controller may be configured to modify a portion of the first boundary located below the reference plane based on the first internal region located on the upper side of the reference plane, and to obtain the second internal region having the second boundary based on the modified first boundary.

[0076] According to the brain image analysis device disclosed in the present application, the controller may be configured to define the volume index as a ratio of the first volume value to the second volume value.

[0077] According to the brain image analysis device disclosed in the present application, the controller may be configured to: acquire a first correction parameter for correcting the first volume value, the first correction parameter being acquired based on scanning conditions under which the brain image is acquired or the position of the target element; calculate the first volume correction value based on the first correction parameter and the first volume value to obtain the first volume correction value; acquire a second correction parameter for correcting the second volume value, the second correction parameter being acquired based on scanning conditions under which the brain image is acquired or the position of the target element; calculate the second volume correction value based on the second correction parameter and the second volume value to obtain the second volume correction value; and calculate the volume index of the target element by calculating the volume index based on the first volume correction value and the second volume correction value.

[0078] According to the brain image analysis device disclosed in the present application, the scanning conditions under which the brain image is acquired may be related to at least one of the magnetic field strength related to the resolution of the brain image of the brain image acquisition device, the manufacturer of the brain image acquisition device, and setting parameters related to the form of the magnetic field generated by the brain image acquisition device.

[0079] The medical image analysis method disclosed in the present application may include: acquiring correction parameters calculated based on a correlation between a first morphological value associated with a target element acquired from a first medical image acquired under first scanning conditions and a second morphological value associated with the target element acquired from a second medical image acquired under second scanning conditions; acquiring a target medical image acquired under the second scanning conditions; acquiring a target region associated with the target element from the target medical image by segmenting the target medical image into a plurality of regions corresponding to a plurality of elements including the target element; acquiring target morphological values ​​associated with the target element based on voxel data corresponding to the target region; acquiring corrected morphological values ​​based on the target morphological values ​​and the correction parameters; and outputting a morphological index based on the corrected morphological values.

[0080] According to the medical image analysis method disclosed in the present application, the correction parameters include parameters for calculating the first morphological value based on the second morphological value, and obtaining the corrected morphological value may include obtaining the corrected morphological value, which is a morphological estimate of the target element under the first scanning conditions, based on the target morphological value and the correction parameters.

[0081] According to the medical image analysis method disclosed in the present application, the correction parameters include parameters associated with a linear function for calculating the first morphological value based on the second morphological value, and obtaining the corrected morphological value may include obtaining the corrected morphological value based on the linear function including the target morphological value and the parameters.

[0082] According to the medical image analysis method disclosed in the present application, the first scan condition and the second scan condition may be related to at least one of setting parameters related to a magnetic field strength associated with a resolution of a medical image of a medical image acquisition device, a manufacturer of the medical image acquisition device, and a form of a magnetic field generated by the medical image acquisition device.

[0083] According to the medical image analysis method disclosed in the present application, the target medical image is acquired from a first object having a first characteristic, and the correction parameters are acquired from the first medical image and the second medical image acquired from a second object having the first characteristic, and the first characteristic may be related to the age or gender of the object.

[0084] According to the medical image analysis method disclosed in the present application, the segmentation may be performed using a neural network configured to obtain multiple regions corresponding to multiple elements based on the target medical image.

[0085] According to the medical image analysis method disclosed in the present application, the method may further include converting the target medical image acquired under the second scanning conditions into a medical image corresponding to the image acquired under the first scanning conditions, and the segmentation may be performed based on the converted medical image.

[0086] According to the medical image analysis method disclosed in the present application, when the first scan condition and the second scan condition are related to magnetic field strength, the correction parameters may include a parameter set for converting the second morphological value associated with a target element obtained from the second medical image acquired under the second magnetic field strength into the first morphological value associated with the target element obtained from the first medical image acquired under the first magnetic field strength, and when the target medical image is acquired under the second magnetic field strength, the corrected morphological value may be obtained based on the target morphological value and the parameter set.

[0087] According to the medical image analysis device disclosed in the present application, the medical image analysis device may include an image acquisition unit that acquires a target medical image; and a controller that provides medical image analysis information based on the target medical image. The controller may be configured to acquire correction parameters calculated based on a correlation between a first morphological value associated with a target element acquired from a first medical image acquired under first scanning conditions and a second morphological value associated with the target element acquired from a second medical image acquired under second scanning conditions, acquire the target medical image acquired under the second scanning conditions, acquire a target region associated with the target element from the target medical image by segmenting the target medical image into a plurality of regions corresponding to a plurality of elements including the target element, acquire target morphological values ​​associated with the target element based on voxel data corresponding to the target region, acquire corrected morphological values ​​based on the target morphological values ​​and the correction parameters, and output morphological indices based on the corrected morphological values.

[0088] According to the medical image analysis device disclosed in the present application, the correction parameters include parameters for calculating the first morphological value based on the second morphological value, and the controller can obtain the corrected morphological value, which is a morphological estimate of the target element under the first scanning conditions, based on the target morphological value and the correction parameters.

[0089] According to the medical image analysis device disclosed in the present application, the correction parameters may include parameters associated with a linear function for calculating the first morphological value based on the second morphological value, and the controller may be configured to obtain the corrected morphological value based on the linear function including the target morphological value and the parameters.

[0090] According to the medical image analysis device disclosed in the present application, the first scan condition and the second scan condition may be related to at least one of setting parameters related to a magnetic field strength associated with the resolution of a medical image of a medical image acquisition device, a manufacturer of the medical image acquisition device, and a form of a magnetic field generated by the medical image acquisition device.

[0091] According to the medical image analysis device disclosed in the present application, the target medical image is acquired from a first object having a first characteristic, and the correction parameters are acquired from the first medical image and the second medical image acquired from a second object having the first characteristic, and the first characteristic may be related to the age or gender of the object.

[0092] According to the medical image analysis apparatus disclosed in the present application, the segmentation may be performed using a neural network configured to obtain multiple regions corresponding to multiple elements based on the target medical image.

[0093] According to the medical image analysis device disclosed in the present application, the controller may be configured to convert the target medical image acquired under the second scanning conditions into a medical image corresponding to the image acquired under the first scanning conditions, and the segmentation may be performed based on the converted medical image.

[0094] The medical image analysis device disclosed in the present application may further include an output module that outputs morphological information acquired based on the morphological indicators and a morphological indicator database, wherein the morphological information includes percentile information for the morphological indicator database of the morphological indicators associated with the object of the target medical image based on the morphological indicators and the morphological indicator database, and the output module may be configured to output the morphological information reflecting the percentile information.

[0095] According to the medical image analysis device disclosed in the present application, when the first scan condition and the second scan condition are associated with magnetic field strength, the correction parameters include a parameter set for converting the second morphological value associated with the target element obtained from the second medical image acquired under the second magnetic field strength into the first morphological value associated with the target element obtained from the first medical image acquired under the first magnetic field strength, and when the target medical image is acquired under the second magnetic field strength, the corrected morphological value may be obtained based on the target morphological value and the parameter set.

[0096] According to the medical image analysis method disclosed in the present application, the method includes: acquiring a target medical image; acquiring target scan conditions related to the target medical image; acquiring target morphological values ​​associated with the target element based on voxel data of a target area corresponding to a target element included in the target medical image; determining a target correction parameter from among one or more correction parameters based on the target scan conditions; and outputting a morphological index based on the determined target correction parameter and the target morphological value. The determining the target correction parameter may include, if the target scan condition corresponds to the first scan condition, determining the target correction parameter as a first correction parameter for correcting a first morphological value associated with the target element acquired under the first scan condition; and, if the target scan condition corresponds to the second scan condition, determining the target correction parameter as a second correction parameter for correcting a second morphological value associated with the target element acquired under a second scan condition different from the first scan condition.

[0097] According to the medical image analysis method disclosed in the present application, outputting the morphological index may include: calculating the first corrected morphological index based on the target morphological value and the first correction parameter when the target scan condition corresponds to the first scan condition; calculating the second corrected morphological index different from the first corrected morphological index based on the target morphological value and the second correction parameter when the target scan condition corresponds to the second scan condition; and outputting the first corrected morphological index when the target scan condition corresponds to the first scan condition; and outputting the second corrected morphological index when the target scan condition corresponds to the second scan condition.

[0098] The medical image analysis method disclosed in the present application may further include obtaining a reference scan condition that serves as a reference for determining the target correction parameters.

[0099] According to the medical image analysis method disclosed in the present application, the first correction parameters include parameters for calculating morphological values ​​acquired under the first scanning conditions into morphological estimates under third scanning conditions, the second correction parameters include parameters for calculating morphological values ​​acquired under the second scanning conditions into morphological estimates under the third scanning conditions, calculating the first corrected morphological index includes obtaining the first corrected morphological value, which is a morphological estimate of the target element under the second scanning condition, based on the target morphological value and the first correction parameters, and calculating the second corrected morphological index includes obtaining the second corrected morphological value, which is a morphological estimate of the target element under the third scanning condition, based on the target morphological value and the second correction parameters.

[0100] According to the medical image analysis method disclosed in the present application, the first correction parameters include a first parameter set associated with a first linear function for calculating morphological values ​​acquired under the first scanning condition into morphological estimates under a third scanning condition, the second correction parameters include a second parameter set associated with a second linear function for calculating morphological values ​​acquired under the second scanning condition into morphological estimates under the third scanning condition, calculating the first corrected morphological index includes obtaining the first corrected morphological value based on the first linear function including the target morphological value and the first parameter set, and calculating the second corrected morphological index includes obtaining the second corrected morphological value based on the second linear function including the target morphological value and the second parameter set.

[0101] According to the medical image analysis method disclosed in the present application, the first scan condition, the second scan condition, and the target scan condition may be associated with at least one of setting parameters related to a magnetic field strength associated with a resolution of a medical image of a medical image acquisition device, a manufacturer of the medical image acquisition device, and a form of a magnetic field generated by the medical image acquisition device.

[0102] According to the medical image analysis method disclosed in the present application, the target medical image is acquired from a subject having a first characteristic, and the first correction parameter or the second correction parameter is acquired from the first medical image and the second medical image acquired from the subject having the first characteristic, and the first characteristic may be related to the age or gender of the subject.

[0103] According to the medical image analysis method disclosed in the present application, obtaining the target morphological value includes obtaining a target region associated with the target element from the target medical image by performing segmentation of the target medical image into a plurality of regions corresponding to a plurality of elements including at least the target element, and obtaining the target morphological value based on voxel data corresponding to the target region, and the segmentation may be performed using a neural network provided to obtain a plurality of regions corresponding to a plurality of elements based on the target medical image.

[0104] According to the medical image analysis method disclosed in the present application, the method may further include converting the target medical image acquired under the target scanning conditions into a medical image corresponding to an image taken under a scanning condition other than the target scanning conditions, and the segmentation may be performed based on the converted medical image.

[0105] According to the medical image analysis method disclosed in the present application, when the first scan condition, the second scan condition, and the target scan condition are associated with magnetic field strengths, the first correction parameters include a first parameter set for correcting the first morphological value associated with the target element acquired under a first magnetic field strength, and the second correction parameters include a second parameter set for correcting the second morphological value associated with the target element acquired under a second magnetic field strength, and determining the target correction parameters may include determining the target correction parameters to the first parameter set when the target scan condition corresponds to the first magnetic field strength, and determining the target correction parameters to the second parameter set when the target scan condition corresponds to the second magnetic field strength.

[0106] According to the medical image analysis device disclosed in the present application, the medical image analysis device includes an image acquisition unit that acquires a target medical image; and a controller that provides medical image analysis information based on the target medical image. The controller is configured to acquire target scan conditions related to the target medical image, acquire target morphological values ​​associated with the target element based on voxel data of a target area corresponding to the target element included in the target medical image, determine a target correction parameter from among one or more correction parameters based on the target scan conditions, and output a morphological index based on the determined target correction parameter and the target morphological value. If the target scan conditions correspond to the first scan conditions, the controller may be configured to determine the target correction parameter as a first correction parameter for correcting a first morphological value associated with the target element acquired under the first scan conditions, and if the target scan conditions correspond to the second scan conditions, the controller may be configured to determine the target correction parameter as a second correction parameter for correcting a second morphological value associated with the target element acquired under second scan conditions different from the first scan conditions.

[0107] According to the medical image analysis device disclosed in the present application, the controller may be configured to calculate the first corrected morphological index based on the target morphological value and the first correction parameter when the target scan condition corresponds to the first scan condition, calculate the second corrected morphological index different from the first corrected morphological index based on the target morphological value and the second correction parameter when the target scan condition corresponds to the second scan condition, and output the first corrected morphological index when the target scan condition corresponds to the first scan condition, and output the second corrected morphological index when the target scan condition corresponds to the second scan condition, thereby outputting the morphological index.

[0108] According to the medical image analysis device disclosed in the present application, the first correction parameters may include parameters for calculating morphological values ​​acquired under the first scanning conditions into morphological estimates under third scanning conditions, and the second correction parameters may include parameters for calculating morphological values ​​acquired under the second scanning conditions into morphological estimates under the third scanning conditions. The controller may be configured to obtain the first corrected morphological values, which are morphological estimates of the target element under the second scanning conditions, based on the target morphological values ​​and the first correction parameters, to calculate the first corrected morphological indices, and to obtain the second corrected morphological values, which are morphological estimates of the target element under the third scanning conditions, based on the target morphological values ​​and the second correction parameters, to calculate the second corrected morphological indices.

[0109] According to the medical image analysis apparatus disclosed in the present application, the first correction parameters may include a first parameter set associated with a first linear function for calculating morphological values ​​acquired under the first scanning conditions into morphological estimates under a third scanning condition, and the second correction parameters may include a second parameter set associated with a second linear function for calculating morphological values ​​acquired under the second scanning conditions into morphological estimates under the third scanning condition. The controller may be configured to obtain the first corrected morphological values ​​and calculate the first corrected morphological indices based on the first linear function including the target morphological values ​​and the first parameter set, and to obtain the second corrected morphological values ​​and calculate the second corrected morphological indices based on the second linear function including the target morphological values ​​and the second parameter set.

[0110] According to the medical image analysis device disclosed in the present application, the first scan condition and the second scan condition may be related to at least one of setting parameters related to a magnetic field strength associated with the resolution of a medical image of a medical image acquisition device, a manufacturer of the medical image acquisition device, and a form of a magnetic field generated by the medical image acquisition device.

[0111] According to the medical image analysis device disclosed in the present application, the target medical image is acquired from a subject having a first characteristic, and the target correction parameters including at least one of the first correction parameter and the second correction parameter are acquired from the first medical image and the second medical image acquired from the subject having the first characteristic, and the first characteristic may be related to the age or gender of the subject.

[0112] According to the medical image analysis device disclosed in the present application, the controller is configured to obtain a target region associated with the target element from the target medical image by performing segmentation of the target medical image into a plurality of regions corresponding to a plurality of elements including at least the target element, and to obtain the target morphological value based on voxel data corresponding to the target region, and the segmentation may be performed using a neural network provided to obtain a plurality of regions corresponding to a plurality of elements based on the target medical image.

[0113] According to the medical image analysis device disclosed in the present application, the controller may be configured to convert the target medical image acquired under the target scanning conditions into a medical image corresponding to an image captured under a scanning condition other than the target scanning conditions, and perform the segmentation based on the converted medical image.

[0114] According to the medical image analysis device disclosed in the present application, when the first scan condition, the second scan condition, and the target scan condition are associated with magnetic field strengths, the first correction parameters include a first parameter set for correcting the first morphological value associated with the target element acquired under a first magnetic field strength, and the second correction parameters include a second parameter set for correcting the second morphological value associated with the target element acquired under a second magnetic field strength, and the controller may be configured to determine the target correction parameters by determining the target correction parameters to be the first parameter set when the target scan condition corresponds to the first magnetic field strength, and determining the target correction parameters to be the second parameter set when the target scan condition corresponds to the second magnetic field strength.

[0115] The brain image analysis method disclosed in the present application may include: acquiring a first correction parameter for correcting a first morphological value associated with a first brain element; acquiring a second correction parameter for correcting a second morphological value associated with a second brain element different from the first brain element; acquiring a target brain image; segmenting the target brain image into a plurality of brain regions, including a first region corresponding to the first brain element, a second region corresponding to the second brain element, and a skull region, to acquire the first region, the second region, and a third region associated with an internal region of the skull region; acquiring a first brain morphological index associated with the first brain element based on voxel data corresponding to the first region, voxel data corresponding to the third region, and the first correction parameter; and acquiring a second brain morphological index associated with the second brain element based on voxel data corresponding to the second region, voxel data corresponding to the third region, and the second correction parameter.

[0116] The brain image analysis method disclosed in the present application further includes obtaining a reference morphological value related to the internal region based on voxel data corresponding to the third region, wherein obtaining the first brain morphological index includes obtaining a first target morphological value related to the first brain element based on voxel data corresponding to the first region, calculating a first corrected morphological value related to the first brain element based on the first target morphological value and the first correction parameter, and calculating the first brain morphological index based on the first corrected morphological value and the reference morphological value, and obtaining the second brain morphological index includes obtaining a second target morphological value related to the second brain element based on voxel data corresponding to the second region, calculating a second corrected morphological value related to the second brain element based on the second target morphological value and the second correction parameter, and calculating the second brain morphological index based on the second corrected morphological value and the reference morphological value.

[0117] According to the brain image analysis method disclosed in the present application, the first morphological value and the second morphological value are values ​​acquired under first scanning conditions, the target brain image is acquired under the first scanning conditions, the first correction parameters include a first parameter for calculating a third morphological value associated with the first brain element acquired under second scanning conditions different from the first scanning conditions based on the first morphological values, the second correction parameters include a second parameter for calculating a fourth morphological value associated with the second brain element acquired under the second scanning conditions based on the second morphological values, calculating the first corrected morphological value includes obtaining the first corrected morphological value, which is a morphological estimate of the first brain element under the second scanning conditions, based on the first target morphological value and the first correction parameters, and calculating the second corrected morphological value includes obtaining the second corrected morphological value, which is a morphological estimate of the second brain element under the second scanning conditions, based on the second target morphological value and the second correction parameters.

[0118] According to the brain image analysis method disclosed in the present application, the first correction parameters include a first parameter set associated with a first linear function for calculating the third morphological value based on the first morphological value, the second correction parameters include a second parameter set associated with a second linear function for calculating the fourth morphological value based on the second morphological value, calculating the first corrected morphological value includes obtaining the first corrected morphological value based on the first target morphological value and a first parameter set associated with the first linear function, and calculating the second corrected morphological value includes obtaining the second corrected morphological value based on the second target morphological value and a second parameter set associated with the second linear function.

[0119] According to the brain image analysis method disclosed in the present application, the first region corresponding to the first brain element may be located adjacent to the skull region compared to the second region corresponding to the second brain element.

[0120] According to the brain image analysis method disclosed in the present application, the first brain element may be associated with an element that performs a first brain function, and the second brain element may be associated with an element that performs a second brain function different from the first brain function.

[0121] According to the brain image analysis method disclosed in the present application, the first scan condition and the second scan condition may be related to at least one of setting parameters related to a magnetic field strength associated with the resolution of a brain image of a brain image acquisition device, a manufacturer of the brain image acquisition device, and a form of a magnetic field generated by the brain image acquisition device.

[0122] According to the brain image analysis method disclosed in the present application, the target brain image is obtained from a first subject having a first characteristic, the first correction parameter and the second correction parameter are obtained from a first brain image and a second brain image obtained from a second subject having the first characteristic, and the first characteristic may be related to the age or gender of the subject.

[0123] According to the brain image analysis method disclosed in the present application, the segmentation may be performed using a neural network provided to obtain a plurality of brain regions corresponding to a plurality of brain elements including the first brain element, the second brain element, and the skull based on the target brain image.

[0124] According to the brain image analysis method disclosed in the present application, the method may further include preprocessing the target brain image, wherein preprocessing the target brain image includes performing preprocessing on the first region using a first method and performing preprocessing on the second region using a second method different from the first method, and the segmentation may be performed based on the preprocessed target brain image.

[0125] According to the brain image analysis method disclosed in the present application, the first corrected morphological value may be a volume value associated with the first brain element acquired in the target brain image, the second target morphological value may be a volume value associated with the second brain element acquired in the target brain image, the first brain morphological index may be calculated using the first corrected morphological value relative to the reference morphological value, and the second brain morphological index may be calculated using the second corrected morphological value relative to the reference morphological value.

[0126] According to the brain image analysis device disclosed in the present application, the device may include an image acquisition unit that acquires a target brain image; and a controller that provides brain image analysis information based on the target brain image, wherein the controller may be configured to acquire a first correction parameter for correcting a first morphological value associated with a first brain element, acquire a second correction parameter for correcting a second morphological value associated with a second brain element different from the first brain element, acquire a target brain image, and perform segmentation of the target brain image into a plurality of brain regions including a first region corresponding to the first brain element, a second region corresponding to the second brain element, and a skull region, thereby acquiring the first region, the second region, and a third region associated with an internal region of the skull region, acquire a first brain morphological index associated with the first brain element based on voxel data corresponding to the first region, voxel data corresponding to the third region, and the first correction parameter, and acquire a second brain morphological index associated with the second brain element based on voxel data corresponding to the second region, voxel data corresponding to the third region, and the second correction parameter.

[0127] According to the brain image analysis device disclosed in the present application, the controller may be configured to obtain a reference morphological value related to the internal region based on voxel data corresponding to the third region; obtain a first target morphological value related to the first brain element based on voxel data corresponding to the first region; calculate a first corrected morphological value related to the first brain element based on the first target morphological value and the first correction parameter; calculate the first brain morphological index based on the first corrected morphological value and the reference morphological value to obtain the first brain morphological index; obtain a second target morphological value related to the second brain element based on voxel data corresponding to the second region; calculate a second corrected morphological value related to the second brain element based on the second target morphological value and the second correction parameter; and calculate the second brain morphological index based on the second corrected morphological value and the reference morphological value to obtain the second brain morphological index.

[0128] According to the brain image analysis device disclosed in the present application, the first morphological value and the second morphological value are values ​​acquired under first scanning conditions, the target brain image is acquired under the first scanning conditions, the first correction parameters include a first parameter for calculating a third morphological value associated with the first brain element acquired under second scanning conditions different from the first scanning conditions based on the first morphological values, and the second correction parameters include a second parameter for calculating a fourth morphological value associated with the second brain element acquired under the second scanning conditions based on the second morphological values, and the controller may be configured to calculate the first corrected morphological value, which is a morphological estimate of the first brain element under the second scanning conditions, based on the first target morphological value and the first correction parameters, and to calculate the second corrected morphological value, which is a morphological estimate of the second brain element under the second scanning conditions, based on the second target morphological value and the second correction parameters.

[0129] According to the brain image analysis device disclosed in the present application, the first correction parameters may include a first parameter set associated with a first linear function for calculating the third morphological value based on the first morphological value, the second correction parameters may include a second parameter set associated with a second linear function for calculating the fourth morphological value based on the second morphological value, calculating the first corrected morphological value may include obtaining the first corrected morphological value based on the first target morphological value and a first parameter set associated with the first linear function, and the controller may be configured to calculate the first corrected morphological value based on the first target morphological value and a first parameter set associated with the first linear function, and calculate the second corrected morphological value based on the second target morphological value and a second parameter set associated with the second linear function.

[0130] According to the brain image analysis device disclosed in the present application, the first region corresponding to the first brain element can be located adjacent to the skull region compared to the second region corresponding to the second brain element.

[0131] According to the brain image analysis device disclosed in the present application, the first brain element can be associated with an element that performs a first brain function, and the second brain element can be associated with an element that performs a second brain function different from the first brain function.

[0132] According to the brain image analysis device disclosed in the present application, the first scan condition and the second scan condition may be related to at least one of setting parameters related to the magnetic field strength associated with the resolution of the brain image of the brain image acquisition device, the manufacturer of the brain image acquisition device, and the form of the magnetic field generated by the brain image acquisition device.

[0133] According to the brain image analysis device disclosed in the present application, the controller determines that the target brain image is acquired from a first subject having a first characteristic, and the first correction parameter and the second correction parameter are acquired from a first brain image and a second brain image acquired from a second subject having the first characteristic, and the first characteristic may be related to the age or gender of the subject.

[0134] According to the brain image analysis device disclosed in the present application, the segmentation may be performed using a neural network provided to obtain a plurality of brain regions corresponding to a plurality of brain elements including the first brain element, the second brain element, and the skull based on the target brain image.

[0135] According to the brain image analysis device disclosed in the present application, the controller may be configured to preprocess the target brain image, and the segmentation may be performed based on the preprocessed target brain image, to perform preprocessing on the first region using a first method, and to perform preprocessing on the second region using a method different from the first method.

[0136] According to the brain image analysis device disclosed in the present application, the first corrected morphological value may be a volume value associated with the first brain element acquired in the target brain image, the second target morphological value may be a volume value associated with the second brain element acquired in the target brain image, the first brain morphological index may be calculated using the first corrected morphological value relative to the reference morphological value, and the second brain morphological index may be calculated using the second corrected morphological value relative to the reference morphological value.

[0137] The above-mentioned objects, features, and advantages of the present application will become more apparent through the following detailed description taken in conjunction with the accompanying drawings. However, the present application can be modified in various ways and can have various embodiments, and the present invention will be described in detail below by illustrating specific embodiments in the drawings.

[0138] The same reference numerals are used throughout the specification to indicate essentially the same elements. Furthermore, elements having the same function within the same concept shown in the drawings of each embodiment will be described using the same reference numerals, and redundant description thereof will be omitted.

[0139] If a detailed description of a known function or configuration related to this application is deemed unnecessary for the gist of this application, the detailed description will be omitted. In addition, numbers (e.g., 1, 2, etc.) used in the description of this specification are merely identification symbols for distinguishing one component from another.

[0140] Furthermore, the suffixes "module" and "section" for components used in the following examples are given or used interchangeably in consideration of making the specification easier to write, and do not have any meanings or roles that are distinct from each other.

[0141] In the following examples, the singular expression also includes the plural expression unless the context clearly indicates otherwise.

[0142] In the following examples, terms such as "include" or "have" mean that the features or components described in the specification are present, and do not preclude the possibility that one or more other features or components may also be added.

[0143] In the drawings, the dimensions of components may be expanded or reduced for ease of explanation. For example, the size and thickness of each component shown in the drawings are arbitrarily shown for the convenience of explanation, and the present invention is not necessarily limited to those shown in the drawings.

[0144] If an embodiment can be alternatively implemented, the order of certain processes may be performed differently than described. For example, two processes described in succession may be performed substantially simultaneously or in the reverse order from that described.

[0145] In the following examples, when elements are connected, it includes not only the case where the elements are directly connected, but also the case where the elements are indirectly connected with an intervening element.

[0146] For example, in this specification, when components are electrically connected, this includes not only cases where the components are directly electrically connected, but also cases where the components are indirectly electrically connected with other components interposed therebetween.

[0147] Hereinafter, a brain image analysis device, a brain image analysis system, and a brain image analysis method according to the present application will be described. Here, the brain image analysis may be performed by performing segmentation on a brain image to obtain a plurality of brain regions, calculating morphological indices associated with the brain, and outputting the morphological indices to a user.

[0148] In this application, various techniques are disclosed to improve the accuracy and reliability of such brain image analysis, such as checking the quality of the brain image to improve the reliability of the brain image analysis results, correcting the morphological values ​​and morphological indices of specific brain regions obtained to improve the accuracy of the brain image analysis results, calculating the morphological values ​​and morphological indices of specific brain regions from the brain image being analyzed to provide personalized analysis results, and providing only selected information to the user to provide a user-friendly report.

[0149] A medical image analysis method, a medical image analysis apparatus, and a medical image analysis system according to an embodiment of the present application will be described below.

[0150] 1 is a schematic diagram of a medical image analysis system according to an embodiment of the present application. Referring to FIG. 1, the medical image analysis system according to an embodiment of the present application may include an image acquisition device 1000 and an image analysis device 2000.

[0151] The image acquisition device 1000 can acquire and transmit images to the image analysis device 2000 over a network.

[0152] For example, the image acquisition device 1000 may be a device for acquiring a magnetic resonance image (MRI). At this time, the magnetic resonance image acquired by the image acquisition device 1000 may be transmitted to the image analysis device 2000 via a network.

[0153] As another example, the image acquisition device 1000 may be a device for acquiring a computed tomography (CT) image. In this case, the computed tomography image acquired by the image acquisition device 1000 may be transmitted to the image analysis device 2000 via a network.

[0154] As another example, the image acquisition device 1000 may be a device for acquiring an image acquired by radiography. In this case, the radiography image acquired by the image acquisition device 1000 may be transmitted to the image analysis device 2000 via a network.

[0155] As another example, the image acquisition device 1000 may be a device for acquiring images acquired by positron emission tomography (PET).

[0156] However, the image capture device 1000 described above is by way of example only and is not intended to be limiting and should be construed to include any suitable device or system utilized in medical imaging.

[0157] The image acquired by the image acquisition device 1000 may be a two-dimensional image, and may include pixel information related to pixel coordinates, color, intensity, etc.

[0158] The image acquired by the image acquisition device 1000 may be a three-dimensional image, which may include pixel information related to the coordinates, color, and intensity of voxels.

[0159] The image acquired by the image acquisition device 1000 may include information related to image alignment. For example, the image acquisition device 1000 may acquire data (ijk) related to the orientation of the captured image in consideration of the orientation of the reference coordinate axis (RAS) of the object 100. Specifically, the image acquisition device may acquire data (ijk) related to the orientation of the captured image in consideration of information (xyz) with respect to the coordinate axes of the image acquisition device and information (RAS) with respect to the reference coordinate axis of the object 100.

[0160] The image acquisition device 1000 can acquire data related to the magnetic field strength of the image, data related to the manufacturer of the image device, data related to the setting parameters of the image device, data related to the object of the image, etc.

[0161] Data related to magnetic field strength may include data related to the strength of a magnetic field applied to a subject when acquiring a medical image using a magnetic field (e.g., MRI). For example, if a target image is acquired under a magnetic field strength condition of 1.5T (Tesla), data corresponding to "1.5T" may be data related to magnetic field strength. However, data corresponding to magnetic field strengths commonly used to acquire medical images, such as 3T, 7T, and 8T, may also be data related to magnetic field strength, without being limited to 1.5T.

[0162] The imaging device setting parameters may include parameters that can be adjusted or controlled by the imaging device to acquire medical images. For example, the imaging device setting parameters may include a repetition time (TR) that refers to the time between successive pulse sequences applied to the same slice, a time to echo (TE) between transmission of an RF pulse and reception of an echo signal, a time constant (T1) associated with the rate at which excited protons return to equilibrium, a time constant (T2) associated with the rate at which excited protons reach equilibrium or fall out of phase with each other, proton density (PD), or any combination thereof.

[0163] However, the term should not be limited to the above examples and should be interpreted to include any parameters related to magnetic field characteristics, and should not be limited to obtaining an image of a subject using a magnetic field, but should be interpreted to include any parameters related to an imaging device using a CT method or a radiation (e.g., X-ray) investigation method.

[0164] At this time, the above-mentioned data can be processed as metadata within the acquired image and sent to the image analysis device 2000, or can be transmitted to the image analysis device 2000 separately from the image.

[0165] The image acquisition device 1000 can be a device for acquiring medical images such as MRI, CT, and X-ray.

[0166] The image acquisition device 1000 is capable of acquiring images under a variety of scanning conditions.

[0167] For example, if the imaging device is an MRI, an image can be acquired under scanning conditions with a magnetic field strength of 1.5 Tesla (hereinafter, T), or under scanning conditions with a magnetic field strength of 3T, or under scanning conditions with a magnetic field strength of 7T or 8T.

[0168] In another example, if the image acquisition device 1000 is an MRI, as described above, images can be acquired under scan conditions with set parameters consisting of the TR (Repetition Time), TE (Time to Echo), proton density (PD) of the image acquisition device 1000, or a time constant (T1) associated with the rate at which excited protons return to equilibrium, a time constant (T2) associated with the rate at which excited protons reach equilibrium or fall out of phase with each other, or any combination thereof.

[0169] In another example, the image acquisition device 1000 may be a device related to CT, and in this case, an image may be acquired under scan conditions of setting parameters related to voltage, current, exposure time, scanning time, projection, and any combination thereof that can be set in the CT device.

[0170] In another example, the image acquisition device 1000 may be an X-ray related device, and an image may be acquired under scan conditions of setting parameters related to tube voltage, tube current, imaging time, distance between the X-ray tube device and the detector (SID), angle of the X-ray tube support device, and any combination thereof that can be set in the X-ray image device.

[0171] In addition, images may be acquired differently depending on the propensity of the applicator who irradiates the radiation. In this case, information regarding the applicator's propensity may be parameterized. Furthermore, the parameter regarding the applicator's propensity may be used as a consideration in relation to correcting morphological values ​​and morphological indices in consideration of the scan conditions described below. In other words, the morphological values ​​and morphological indices may be corrected according to the applicator's propensity. More specifically, the image analysis device 2000 according to an embodiment may acquire identification information of the applicator in relation to the scan conditions. Furthermore, the image analysis device 2000 may be implemented to acquire correction parameters for correcting morphological values ​​and morphological indices based on the identification information of the applicator. For example, the image analysis device 2000 may acquire correction parameters for estimating morphological values ​​calculated from a target image acquired from a first practitioner to morphological values ​​associated with a second practitioner (or correction parameters for estimating morphological values ​​calculated from a target image acquired from a second practitioner to morphological values ​​associated with a first practitioner) based on a correlation between a first morphological value calculated from a first image taken by a first practitioner and a second morphological value calculated from a second image taken by a second practitioner.

[0172] In another example, the image acquisition device 1000 may be an MRI device or CT device using positron emission tomography (PET). When using PET imaging, a chemical related to a radiopharmaceutical may be used. For example, to measure amyloid beta, a chemical (or tracer) such as 18F-florbetapir, 18F-florbetaben, 18F-flutemetamol, or 18F-florapronol may be used. To measure tau protein, a chemical (or tracer) such as 18F-flortaucipir may be used. The chemicals used may vary depending on the manufacturer of the image acquisition device. In this case, information about the chemicals used may be parameterized. For the image acquisition device, the parameters for the chemicals used may be used as scan conditions and may be used as a consideration factor in correcting morphological values ​​or morphological indices taking into account the scan conditions described below. In other words, the morphological values ​​and morphological indices can be corrected based on the manufacturer of the image acquisition device and the chemicals used. More specifically, the image analysis device 2000 can acquire information on the manufacturer of the image acquisition device and the chemicals used, and can be implemented to acquire correction parameters for correcting the morphological values ​​and morphological indices based on the information on the manufacturer of the image acquisition device and the chemicals used.For example, the image analysis device 2000 can obtain a first correction parameter for estimating a morphological value calculated from a target image acquired using a first drug to a morphological value acquired using a second drug (or a correction parameter for estimating a morphological value calculated from a target image acquired using a second drug to a morphological value acquired using the first drug) based on a correlation between a first morphological value calculated from a first image taken using a first drug (e.g., 18F-florbetapir) and a second morphological value calculated from a second image taken using a second drug (e.g., 18F-florbetaben). Furthermore, the image analysis device 2000 may acquire a second correction parameter for extrapolating the morphological values ​​calculated from the target image acquired using the second chemical to morphological values ​​acquired using a third chemical (e.g., 18F-flutemetamol) based on the correlation between the second morphological value calculated from the second image acquired using the second chemical and the third morphological value calculated from the third image acquired using the third chemical (or a correction parameter for extrapolating the morphological values ​​calculated from the target image acquired using the third chemical to morphological values ​​acquired using the second chemical). In this case, the first correction parameter may be different from the second correction parameter.

[0173] Furthermore, the image acquisition device 1000 may acquire images under scan conditions including at least one setting parameter, i.e., images may be acquired under scan conditions based on various combinations of the setting parameters described above.

[0174] For example, an image may be acquired under a first scan condition with a first combination of setting parameters, or under a second scan condition with a second combination of setting parameters and a first magnetic field strength, etc. Images acquired by the image acquisition device 1000 may be images acquired under various scan conditions with various combinations of setting parameters.

[0175] Furthermore, the image acquisition device 1000 may be configured with a plurality of image acquisition devices 1000 .

[0176] In this case, the multiple image acquisition devices 1000 may be devices manufactured by different manufacturers. Images acquired by devices manufactured by different manufacturers may have different characteristics, such as brightness and intensity, even if the scanning conditions and setting parameters are the same. Therefore, even if medical images are acquired of the same subject, morphological indices based on the medical images may differ depending on the manufacturer's device.

[0177] Therefore, a correction parameter acquisition operation is required to control the scan conditions and setting parameters or the variables of the imaging device manufacturer using a correction parameter acquisition device 2400 according to an embodiment of the present application, which will be described later.

[0178] The images acquired by the image acquisition device 1000 may include information related to the anatomical structure of a particular body part. Furthermore, the particular body part may correspond to any body part for which medical imaging can be utilized. For convenience of explanation, the drawings and specification below will focus on images related to the brain, but this is for illustrative purposes only, and various embodiments disclosed herein may also be applied to any suitable body part other than the brain (e.g., lungs, breasts, heart, joints, blood vessels, etc.).

[0179] Meanwhile, the image acquisition device 1000 according to an embodiment of the present application may be implemented in the form of a server. In this case, the server may be configured to store medical images and information related to the medical images. The server may also be configured to modify or process the medical images and information related to the medical images.

[0180] Additionally, the medical images may be stored in the memory of the image analyzer 2000 or on a server and used to calculate correction parameters, perform quality control, and output analysis results, as will be described in more detail below.

[0181] Hereinafter, an image analysis device 2000, a learning device 2200, a correction parameter acquisition device 2400, and an output device 2600 for image analysis according to an embodiment of the present application will be described with reference to Fig. 2. Fig. 2 is a schematic diagram of the image analysis device 2000, the learning device 2200, the correction parameter acquisition device 2400, and the output device 2600 for medical image analysis according to an embodiment of the present application.

[0182] The image analysis device 2000 can perform operations such as segmenting an image acquired from the image acquisition device 1000 using an artificial neural network trained by the learning device 2200, and calculating morphological indices of target elements contained in the image.

[0183] The learning device 2200 may perform operations using multiple image datasets to train a neural network model for image segmentation, or a neural network model for image quality control, or the like.

[0184] The correction parameter acquisition unit 2400 can perform an operation of calculating correction parameters for correcting morphological values ​​or morphological indices associated with target elements of a target image (or object image) acquired by the image analysis unit 2000 using multiple image data sets and data regarding scanning conditions.

[0185] The image analysis device 2000, the learning device 2200, and the correction parameter acquisition device 2400 shown in FIG. 2 may be implemented to transmit and receive data to and from each other using any communication method.

[0186] For example, the image analysis device 2000, the learning device 2200, and the correction parameter acquisition device 2400 may be implemented to share a server.

[0187] 2, the image analysis device 2000, the learning device 2200, and the correction parameter acquisition device 2400 are shown as separate devices, but this is merely an example, and the image analysis device 2000, the learning device 2200, and / or the correction parameter acquisition device 2400 could be implemented in a single device. Alternatively, some of the image analysis device 2000, the learning device 2200, and the correction parameter acquisition device 2400 may be provided as separate devices, and the remaining devices may be implemented in a single device.

[0188] The configuration of the image analysis device 2000 according to one embodiment of the present application will be described below with reference to Fig. 3. Fig. 3 is a block diagram of the image analysis device 2000 according to one embodiment of the present application.

[0189] The image analysis device 2000 according to an embodiment of the present application may include a first communication module 2010 , a first memory 2020 , and a first controller 2030 .

[0190] The first communication module 2010 can communicate with any external device including the image acquisition device 1000, the learning device 2200, the correction parameter acquisition device 2400, and the output device 2600 described below. In other words, the image analysis device 2000 can transmit and receive images from the image acquisition device 1000 via the first communication module 2010, or transmit and receive data with external devices including the learning device 2200, the parameter acquisition device 2400, and the output device 2600 described below, a repeater, a server, etc.

[0191] For example, the image analysis device 2000 may receive, via the first communication module 2010, an acquired image from the image acquisition device 1000, information about a trained neural network model from the learning device 2200, and information related to calculated correction parameters from the correction parameter acquisition device 2400. In another example, the image analysis device 2000 may transmit information related to image scanning conditions to the correction parameter acquisition device 2400 or information related to analysis results to the output device 2600 via the first communication module 2010. In another example, the image analysis device 2000 may connect to the Internet via the first communication module 2010 and upload various data related to images, information related to scanning conditions, and information related to analysis results.

[0192] The first communication module 2010 is roughly divided into a wired type and a wireless type. Since the wired type and the wireless type each have advantages and disadvantages, in some cases, the image analysis device 2000 may be provided with both a wired type and a wireless type.

[0193] Here, in the case of a wired type, LAN (Local Area Network) or USB (Universal Serial Bus) communication is a typical example, but other methods are also possible.

[0194] In addition, in the case of a wireless type, a communication method of the WPAN (Wireless Personal Area Network) series such as Bluetooth or Zigbee can be mainly used. However, the wireless communication protocol is not limited thereto, and the wireless type communication module can also use a communication method of the WLAN (Wireless Local Area Network) series such as Wi-Fi or other known communication methods.

[0195] The first memory 2020 can store various types of information. Various types of data can be temporarily or semi-permanently stored in the first memory 2020. Examples of the first memory 2020 include a hard disk drive (HDD), a solid state drive (SSD), a flash memory, a read-only memory (ROM), and a random access memory (RAM).

[0196] The first memory 2020 may be provided in a form built into the image analysis device 2000 or in a form that is detachable. The first memory 2020 may store various data necessary for the operation of the image analysis device 2000, including an operating system (OS) for driving the image analysis device 2000 and programs for operating each component of the image analysis device 2000. For example, the first memory 2020 may store various data related to images, information related to scan conditions, and information related to analysis results.

[0197] The first controller 2030 may control the overall operation of the image analysis device 2000. For example, the first controller 2030 may load a program for the operation of the image analysis device 2000 from the first memory 2020 and execute the program.

[0198] The first controller 2030 may be realized by hardware, software, or a combination thereof, such as a CPU (Central Processing Unit) or a similar device. In terms of hardware, it may be provided in the form of an electronic circuit that processes electrical signals and performs control functions, and in terms of software, it may be provided in the form of a program or code that drives a hardware circuit.

[0199] Meanwhile, referring back to FIG. 3, the image analysis device 2000 according to an embodiment of the present application may include an input module 2040 and an output module 2050 .

[0200] At this time, the image analysis device 2000 can acquire user inputs and output information corresponding to the user inputs using the input module 2040 and the output module 2050. For example, the image analysis device 2000 can acquire user inputs requesting data acquisition, user inputs instructing preprocessing, user inputs related to image segmentation, and user inputs regarding reference scan conditions related to calculation of morphological indices using the input module 2040, and output corresponding information via the output module 2050.

[0201] For example, a user can input conditions and settings related to analysis of the image analysis device 2000 via the input module 2040 .

[0202] For example, a user can set reference scan conditions associated with correction parameters for correcting morphological values ​​or morphological indices acquired from a target image via the input module 2040. In this case, the image analysis device 2000 can perform correction of the morphological values ​​or morphological indices based on the reference scan conditions received from the input module 2040.

[0203] The input module 2040 can be implemented in various forms, such as a mouse, a keyboard, a touchpad, and the like.

[0204] The output module 2050 may be provided to output information on the image analysis operation of the image analysis device 2000, image analysis results, and the like.

[0205] For example, when the image analysis device 2000 performs an operation to check the quality of an image, a notification window can be provided via the output module 2050 that artifacts are present in the target image.

[0206] According to another example, when the image analysis device 2000 performs an operation to check the quality of an image, a notification window can be provided via the output module 2050 to select whether or not serious artifacts exist in the target image and to analyze the image.

[0207] Also, in another example, if the image analysis device 2000 performs a segmentation operation on the target image, the results of the segmentation may be provided via the output module 2050 .

[0208] In another example, when the image analysis device 2000 completes the analysis of the target image, the analysis results of the target image can be provided via the output module 2050 .

[0209] The output module 2050 can be implemented in various forms, such as a display.

[0210] The image analysis device 2000 may further include a user interface for receiving user input via an input module 2040 and outputting information corresponding to the user input via an output module 2050 .

[0211] Although FIG. 3 illustrates an image analysis device 2000 according to one embodiment of the present application including an input module 2040 and an output module 2050, this is merely an example, and an image analysis device 2000 can be provided in which the input module 2040 and the output module 2050 are omitted.

[0212] Meanwhile, the image analysis device 2000 according to an embodiment of the present application may be implemented in the form of a server. In this case, the server may be configured to store the medical images and information related to the medical images transmitted from the image acquisition device 1000. The server may also be implemented to modify or process the medical images and information related to the medical images transmitted from the image acquisition device 1000.

[0213] Furthermore, the server of the image analysis device 2000 may be implemented separately from the server of the image acquisition device 1000, but is not limited thereto, and the server of the image acquisition device 1000 and the server of the image analysis device 2000 may be implemented in a single form. In other words, the image acquisition device 1000 and the image analysis device 2000 may be implemented in a form having a common server.

[0214] Meanwhile, referring again to FIG. 2, the image analysis device 2000 according to one embodiment of the present application may be implemented to communicate with an output device 2600 .

[0215] The output device 2600 according to an embodiment of the present application may be implemented to receive the analysis results of the image analysis device 2000 and output the image analysis results to a user in a visual form.

[0216] In one embodiment, the output device 2600 according to one embodiment of the present application may be configured to receive information related to image quality from the image analysis device 2000 and output the information to a user. For example, the output device 2600 may be configured to output an index related to image quality or a comment related to the reliability of the image analysis result related to the image quality.

[0217] In one embodiment, the output device 2600 according to one embodiment of the present application may be implemented to receive analysis results associated with morphological indices from the image analysis device 2000 and output them to a user.

[0218] For example, the output device 2600 may be configured to output to a user the results of comparative analysis of the analysis results associated with the morphological indices of the subject with statistical data of a comparison subject population. In this case, the results of comparative analysis of the statistical data of the comparison subject population may be processed into statistical information in any suitable form, such as a graph, and output. The output device 2600 may also be configured to provide analysis results associated with the morphological indices of multiple parts of the subject together.

[0219] According to another example, the output device 2600 may be provided to output analysis results associated with morphological indicators of a subject, and to output information related to where the body parts associated with the morphological indicators correspond in a visual form.

[0220] In one embodiment, the output device 2600 according to one embodiment of the present application may be configured to receive an image segmentation result from the image analysis device 2000 and output the result to a user. For example, the segmentation result of the image analysis device 2000 may be a labeling result within the image. In this case, the image analysis device 2000 or the output device 2600 may visually process the image according to the anatomical structure of the body based on the labeling result, and the output device 2600 may be configured to output an image in which the anatomical structure of the body is visually segmented to a user.

[0221] In one embodiment, the output device 2600 according to one embodiment of the present application may be implemented to receive reports from the image analysis device 2000 and output them to a user.

[0222] The configuration and operation of an output device 2600 according to an embodiment of the present application will be described below with reference to Fig. 4. Fig. 4 is a block diagram of an output device according to an embodiment of the present application.

[0223] The output device 2600 according to an embodiment of the present application may include a second communication module 2610, a second memory 2620, and a second controller 2630.

[0224] The second communication module 2610 can communicate with any external device including the image acquisition device 1000 and the image analysis device 2000. In other words, the output device 2600 can transmit and receive images from the image acquisition device 1000 and transmit and receive image analysis results from the image analysis device 2000 via the second communication module 2610. In addition, the output device 2600 can transmit and receive data to and from any external device including a repeater, etc. via the second communication module 2610.

[0225] For example, the output device 2600 may receive, via the second communication module 2610, images acquired from the image acquisition device 1000, data related to the reliability of analysis results associated with the quality of images acquired from the image analysis device 2000, data on image segmentation results, data such as morphological indices, etc. In another example, the output device 2600 may transmit data related to user input received from the input module 2640 (described below) to the image analysis device 2000 or transmit data processed by the output device 2600 to the image analysis device 2000 via the second communication module 2610. In another example, the output device 2600 may connect to the Internet via the second communication module 2610 to upload data related to user input or data processed on the output device 2600.

[0226] The second communication module 2610 can be broadly divided into a wired type and a wireless type. Since the wired type and the wireless type each have advantages and disadvantages, in some cases, the image analysis device 2000 can be provided with both a wired type and a wireless type.

[0227] Here, in the case of a wired type, LAN (Local Area Network) or USB (Universal Serial Bus) communication is a typical example, but other methods are also possible.

[0228] In addition, in the case of a wireless type, a communication method of the WPAN (Wireless Personal Area Network) series such as Bluetooth or Zigbee can be mainly used. However, the wireless communication protocol is not limited thereto, and the wireless type communication module can also use a communication method of the WLAN (Wireless Local Area Network) series such as Wi-Fi or other known communication methods.

[0229] The second memory 2620 can store various types of information. Various types of data can be temporarily or semi-permanently stored in the second memory 2620. Examples of the second memory 2620 include a hard disk drive (HDD), a solid state drive (SSD), a flash memory, a read-only memory (ROM), and a random access memory (RAM).

[0230] The second memory 2620 may be provided in a form built into the output device 2600 or in a form that is detachable. The second memory 2620 may store various data necessary for the operation of the output device 2600, including an operating system (OS) for driving the output device 2600 and programs for operating each component of the output device 2600. For example, the second memory 2620 may store various data related to images, information related to analysis results, and user input data.

[0231] The second controller 2630 may control the overall operation of the output device 2600. For example, the second controller 2630 may load a program for operating the output device 2600 from the second memory 2620 and execute the program.

[0232] The second controller 2630 may be realized by hardware, software, or a combination thereof, such as a CPU (Central Processing Unit) or similar device. In terms of hardware, it may be provided in the form of an electronic circuit that processes electrical signals and executes control functions, and in terms of software, it may be provided in the form of a program or code that drives a hardware circuit.

[0233] Meanwhile, referring back to FIG. 4, an output device 2600 according to an embodiment of the present application may include an input module 2640 and an output module 2650 .

[0234] At this time, the output device 2600 can obtain a user input using the input module 2640 and the output module 2650 and output information corresponding to the user input.

[0235] For example, the output device 2600 can utilize the input module 2640 to receive user inputs requesting data acquisition, user inputs directing preprocessing, user inputs related to image segmentation, and user inputs related to reference scan conditions relevant to the calculation of morphological indices, and output corresponding information via the output module 2650.

[0236] In this case, the output device 2600 may further include a user interface for receiving a user input via an input module 2640 and outputting information corresponding to the user input via an output module 2650 .

[0237] In one embodiment, a user can input conditions or settings related to the output of the output device 2600 via the input module 2640, or select a portion of the multiple pieces of output data.

[0238] For example, a user may set reference scan conditions associated with correction parameters for correcting morphological values ​​or morphological indices acquired from a target image via the input module 2640. In this case, the output device 2600 may be configured to output analysis results corresponding to the corrected morphological values ​​or morphological indices based on the reference scan conditions (e.g., magnetic field strength, manufacturer, setting parameters of the image device, etc.) received from the input module 2640.

[0239] For example, if a user changes the magnetic field strength and wishes to receive image analysis results corresponding to the changed magnetic field strength, the scan conditions for the magnetic field strength can be input via the input module 2640. In this case, the output device 2600 can be configured to output image analysis results corrected based on the scan conditions for the magnetic field strength input by the user. In this case, the output device 2600 can transmit input data related to the scan conditions for the magnetic field strength input by the user to the image analysis device 2000, and the image analysis device 2000 can modify the image analysis results based on the correction parameters acquired based on the received user input and transmit the modified image analysis results to the output device 2600. Of course, the output device 2600 can be configured to receive the correction parameters from the image analysis device 2000 and modify and output the image analysis results.

[0240] This will be described in more detail below in connection with FIGS. 66 and 67.

[0241] However, the above content is merely an example and is not limited thereto, and the output device 2600 may be implemented to receive user input via the input module 2640 in order to output any type of information that the user wishes to provide.

[0242] The input module 2640 may be implemented in various forms, such as a mouse, a keyboard, a touchpad, etc.

[0243] An output module 2650 may be provided to output image analysis results received from the image analysis device 2000, etc.

[0244] For example, the output device 2600 may output, via the output module 2650, information regarding whether artifacts are present in the target image received from the image analysis device 2000 and the reliability of the analysis results in text or any suitable visual form.

[0245] According to another example, the output device 2600 may output the segmentation results of the target image received from the image analysis device 2000 via the output module 2650. In this case, the segmentation results may be implemented to separate and output the anatomical structures of the body in a visual graphic form.

[0246] In another example, the output device 2600 may output the analysis results of the target image received from the image analysis device 2000 via the output module 2650. At this time, the analysis results of the target image may be output in the form of a report. Furthermore, the morphological values ​​and morphological indices of the anatomical structures included in the analysis results of the target image may be output using a statistical technique such as a graph. Furthermore, the morphological values ​​and morphological indices of the anatomical structures included in the analysis results of the target image may be output using a statistical technique such as a graph after comparing them with statistical data of a comparison subject population.

[0247] The output module 2650 may be implemented in various forms, such as a display.

[0248] In accordance with the above, the image analysis device 2000 and the output device 2600 have been described as separate devices. However, this is for illustrative purposes only, and the image analysis device 2000 and the output device 2600 may be implemented as a single device.

[0249] The image analysis device 2000 receives medical images transmitted from the image acquisition device 1000 and can perform pre-processing such as alignment of the medical images, normalization of the brightness or intensity of the medical images, and noise removal.

[0250] The image analysis device 2000 can also receive medical images transmitted from the image acquisition device 1000 and perform segmentation of the medical images. In this case, the medical image segmentation according to one embodiment of the present application may be performed using a trained neural network model.

[0251] The image analysis device 2000 according to an embodiment of the present application can perform an operation to control the quality of medical images, thereby increasing the reliability of medical image analysis results.

[0252] The image analysis device 2000 according to an embodiment of the present application may perform an operation of calculating morphological indices based on medical images acquired from a subject. Therefore, the image analysis device 2000 according to an embodiment of the present application may calculate accurate morphological indices in relation to the anatomical structure of the subject's body. This advantageously provides an accurate and objective indicator of a subject's disease. In particular, since the morphological indices can be calculated by image segmentation without aligning the brain image associated with the subject to a standard brain model, a personalized diagnostic aid indicator related to brain disease may be provided.

[0253] The image analysis device 2000 according to an embodiment of the present application can perform an operation of correcting morphological values ​​or morphological indices taking into account the scan conditions or body parts under which the medical images are acquired. Therefore, the image analysis device 2000 according to an embodiment of the present application can calculate accurate morphological indices in relation to the anatomical structure of the subject's body. In other words, although there may be some error in the morphological values ​​calculated depending on the scan conditions or body parts, the image analysis device 2000 according to an embodiment of the present application can correct the primarily calculated morphological values ​​or morphological indices taking into account the scan conditions or body parts. This has the advantageous effect of providing accurate and objective indices for the subject's disease.

[0254] The image analysis device 2000 according to an embodiment of the present application can prioritize various image analysis results according to the state of the subject or the diagnostic field, and output the image analysis results selected based on the priority. Therefore, the image analysis device 2000 according to an embodiment of the present application can selectively provide the user with necessary index information from various index information obtained through the analysis of the target medical image, thereby improving usability.

[0255] Some operations performed by one embodiment of image analysis device 2000 are described in more detail below.

[0256] For convenience of explanation, the following description will focus on an embodiment analyzing images related to the brain, but the present disclosure is not limited to the brain, and various embodiments disclosed herein may be applied to medical images of any suitable body part other than the brain.

[0257] In addition, in the following, the terms medical image, brain image, subject image (or target image), etc. are used interchangeably, but this is for the convenience of explanation only, and medical image, brain image, subject image (or target image) should be interpreted as referring to the images analyzed by both image analysis devices (2000, 3000, 4000).

[0258] Furthermore, the following content may be performed by an image analyzer (2000, 3000, 4000) according to an embodiment of the present application. The reference numerals indicating the image analyzers are merely used for the convenience of explanation to distinguish and explain the operations of the image analyzers, and the reference numerals do not limit the image analyzers.

[0259] The image analysis device 2000 according to one embodiment of the present application can acquire brain images and information associated with the brain images.

[0260] Specifically, the image analysis device 2000 can acquire brain images from the image acquisition device 1000. The image analysis device 2000 can also acquire information related to the brain images from the image acquisition device 1000 or any external device (e.g., a server).

[0261] More specifically, the image analysis device 2000 can acquire brain images and data related to the brain images from the image acquisition device 1000 via the first communication module 2010 .

[0262] At this time, the format of the brain image may be in various medical image formats, for example, the brain image may be in DICOM, Nifti (Neuroimaging Ingormatics Technology Initiative), or any other suitable format.

[0263] Data associated with a brain image may include data contained in the brain image, data related to the scanning conditions under which the brain image was acquired, and data about the subject of the brain image.

[0264] In this case, the data contained in the brain image may be data related to pixels or voxels contained in the brain image, data related to the orientation of the brain image, and any structured metadata for the brain image.

[0265] In particular, data relating to the scan conditions under which the brain image was acquired or data relating to the subject of the brain image, as described below, may be structured for the brain image as metadata.

[0266] On the other hand, the data related to the scanning conditions under which the brain image is acquired may be the magnetic field strength of the image acquisition device 1000 under which the brain image is acquired, the setting parameters of the image acquisition device 1000, or data related to the manufacturer of the image acquisition device.

[0267] Since the morphological values ​​obtained from the brain image may be affected by the scan conditions, the image analysis device 2000 according to an embodiment of the present application may acquire data related to the scan conditions and perform corrections for the morphological values, as will be described in detail below with reference to Figures 58 to 67.

[0268] The data on the subject of the brain image may be personal information or medical information on the subject of the target image analyzed by the image analysis device 2000. For example, the data on the subject of the brain image may include personal information related to the subject's gender, age, etc., and various medical information related to a brain disease (e.g., dementia, Alzheimer's disease, depression, stroke, etc.) or information related to the underlying disease.

[0269] In particular, because gender and age are important variables in brain diseases, data related to the gender and age of the subject can serve as a basis for the image analysis device 2000 to determine or acquire correction parameters. For example, if the target image to be analyzed is acquired from a first subject having first characteristics related to gender and age, the image analysis device 2000 can determine correction parameters for correcting morphological values ​​and morphological indices acquired from the target image using a first brain image acquired from a second subject having the first characteristics and correction parameters acquired from the second brain image. For example, the image analysis device 2000 can determine correction parameters for correcting morphological values ​​and morphological indices acquired from the target image using correction parameters acquired from a second subject having a similar age and the same gender as the first subject of the target image.

[0270] Alternatively, data relating to the gender and age of the subject can be taken into account when using statistical techniques to output relative percentiles of the subject's morphological indices according to gender and age.

[0271] Additionally, the image analysis device 2000 according to an embodiment of the present application can obtain information related to behaviors associated with brain image analysis.

[0272] Specifically, the image analysis device 2000 can acquire information related to a brain-related template for pre-processing and aligning brain images from any external device, as will be described in detail below with reference to FIGS. 5, 6, and 48.

[0273] Additionally, the image analysis device 2000 can obtain information related to a brain atlas, which serves as a reference for segmenting brain images, from any external device.

[0274] By way of example, the information related to a brain atlas can be information related to an atlas related to brain structure, such as Automated Anatomical Labeling (Tzourio-Mazoyer 2002), Desikan-Killiany Atlas (Desikan 2006), Destrieux Atlas (Destrieux 2010), Harvard-Oxford cortical / subcortical atlases (Makris 2006), MICCAI 2012 Multi-Atlas Labeling Workshop and Challenge (Neuromorphometrics), Hammersmith atlas (Hammers 2003, Gousias 2008, Faillenot 2017), HCP MMP 1.0 (Glasser 2016), JuBrain / Juelich histological atlas (Eickhoff 2005), or MarsAtlas (Auzias 2016).

[0275] As another example, the information associated with a brain atlas may be information associated with an atlas related to brain function. For example, functional atlases related to brain function include Mindboggle 101 (Klein 2012), Cortical Area Parcellation from Resting-State Correlations (Gordon 2016), Consensual Atlas of Resting-State Network (CAREN, Doucet 2019), Brainnetome Atlas parcellation (Fan 2016), Local-Global Parcellation of the Human Cerebral Cortex (Schaefer 2018), Human Motor Area Template (Mayka 2005), Sensorimotor Area Tract Template (Archer 2017), AICHA: An atlas of intrinsic connectivity of homotopic areas (Joliot 2015), Yeo 2011 functional parcellations (Yeo 2011), PrAGMATiC (Huth 2016), fMRI-based random parcellations (Craddock 2011), and voxelwise parcellations. (Lead-DBS), SUIT Cerebellar parcellation (Diedrichsen 2006), or Buckner functional cerebellar parcellation (Buckner 2011).

[0276] The image analysis device 2000 can transmit information about the brain atlas associated with the brain described above to the learning device 2200, and the learning device 2200 can be configured to learn a neural network model for segmentation of brain images based on the information about the brain atlas.

[0277] However, the information on the atlases related to brain structure or brain function described above is merely exemplary, and the image analysis device 2000 may be implemented to obtain any appropriate atlas information related to the brain and use it as a basis for training the artificial neural network model for image segmentation in the learning device 2200.

[0278] The image analysis device 2000 can receive user input from the input module 2040 of the image analysis device 2000 or the input module 2640 of the output device 2600 .

[0279] The image analysis device 2000 may acquire user input related to a disease to be diagnosed. For example, the image analysis device 2000 may acquire user input related to a brain disease (e.g., dementia, depression, stroke, etc.) corresponding to a disease to be diagnosed associated with a target image.

[0280] The image analyzer 2000 may acquire user input related to a patient. For example, the image analyzer 2000 may acquire user input related to the gender, age, name, etc. of a patient associated with a target image.

[0281] The image analysis device 2000 can obtain user input related to image analysis.

[0282] For example, the image analysis device 2000 may receive user input related to image pre-processing, such as correcting the intensity of a target image or removing noise.

[0283] For example, the image analysis device 2000 can receive user input related to segmentation, such as to modify label data output from a brain atlas or neural network model that is taken into account in the segmentation.

[0284] For example, the image analysis device 2000 may receive user input regarding scan conditions under which a target image was captured via an input module. Information regarding scan conditions under which a target image was captured may be included in metadata of the target image and obtained, or may be obtained through user input.

[0285] For example, the image analyzing device 2000 may receive, via an input module, a user input for reference scan conditions associated with the criteria for correction parameters for correcting morphological values ​​or morphological indices associated with the brain calculated based on the segmentation results. In this case, the image analyzing device 2000 may obtain the correction parameters based on the user input for the reference scan conditions, etc., and correct the morphological values ​​or morphological indices associated with the brain. This will be described in detail below with reference to FIGS. 66 and 67.

[0286] In another example, when outputting image analysis results, the image analyzing device 2000 may receive user input via an input module to select or prioritize some of the results to be output. In this case, the image analyzing device 2000 may be implemented to selectively output information on the analysis results to be output based on the user input, or to output analysis results that reflect the user's priorities. This will be described in detail below with reference to FIGS. 68 to 81.

[0287] The data acquired by the image analysis device 2000 may be stored in the first memory 2020 of the image analysis device 2000 or in any external device (e.g., a server) of the image analysis device 2000. The data acquired by the image analysis device 2000 may also be transmitted to the learning device 2200 or the correction parameter acquisition device 2400. Alternatively, the data acquired by the image analysis device 2000 may be transmitted to the output device 2600 or to any external device (e.g., a server).

[0288] The image analysis device 2000 according to an embodiment of the present application can perform image pre-processing. The image analysis device 2000 can perform pre-processing to improve the accuracy of image analysis. The image analysis device 2000 may be provided to perform image pre-processing to derive more accurate segmentation results before performing a segmentation operation on the image.

[0289] As an example, the image analysis device 2000 may be provided to convert the format of the image acquired from the image acquisition device 1000. Specifically, by unifying the format of the images to be analyzed, the neural network model can be trained more stably and accurately. More specifically, performing image analysis using images having the same format as the images used to train the neural network model is more stable and accurate. Therefore, the image analysis device 2000 according to an embodiment of the present application may be provided to convert the format of the image acquired from the image acquisition device 1000.

[0290] For example, the image analysis device 2000 according to an embodiment of the present application may convert an image of a first format acquired from the image acquisition device 1000 into an image of a second format. For example, the image format acquired from the image acquisition device 1000 may be the DICOM format, which is commonly used for medical images. In this case, image segmentation and morphological indices may be calculated based on the image in the DICOM format.

[0291] However, in analyzing brain images, it is relatively easy to analyze images or train an artificial neural network using brain images in Nifti format in a brain image analysis system. Therefore, an image analysis device 2000 according to an embodiment of the present application may be provided to perform an operation of converting the format of images acquired from the image acquisition device 1000 into the Nifti format.

[0292] However, the above-mentioned formats are merely examples, and the image analysis device 2000 may be provided to perform a conversion operation to any suitable format other than the Nifti format, if necessary. Also, the format of the image acquired from the image acquisition device 1000 may be a medical image in any format other than the DICOM format, and in this case, the image analysis device 2000 may be provided to perform a conversion operation to any suitable format.

[0293] For example, the image analysis device 2000 may be provided to remove noise or correct artifacts that may be present in the images acquired from the image acquisition device 1000. For example, a blurring technique and a technique using a median filter may be used to remove noise. The image analysis device 2000 may derive more accurate image segmentation results by removing noise and correcting artifacts, and may calculate morphological indices based on the segmentation results with improved accuracy, thereby providing objective diagnostic support indices for brain-related diseases with high reliability.

[0294] As an example, the image analysis device 2000 may be provided to perform an operation of correcting the intensity of the image acquired from the image acquisition device 1000. By appropriately correcting the intensity, noise that may be present in the image can be removed, and an image specific to the anatomical structure of the brain to be analyzed can be acquired.

[0295] As an example, the image analysis device 2000 may be provided to perform an operation of smoothing the image acquired from the image acquisition device 1000. For example, a technique using blurring or a Gaussian filter may be used as a method of smoothing the image.

[0296] As an example, the image analyzer 2000 may be provided to adjust the proportions of an image acquired from the image acquisition device 1000 or to perform an image cropping operation. For example, the image analyzer 2000 may be implemented to utilize any suitable cropping technique to crop the image. Alternatively, the image analyzer 2000 may be implemented to utilize any suitable image resizing technique to adjust the proportions of the image, such as on-demand image resizing, Lambda image resizing, a resizing method using a CILanczosScaleTransform filter, or a resizing method using a CIFilter.

[0297] For example, the image analysis device 2000 may perform an operation of converting an image acquired from the image acquisition device 1000 into an image acquired under scanning conditions different from the scanning conditions under which the image was acquired. For example, the image analysis device 2000 may be provided to perform an operation of converting an image acquired under a first scanning condition of an MRI device into an estimated image that would have been acquired under a second scanning condition of the MRI device. Such an image conversion operation may be performed using MR conversion technology or an artificial intelligence model. However, without being limited thereto, the image analysis device 2000 may be provided to perform image conversion taking into account the scanning conditions using any appropriate software or image processing technology.

[0298] For example, the image analysis device 2000 may be implemented to perform a preprocessing operation corresponding to a preprocessing operation on an image performed by the learning device 2200, which will be described later. For example, if a neural network model is trained on an image using a first preprocessing technique in the learning device 2200, the image analysis device 2000 may be implemented to preprocess a target image using a preprocessing technique corresponding to the first preprocessing technique. This may allow image segmentation using a neural network model to be more stably and accurately performed.

[0299] Hereinafter, with reference to FIGS. 5 and 6, the image alignment related operations of the image analysis device 2000 according to one embodiment of the present application will be described.

[0300] FIG. 5 is a diagram showing an example of the image alignment operation of the image analysis device 2000.

[0301] FIG. 6 is a diagram showing an example of the image alignment operation of the image analysis device 2000.

[0302] The image analysis device 2000 may be implemented to perform operations for aligning the brain images based on data contained in the brain images relating to the orientation of the brain images.

[0303] As an example, the image analysis device 2000 may be implemented to perform an operation to align the brain images before performing an operation to segment the images.

[0304] 5, the image acquisition device 1000 may acquire data (i, j, k) related to the direction of the captured image in consideration of the direction of the reference coordinate axis (RAS) of the object 100. Specifically, the image acquisition device 1000 may acquire data (i, j, k) related to the direction of the captured image in consideration of information (x, y, z) on the coordinate axes of the image acquisition device 1000 and information (RAS) on the reference coordinate axis of the object 100.

[0305] Data related to the orientation of an image can be structured in the metadata for the image, or can be communicated to the image analysis device 2000 separately from the image.

[0306] The image analysis device 2000 may be implemented to align the image to correspond to the RAS (Right-Anterior-Superior) direction of the object 100 based on the data (ijk) related to the image direction.

[0307] According to the image alignment operation of the image analysis device 2000 described above, the brain images that form the basis of the segmentation can be aligned in a common direction before the segmentation operation is performed, thereby preventing inaccurate segmentation results and ensuring the stability of the neural network segmentation operation.

[0308] The image analyzer 2000 may be implemented to perform spatial normalization of the brain images.

[0309] Specifically, an artificial neural network model for brain image segmentation, etc., can be stably operated on brain images that correspond to the spatial distribution of the training images used as “training data.” In other words, if the brain images differ from the spatial distribution of the training images, there is a possibility that the trained artificial neural network will not be stably operated.

[0310] Therefore, the image analysis unit 2000 may be implemented to perform spatial normalization of the brain images in order to reduce the spatial uncertainty of the brain images.

[0311] For example, the image analysis device 2000 according to an embodiment of the present application may be implemented to perform spatial normalization of the brain image based on a brain template. Specifically, the image analysis device 2000 can convert the coordinates of the brain image by matching the brain image to the brain template so that the spatial distribution of the brain image is optimal for the artificial neural network model.

[0312] For example, the brain template for matching the brain image can be an MNI template, a Talairach template, or the like.

[0313] As described above, the image analysis device 2000 performs an operation of matching a brain image to a brain template. Here, "matching" does not mean matching the internal brain elements of the brain image to the internal brain elements of the brain template, but may simply mean matching the positional space of the brain image.

[0314] The spatial normalization operation of the image analysis device 2000 described above can remove spatial uncertainty about the image, ensuring the stability of the segmentation operation using a neural network, and obtaining segmentation results with improved accuracy.

[0315] Furthermore, by matching the brain image to the brain template, data associated with transformed coordinates can be generated. The data associated with the transformed coordinates can be used to convert the brain image to its original coordinates after the segmentation described below is completed. In other words, the image analysis device 2000 may be implemented to perform an operation of converting the segmentation results obtained based on the spatial distribution of the brain template to correspond to the original brain image using the data associated with the transformed coordinates.

[0316] Through the operation of the image analysis device 2000, the user can be provided with the analysis results of the brain image relative to the original coordinates.

[0317] For example, when morphological indices associated with a brain image are output through an output module, they may be output as morphological indices corrected using correction parameters, etc. Conversely, when visual information about the brain (e.g., info.13 shown in FIG. 57 or segmentation results, etc.) is output through an output module, an inversely transformed image based on the transformed coordinates described above, i.e., visual information associated with the brain based on the original brain image, may be output.

[0318] Hereinafter, the image segmentation-related operations of the learning device 2200 and the image analysis device 2000 according to one embodiment of the present application will be described with reference to FIGS.

[0319] According to one embodiment of the present application, the image segmentation operation may be performed using a trained neural network model, although any suitable method may be used to achieve the image segmentation operation according to one embodiment of the present application without using a neural network model.

[0320] The following description will focus on the operation of training a neural network model for image segmentation and the execution of image segmentation using the trained neural network model.

[0321] Please refer to Figure 7. Figure 7 is a diagram illustrating a process flow for image segmentation according to one embodiment of the present application.

[0322] Referring to FIG. 7, a process for image segmentation according to one embodiment of the present application may include a process P1000 for training an artificial neural network model for image segmentation, and a process P2000 for segmenting a target image using the trained artificial neural network model.

[0323] At this time, the learning process P1000 can be realized by a learning device 2200 according to an embodiment of the present application.

[0324] Also, the segmentation process P2000 can be implemented by an image analysis device 2000 according to one embodiment of the present application.

[0325] At this time, the parameters of the neural network model obtained by the learning process P1000 implemented in the learning device 2200 can be transmitted to the image analysis device 2000 via any suitable communication module.

[0326] The image analysis device 2000 may then be implemented to perform segmentation of the target image based on the parameters of the neural network model obtained by the training process P1000.

[0327] The training process P1000 according to one embodiment of the present application may include a process P1100 for acquiring an image dataset, a process P1200 for training a neural network model, a process P1300 for validating the neural network model, and a process P1400 for acquiring parameters of the neural network model.

[0328] A method for training a neural network model of the training device 2200 according to an embodiment of the present application will be described below with reference to Fig. 8. Fig. 8 is a flowchart illustrating a method for training a neural network model of the training device 2200 according to an embodiment of the present application.

[0329] Referring to FIG. 8, a neural network model training method of a training device 2200 according to one embodiment of the present application may include step S1100 of acquiring an image dataset, step S1200 of screening the image dataset, step S1300 of preprocessing and aligning the image dataset, step S1400 of training and verifying the neural network model, and step S1500 of acquiring neural network model parameters.

[0330] In step S1100 of acquiring an image dataset, the learning device 2200 according to an embodiment of the present application can acquire the image dataset from the image acquisition device 1000 or any external device.

[0331] Please refer to Figure 9. Figure 9 is an exemplary structural diagram of an image dataset according to one embodiment of the present application.

[0332] The image dataset DS acquired by the learning device 2200 can include at least one or more image data. In other words, the image dataset DS acquired by the learning device 2200 can include at least one or more image data, such as the first image data ID1, the second image data ID2, and the nth image data IDn.

[0333] At this time, the image data included in the image dataset DS acquired by the learning device 2200 may include data related to images and labels.

[0334] For example, referring to FIG. 9, first image data ID1 included in image data set DS may include data associated with a first image I1 and a first label L1.

[0335] Specifically, the first label L1 can be obtained by manually labeling the first image I1 from a clinician who can diagnose the brain disease, or the first label L1 can be automatically labeling using any suitable image segmentation technique.

[0336] Data related to images and labels contained in the image data can serve as a basis for training an artificial neural network model and validating the artificial neural network model in connection with a training method according to one embodiment of the present application.

[0337] Meanwhile, the image data included in the image dataset DS may include data related to the scan conditions. In this case, the data related to the scan conditions may be data related to magnetic field strength, setting parameters of the image device, and / or data related to the manufacturer of the image device, as described above. In addition, the data related to the scan conditions may be structured as metadata for the image data.

[0338] 9, the first image data ID1 may include data related to a first scan condition (SC1) that is related to the scan condition under which the first image data ID1 was captured. For example, if the first image data ID1 was captured under a magnetic field strength of 3 T, information related to the magnetic field strength corresponding to 3 T may be structured as metadata for the first image data ID1 and acquired by the learning device 2200 through the dataset acquisition step S1100.

[0339] Data related to the scanning conditions contained in the image data may be taken into consideration to obtain correction parameters for correcting morphological values ​​or morphological indices corresponding to target elements obtained from the target image described below.

[0340] Although Figure 9 only shows the data contained in the first image data ID1, this is merely an example, and the image data of the image data set including the second image data ID2 or the nth image data IDn, etc., may include data related to the image, label, and scanning conditions.

[0341] Additionally, the learning device 2200 may be implemented to obtain information related to a brain atlas relating to brain structure or brain function for segmentation associated with brain images.

[0342] Specifically, the learning device 2200 can obtain information related to the brain atlas relating to the brain structure or brain function described above from the image analysis device 2000 or any external device.

[0343] In this case, the learning device 2200 may be realized to train an artificial neural network model (described later) or to verify the artificial neural network model, taking into account information related to the brain atlas.

[0344] In step S1200 of screening the image dataset, the learning device 2200 according to one embodiment of the present application may be implemented to perform an operation of screening the image dataset acquired in step S1100 of acquiring the image dataset, or selecting only a portion of the image data from the image data included in the image dataset.

[0345] For example, some image data in the acquired image data set may not be suitable for training an artificial neural network model for segmentation. For example, some image data may contain severe artifacts or severe noise. Such image data may not be suitable for training an artificial neural network model.

[0346] Thus, the learning device 2200 may be implemented to screen image data included in an acquired image dataset and select image data that is effective for training an artificial neural network model.

[0347] In the step S1300 of pre-processing and aligning the image dataset, the learning device 2200 according to one embodiment of the present application may be implemented to perform pre-processing operations to remove noise or artifacts from the images included in the image dataset or to correct the intensity of the images.

[0348] Furthermore, the learning device 2200 according to an embodiment of the present application may be implemented to perform operations of aligning images based on data related to the orientation of the images, or matching images to a brain template, and performing spatial normalization to align images.

[0349] In this regard, the preprocessing operation of the image analysis device 2000 described above and the image alignment operation described in relation to Figures 5 and 6 may also be provided to be implemented in the learning device, or may be implemented by performing the image preprocessing and alignment operations in the image analysis device 2000 and then transmitting the results to the learning device 2200 via data transmission and reception between the learning device 2200 and the image analysis device 2000.

[0350] In the neural network model training and validation step S1400, the training device 2200 for image segmentation according to an embodiment of the present application can train an artificial neural network model for image segmentation.

[0351] Specifically, the artificial neural network model may include an input layer for receiving input of image data, an output layer for outputting a labeling result, which is a segmentation result, and a hidden layer including at least one node.

[0352] In this case, the learning device 2200 may be implemented to receive input of image data contained in the acquired image dataset via an input layer, and to obtain labeling results for the acquired image data by the neural network model via an output layer.

[0353] For example, the learning device 2200 may be implemented to train an artificial neural network configured to receive first image data ID1 as input and output a first label (L1') via an output layer. The learning device 2200 may also receive second image data ID2 as input via an input layer and obtain a second label (L2') output via an output layer.

[0354] Furthermore, the learning device 2200 according to an embodiment of the present application can train an artificial neural network by taking into consideration the above-mentioned brain atlas related to the structure of the brain and the brain atlas related to the function of the brain.

[0355] According to one embodiment, a learning device can train a neural network model to segment a brain image based on a predetermined brain atlas. The brain atlas can include a plurality of brain regions, including a first brain region and a second brain region. The learning device can use an image labeled with a first region corresponding to the first brain region and a second region corresponding to the second brain region to train a neural network model to obtain the first region and the second region from the image as input.

[0356] For example, the learning device 2200 according to an embodiment of the present application may be implemented to train an artificial neural network model for segmentation of image data included in an image dataset based on the Desikan-Killiany Atlas (Desikan 2006), which is an atlas used to obtain regions corresponding to a plurality of brain regions, including a first brain region and a second brain region, of the cerebral cortex.

[0357] In this case, the learning device 2400 can learn a neural network model for segmenting the cortical regions of the brain by taking into account the Desikan-Killiany Atlas (Desikan 2006) and using image data in which a first region corresponding to a first brain region and a second region corresponding to a second brain region are labeled.

[0358] However, the above-mentioned brain atlases are merely examples, and any suitable brain atlas may be considered depending on the purpose and region of interest of the image data segmentation.

[0359] Furthermore, the learning device 2200 according to an embodiment of the present application may be implemented to train a neural network model according to the scanning conditions under which the image data was captured.

[0360] For example, the training device 2200 may be implemented to use a first neural network model to segment regions from a first image acquired under a first scanning condition. Conversely, the training device 2200 may be implemented to use a second neural network model to segment regions from a second image acquired under a second scanning condition. In other words, the training device 2200 may train neural network models differently based on the scanning conditions under which the images were taken.

[0361] Furthermore, in the segmentation process P2000, the image analysis device 2000 according to one embodiment of the present application may be implemented to segment a target image acquired under a first scanning condition using a trained first neural network model, and to segment a target image acquired under a second scanning condition using a trained second neural network model.

[0362] As a result, the image analysis device 2000 according to an embodiment of the present application can perform segmentation of the target image using an optimal neural network model for each scan condition, thereby obtaining multiple regions more stably and accurately.

[0363] In the following, with reference to FIGS. 10-11, an example of an artificial neural network model that can be used by a learning device 2200 according to one embodiment of the present application will be described.

[0364] FIG. 10 is an example of an artificial neural network model that can be used by a learning device 2200 according to an embodiment of the present application.

[0365] FIG. 11 is another example of an artificial neural network model that can be used by the learning device 2200 according to one embodiment of the present application.

[0366] Referring to FIG. 10, a learning device 2200 according to one embodiment of the present application can utilize U-net in an artificial neural network for image segmentation.

[0367] The U-net used for image segmentation may be configured with an architecture including a contraction path and an expansion path.

[0368] Specifically, the contraction path of the U-net may be configured to perform two convolutions and max pooling consecutively, allowing features related to the image to be extracted.

[0369] However, in the contraction path, the magnitude of the feature map also decreases, so the U-net may be configured to additionally include an expansion path to restore the magnitude of the feature map.

[0370] The U-net extension path can be configured to perform up-convolution and two convolutions consecutively, allowing the size of the image and feature map to be extracted.

[0371] Furthermore, the U-net is architectured to concatenate feature maps at the same level, allowing it to provide location information related to features from the contraction pathway to the expansion pathway.

[0372] At this time, based on the difference between the label of the input image and the label of the output segmentation map, the parameters or weights of at least one node in the layer containing the U-net can be adjusted so that the difference between the label of the input image and the label of the output segmentation map is minimized.

[0373] 11, a learning device 2200 according to an embodiment of the present application can utilize U-net++ in an artificial neural network for image segmentation. U-net++ is an artificial neural network model that uses the dense block idea of ​​DenseNet to improve the performance of U-net, and differs from U-net in that a convolutional layer exists in the skip path to bridge the semantic gap between the feature maps of the encoder and decoder, and a dense skip connection exists in the skip path to improve the gradient flow.

[0374] Specifically, the learning device 2200 may be implemented to input an input image to an input layer of a U-net++ neural network model and obtain label information output through an output layer. In this case, the learning device 2200 may adjust the parameters or weights of at least one node in a hidden layer included in the U-net++ based on the difference between the label information included in the input image and the label information output from the neural network model.

[0375] Specifically, the learning device 2200 is implemented to repeatedly perform the operation of adjusting the parameters or weights of at least one node described above, and can obtain the parameters or weights of the node that minimizes the difference between the label information included in the input image and the label information output from the neural network model.

[0376] As described above, the learning device 2200 according to an embodiment of the present application can perform an operation of training an artificial neural network model based on the output label results.

[0377] Specifically, step S1400 of training the artificial neural network model may obtain data related to labels included in the image data obtained from step S1100 of obtaining the image data set.

[0378] In this case, the learning device 2200 may be implemented to train an artificial neural network model based on image data and label data output through an output layer of the neural network model.

[0379] More specifically, the learning device 2200 may be implemented to train a neural network model by adjusting the weight or parameter of at least one node included in a hidden layer of the neural network model based on the difference between the label data included in the image data and the label data output through the output layer of the neural network model.

[0380] For example, the learning device 2200 may input the first image data ID1 to an input layer of an artificial neural network to acquire label data corresponding to the first A label L1A. In this case, the learning device may train a neural network model based on the label data corresponding to the first label L1 included in the first image data ID1 and the label data related to the first A label L1A. For example, the learning device 2200 may be implemented to train the neural network model by adjusting the weight or parameter of at least one node included in a hidden layer of the neural network model based on the difference between the first label L1 and the first A label L1A.

[0381] As another example, the learning device 2200 may input the ith image data IDi to an input layer of an artificial neural network to obtain label data corresponding to the ith A-label LiA. In this case, the learning device may train a neural network model based on the label data corresponding to the ith label Li included in the ith image data IDi and the label data related to the iA-label LiA. For example, the learning device 2200 may be implemented to train a neural network model by adjusting the weight or parameter of at least one node included in a hidden layer of the neural network model based on the difference between the ith label Li and the iA-label LiA. Here, i may be any number.

[0382] In step S1400 of validating the artificial neural network model, the learning device 2200 according to an embodiment of the present application can validate the artificial neural network model.

[0383] For example, the learning device 2200 according to an embodiment of the present application may obtain label data output through a trained neural network model based on at least one image data included in the image dataset DS. At this time, the learning device 2200 may verify the trained neural network model based on the label data associated with the at least one image data and the label data output through the trained neural network model.

[0384] For example, the learning device 2200 can verify whether the parameters or weights of the nodes in the hidden layer of the trained neural network model are appropriate by comparing the similarity between label data associated with at least one image data and the label data output through the trained neural network model.

[0385] In step S1500 of obtaining an artificial neural network model, the learning device 2200 according to an embodiment of the present application can obtain a neural network model including at least one node having weights or parameters that minimize the difference between data related to labels included in the image data and data related to labels output by the artificial neural network by repeatedly performing operations of learning the artificial neural network model on image data included in the image dataset and verifying the artificial neural network model.

[0386] The obtained node parameters and weights can be used in the segmentation process P2000 artificial neural network model for image segmentation.

[0387] While the above description has focused on segmentation using an artificial neural network, the image analysis device 2000 disclosed in this application can utilize various image segmentation algorithms, including image segmentation using an artificial neural network.

[0388] As an example, the image segmentation algorithm may be provided as a machine learning model. A representative example of a machine learning model may be an artificial neural network. Specifically, a representative example of an artificial neural network is a deep learning series artificial neural network including an input layer that receives input data, an output layer that outputs results, and a hidden layer that processes data between the input layer and the output layer. Specific examples of artificial neural networks include convolutional neural networks, recurrent neural networks, deep neural networks, and generative adversarial networks. In this specification, the term "artificial neural network" should be interpreted in a comprehensive sense to include the above-mentioned artificial neural networks, various other forms of artificial neural networks, and artificial neural networks in combinations thereof, and is not necessarily a deep learning series.

[0389] Furthermore, the machine learning model does not necessarily have to be in the form of an artificial neural network model, but may also include nearest neighbor algorithms (KNN), random forests, support vector machines (SVM), principal component analysis (PCA), etc., and the above-mentioned techniques may include all ensembled forms and forms combined in various other ways. Meanwhile, in embodiments that mainly refer to an artificial neural network, unless otherwise specified, the artificial neural network may be replaced with another machine learning model.

[0390] Furthermore, in this specification, the image segmentation algorithm is not necessarily limited to a machine learning model, that is, the image segmentation algorithm may include various judgment and decision algorithms that are not machine learning models.

[0391] Thus, in this specification, image segmentation algorithms should be understood in a comprehensive sense to include all forms of algorithms that utilize image data to perform segmentation.

[0392] Referring again to FIG. 7, a segmentation process P2000 according to one embodiment of the present application may include a data acquisition process P2100 and a segmentation process P2200 that utilizes a trained neural network model.

[0393] The segmentation process P2000 may be implemented by an image analysis device 2000 according to one embodiment of the present application.

[0394] Hereinafter, an image segmentation operation using a neural network model of the image analysis device 2000 according to an embodiment of the present application will be described with reference to Figure 12. Figure 12 is a flowchart of an image segmentation method using a neural network model of the image analysis device 2000 according to an embodiment of the present application.

[0395] Referring to FIG. 12, an image segmentation method using a neural network model of an image analysis device 2000 according to one embodiment of the present application may include step S2100 of acquiring target image data, step S2200 of acquiring segmentation information using a trained neural network, and step S2300 of outputting the segmentation information.

[0396] Specifically, in step S2000 of acquiring target image data, the image analysis device 2000 can acquire a target image from the image acquisition device 1000. In addition, the image analysis device 2000 can acquire target object information related to the target image or information related to the scanning conditions under which the target image was captured from the image acquisition device 1000 or any external device.

[0397] At this time, target object information associated with the target image or information related to the scanning conditions under which the target image was captured may be taken into consideration when obtaining the correction parameters, which will be described in detail with reference to Figures 58 to 67.

[0398] Please refer to Fig. 13. Fig. 13 is an exemplary structural diagram of a target image according to one embodiment of the present application.

[0399] For example, the target image data acquired by the image analysis device 2000 according to an embodiment of the present application may include information about the target image TI, such as pixel coordinates, intensity, color, etc.

[0400] As another example, the target image data may include target object information TO. For example, the information on the target object information TO may be information on personal information of a target (e.g., a patient undergoing a brain disease examination) associated with the target image TI. For example, the information on the target object information TO may be information related to the name, age, gender, etc. of the target (e.g., a patient undergoing a brain disease examination). In this case, the image analysis device 2000 may obtain the information on the target object information TO from any external device. Alternatively, the image analysis device 2000 may obtain the information on the target object information TO by recognizing structured metadata for the target image data.

[0401] As another example, the object image data may include information about the object scan conditions TSC. For example, the information about the object scan conditions TSC may be information related to the scan conditions under which the object image TI was captured. For example, the information about the object scan conditions TSC may be information related to the magnetic field strength under which the object image TI was captured, the setting parameters of the imaging device under which the object image TI was captured, or the manufacturer of the imaging device under which the object image TI was captured. In this case, the image analysis device 2000 may obtain the information about the object scan conditions TSC from any external device. Alternatively, the image analysis device 2000 may obtain the information about the object scan conditions TSC by obtaining structured metadata for the object image data.

[0402] Referring again to FIG. 7, the image analysis device 2000 may be implemented to input target image data acquired in the data acquisition process P2100 into an input layer of a trained neural network model.

[0403] In this case, the image analysis device 2000 may be implemented to utilize the node weights and / or node parameters of the artificial neural network model obtained based on the learning process P1000 implemented in the above-mentioned learning device 2200 in the artificial neural network model for segmenting the target image data.

[0404] 12 , in step S2200 of obtaining segmentation information using the trained neural network model, the trained neural network model based on the node weights and / or node parameters obtained from the learning device 2200 may be provided to receive target image data via an input layer, and output a result of labeling the target image TI via an output layer as a result of segmenting the target image TI. In this case, the image analysis device 2000 can obtain segmentation information related to the labeling of the target image via the output layer of the trained neural network model.

[0405] At this time, the results output through the output layer of the artificial neural network model may include a plurality of target regions obtained from the target image of the target image data.

[0406] The results output through the output layer of the neural network model may be in the form of labeling corresponding to multiple target regions obtained from the target image. For example, the results output through the output layer of the neural network model may be in the form of label data including a first label defining a first region and a second label defining a second region obtained from the target image.

[0407] In this case, the result output through the output layer of the neural network model may be in the form of overlaying a first color on a first region of the target image based on the first label and a second color on a second region of the target image, thereby making it easier to distinguish between the first and second regions. However, the above content is merely an example, and the output result may be configured in any form for distinguishing between the first and second regions.

[0408] In addition, the segmentation information output through the output layer of the neural network model may be in the form of dividing the target image into a plurality of regions based on a predetermined brain atlas. For example, as described above, the learning device may train the neural network model to divide the training image into a first region corresponding to a first brain region and a second region corresponding to a second brain region based on the predetermined brain atlas. In this case, since the neural network model trained based on the predetermined brain atlas is used, the segmentation information output through the output layer of the neural network model may be output in the form of dividing the target image into a plurality of regions including the first region corresponding to the first brain region of the target image and the second region in the second brain region.

[0409] The image analysis device 2000 can perform an operation of calculating morphological values ​​corresponding to a specific region based on the segmentation information output through the output layer.

[0410] For example, the image analysis device 2000 may be implemented to calculate a morphological value indicative of a morphological character associated with the first brain region based on the segmentation information associated with the first region output via the output layer, for example, the morphological character may be related to volume, thickness, length, shape, etc.

[0411] In step S2300 of outputting the segmentation information, the image analysis device 2000 may be implemented to overlay visual graphics for multiple brain regions based on the segmentation information output via the output layer and display them to the user via the output module 2650 of the output device 2600 or the output module 2050 of the image analysis device 2000.

[0412] Please refer to Figure 14. Figure 14 is an illustration of an image obtained based on the segmentation information obtained by the segmentation process of the image analysis device 2000 according to one embodiment of the present application.

[0413] Specifically, the upper diagram of Fig. 14 shows an example of an image output based on the segmentation result for an image acquired from T1-MRI, and the lower diagram of Fig. 14 shows an example of an image output based on the segmentation result for an image acquired from T2-Flair MRI.

[0414] For example, the image analysis device 2000 according to one embodiment of the present application can obtain segmentation information corresponding to the frontal lobe, temporal lobe, parietal lobe, occipital lobe, lateral ventricle, amygdala, hippocampus region, etc. through a segmentation process on an image obtained from T1-MRI.

[0415] For example, the image analysis device 2000 according to one embodiment of the present application can obtain segmentation information corresponding to white matter, gray matter, ventricles, WMH (white matter hyperintensity) regions, etc. through a segmentation process on images obtained from T2-Flair MRI.

[0416] According to the image segmentation operation of the image analysis device 2000 described above, a user can visually check the segmentation result and easily confirm the segmentation result, and an advantageous effect is provided in that the user's understanding of auxiliary indicators for brain disease diagnosis can be improved.

[0417] On the other hand, the artificial neural network model used in FIG. 7 may be realized by at least one artificial neural network model.

[0418] A flow chart of a process for image segmentation using at least one artificial neural network model according to an embodiment of the present disclosure will be described below with reference to Figures 15 through 17. Specifically, the following description will focus on the characteristic features of using multiple artificial neural network models in connection with the process P1200 for training an artificial neural network model of Figure 7. The details described with reference to Figures 7 through 14 can also be applied by analogy to the embodiments described below with reference to Figures 15 through 17.

[0419] FIG. 15 is a diagram illustrating the process flow for image segmentation according to this embodiment.

[0420] As an example, referring to FIG. 15, the neural network model training process P1200 of the training process P1000 of FIG. 7 may be comprised of a process P1210 for training a first neural network model and a process P1220 for training a second neural network model.

[0421] For example, a first neural network model may be trained to acquire a first region corresponding to a first brain region and a second region corresponding to a second brain region. In this case, a second neural network model may be trained to acquire a third region and a fourth region included in the first region that can be acquired from the first neural network model. In other words, the first neural network model may be trained to acquire regions corresponding to macroscopic structures of the brain (e.g., a skull region, a cerebrospinal fluid (CSF) region, a cortex region, and a medulla region), and the second neural network model may be trained to acquire regions corresponding to detailed structures of the brain (e.g., a region corresponding to brain elements located in the cortex, a region corresponding to brain elements located in the medulla).

[0422] As another example, a first neural network model may be trained to obtain regions corresponding to brain regions corresponding to a first brain atlas, while a second neural network model may be trained to obtain regions corresponding to brain regions corresponding to a second brain atlas.

[0423] As another example, the first neural network model may be trained to acquire regions corresponding to macroscopic structures of the brain (e.g., skull region, cerebrospinal fluid (CSF) region, cortex region, and medulla region). Conversely, the second neural network model may be trained to acquire regions corresponding to brain regions corresponding to a brain atlas. For example, the second neural network model may be trained to acquire regions corresponding to multiple brain regions based on the Desikan-Killiany Atlas (Desikan 2006), and the second neural network model trained based on the Desikan-Killiany Atlas (Desikan 2006) may be trained to acquire regions including the frontal lobe, temporal lobe, parietal lobe, occipital lobe, lateral ventricle, amygdala, and hippocampus.

[0424] The process P1210 for training the first neural network model and the process P1220 for training the second neural network model may be performed independently.

[0425] Specifically, the data sets used in the process P1210 for training the first neural network model and the process P1220 for training the second neural network model may be independent.

[0426] Please refer to Fig. 16. Fig. 16 is an exemplary structural diagram of an image dataset according to one embodiment of the present application. Specifically, the image dataset shown on the left side of Fig. 16 may be an image dataset for training a first neural network model. Conversely, the image dataset shown on the right side of Fig. 16 may be an image dataset for training a second neural network model.

[0427] For example, the image data set DS for training the first neural network model may include first image data including information associated with a 1a label L1A associated with a first image I1 and a first region, whereas the image data set DS for training the second neural network model may include first image data including information associated with a 1b label L1b associated with the first image I1 and a second region.

[0428] In this case, the learning device may use a first neural network model to input the first image data ID1 to an input layer of the artificial neural network and obtain output label data corresponding to the 1a label L1a' associated with the first region. The learning device may train the artificial neural network model based on the label data corresponding to the label L1A associated with the first region included in the first image data ID1 and the output label data corresponding to the 1a' label L1a' associated with the first region. For example, the learning device 2200 may be configured to adjust the weight or parameters of at least one node included in the hidden layer of the neural network model based on the difference between the 1a label L1A and the 1a' label L1a'. Furthermore, the parameters associated with the first neural network model associated with the first region may be obtained by repeatedly performing the above-described process of training the first neural network model.

[0429] On the other hand, the learning device can use a second neural network model to input the first image data to an input layer of the artificial neural network and obtain output label data corresponding to the 1b' label L1b' associated with the second region. In this case, the learning device can train the artificial neural network model based on the label data corresponding to the 1b label L1b associated with the second region included in the first image data ID1 and the output label data corresponding to the 1b label L1b' associated with the second region. For example, the learning device 2200 can be implemented to adjust the weight or parameters of at least one node included in the hidden layer of the neural network model based on the difference between the 1b label L1b and the 1b' label L1b'. Furthermore, the parameters associated with the second neural network model associated with the second region can be obtained by repeatedly performing the above-described process of training the second neural network model.

[0430] In addition, parameters associated with the first neural network model obtained in the learning process P1000 may be utilized in a first neural network model for image segmentation corresponding to a first region in the segmentation process P2000 (P2210), and parameters associated with the second neural network model obtained in the learning process P1000 may be utilized in a second neural network model for image segmentation corresponding to a second region in the segmentation process P2000 (P2220).

[0431] 16, the label 1a included in the image data related to the first neural network model is a label related to the first region, and the label 1b included in the image data related to the second neural network model is a label related to the second region, but this is merely an example and is not limiting. For example, the image data of the first and second neural network models may include label information related to the first and second regions, and only the label information related to the first region may be used when training the first neural network model, and only the label information related to the second region may be used when training the second neural network model.

[0432] Furthermore, although the first neural network model and the second neural network model have been described as being independent from each other in the above description, the present invention is not limited to this and the first neural network model and the second neural network model may share at least some layers. In other words, the first neural network model and the second neural network model may include at least one common layer.

[0433] Meanwhile, the process of training the first neural network model and the process of training the second neural network model may be independent, or may be trained in a correlated manner. Here, the term "training in a correlated manner" may refer to using output data from one of the two neural network models as input data for the other neural network model, and may include any form of using any data generated by one of the two neural network models in the other neural network model.

[0434] Please refer to Figure 17. Figure 17 is a diagram illustrating a process flow for image segmentation according to an embodiment of the present application.

[0435] As an example, the process P1200 for training a neural network model of the learning process P1000 according to an embodiment of the present application may include a process (P1211) for training a first neural network model and a process (P1221) for training a second neural network model, and may be implemented such that data related to the results output from the first neural network model is input as input data to the second neural network model.

[0436] For example, the learning device 2200 may be configured to allow the second neural network model to receive label data associated with the first region as an input. In this case, the label data associated with the first region may be obtained by manually labeling the first region corresponding to the first brain region in the image data, or by using any automatic labeling software.

[0437] For example, the label data associated with the first region may be data manually labeled by a clinician for image data that is input into the first neural network model.

[0438] In another example, a first neural network model can be trained to output label data associated with a first region, and the label data associated with the first region input to a second neural network model can then be the label data output from the first neural network model.

[0439] The second neural network model can then be trained to output label data associated with the second region based on the label data associated with the first region and the image data.

[0440] Therefore, the learning device 2200 according to an embodiment of the present application may be implemented to acquire label data related to a first region through a first neural network model and acquire label data related to a second region through a second neural network model, and thus the learning device 2200 can train the first neural network model and the second neural network model to acquire label data related to the first region and the second region.

[0441] The learning device 2200 may also be provided to adjust weights or parameters of at least one node of the first neural network model or adjust weights or parameters of at least one node of the second neural network model based on a difference between label data related to labels associated with the first and second regions and label data related to the first and second regions of the image dataset.

[0442] Additionally, the parameters associated with the first and second neural network models obtained in the training process P1000 may be utilized in the first and second neural network models for image segmentation corresponding to the first and second regions in the segmentation process P2000.

[0443] In this case, the first or second region may be a region associated with the anatomical structure of the brain, or may be one of the regions demarcated based on the above-mentioned brain atlas, or may be a region showing a meaningful association with a brain disease.

[0444] However, the above-described neural network model training process is merely illustrative, and various combinations and connections of at least one or more neural network models having any suitable form, type, and parameters may be used to train image segmentation operations, and the neural network models may be verified to obtain the parameters of the neural network models.

[0445] Meanwhile, the image analysis device 2000 according to one embodiment of the present application may be implemented to refresh or update the trained artificial neural network model in the segmentation process P2000.

[0446] For example, the image analysis device 2000 may be implemented to allow the segmentation information obtained by segmenting a target image using the segmentation process P2200 using a trained neural network model to be modified manually or using any software. In this case, the image analysis device 2000 may be implemented to update or renew the artificial neural network model by modifying the weights of at least one or more nodes or the parameters of at least one or more nodes of the trained neural network model based on the modified segmentation information.

[0447] According to an embodiment, an image analyzer may perform a function of determining image quality. The image analyzer may have a quality standard for determining image quality and determine whether an input image satisfies the quality standard.

[0448] More specifically, the image quality assessment may refer to determining whether a medical image acquired by an image acquisition device has a quality equal to or greater than a certain level. In other words, the image quality assessment may refer to determining whether medical information with a reliability equal to or greater than a certain level can be acquired from the medical image acquired by the image acquisition device.

[0449] The image quality assessment may be performed in conjunction with the sub-operations that make up the image analysis operation, for example, the image quality assessment operation may be performed in conjunction with the image acquisition operation, image pre-processing operation, image segmentation operation, and / or medical information output operation described herein.

[0450] The image quality determination operation may be performed before or after the execution of at least one detail operation. The image quality determination operation may be performed based on information obtained from the execution of the detail operation. The image quality determination operation may be performed to determine whether a criterion for executing the detail operation is met.

[0451] The image quality assessment may be performed based on image data. The image quality assessment may be performed based on raw image data. The image quality assessment may be performed based on pre-processed medical images. As another example, the image quality assessment may be performed based on the results of segmenting the medical images. As another example, the image quality assessment may be performed based on the results of analyzing the medical images.

[0452] The image quality determination may be performed based on non-image data associated with the image data, including metadata information about the medical image, artifact information included in the medical image, and / or region of interest information obtained through medical image segmentation.

[0453] According to one embodiment, the image quality determination may be performed separately for each corresponding sub-operation.

[0454] For example, the first quality judgment may be performed based on raw data before a preprocessing operation. In this case, the preprocessing operation may be performed on images that satisfy the first quality criterion. Also, for example, the second quality judgment may be performed based on raw data or preprocessed images before a segmentation operation. In this case, the segmentation operation may be performed on images that satisfy the second quality criterion. Also, for example, the third quality judgment may be performed based on raw data, preprocessed images, or image segmentation results before image analysis. Image analysis may be performed on images that satisfy the third quality criterion. The first, second, and third quality criteria may be different from each other.

[0455] Information acquired through the image quality determination may be output. The information acquired through the image quality determination may include information serving as a basis for the image quality determination. The information serving as a basis for the image quality determination may include whether a defect exists in the image, image type information, image acquisition conditions, etc. The information acquired through the image quality determination may be used to generate or provide information related to whether a subsequent operation is to be performed.

[0456] A specific example of image quality determination will be described later.

[0457] The image analysis device 2000 according to an embodiment of the present application may be provided to perform an operation of calculating a brain morphological index of a subject based on a segmentation result of the subject image.

[0458] Specifically, the image analysis device 2000 can obtain an internal region of the skull and a region corresponding to the target element based on the segmentation result, and can additionally perform an operation of correcting the boundary corresponding to the internal region of the skull to calculate the morphological index of the target element.

[0459] Additionally, the image analyzer 2000 can perform additional operations to align the brain image to correct for the boundaries of the internal region of the skull.

[0460] The operation of calculating the subject's brain morphological indexes in the image analysis device 2000 according to an embodiment of the present application will be described in detail below with reference to FIGS.

[0461] The image analysis device 2000 according to an embodiment of the present application may be provided to perform an operation of correcting the morphological values ​​or morphological indices of the brain calculated based on the segmentation results of the target image.

[0462] Specifically, the image analysis device 2000 may acquire a region corresponding to the target element and pixel or voxel data corresponding to the target element based on the segmentation result, and may acquire a morphological value of the target element based on the pixel or voxel data corresponding to the target element.

[0463] In addition, the image analysis device 2000 can perform an operation to correct the morphological values ​​of the target elements taking into account the scanning conditions under which the target image was acquired and the position of the target elements within the target image in order to output more accurate morphological indicators.

[0464] In this case, the image analysis device 2000 may be provided to acquire correction parameters taking into account the scanning conditions and the position of the target element from the correction parameter acquisition device 2400 in order to correct the morphological value of the target element, and to perform an operation of outputting a morphological index of the target element based on the correction parameters and the morphological value of the target element.

[0465] The operation of the image analysis device 2000 according to one embodiment of the present application for correcting the brain morphological values ​​or morphological indices will be described in detail below with reference to Figures 58-67.

[0466] An image output device according to an embodiment may provide a user with auxiliary diagnostic information based on various medical information acquired through image analysis. Here, the auxiliary diagnostic information may include information acquired through processing various medical information acquired from medical images. For example, the auxiliary diagnostic information may include information acquired by processing medical information, such as diagnosis information, analysis information, and prescription information based on the medical information.

[0467] The auxiliary diagnostic information providing operation may be performed together with the sub-operations constituting the image output operation. For example, the auxiliary diagnostic information providing operation may be performed together with the image acquiring operation, the operation of acquiring medical information from the image, the operation of acquiring auxiliary diagnostic information from the acquired information, the operation of outputting the auxiliary diagnostic information, and / or the operation of providing a comment based on the auxiliary diagnostic information, as described herein. Each of the above operations will be described in detail below.

[0468] According to another embodiment, an image output device may selectively provide necessary index information to a user from various medical information acquired through image analysis, where selective information provision may include selectively providing only necessary medical information to a user from various medical information that can be acquired through the image analysis device.

[0469] The selective information providing operation may be performed together with the sub-operations constituting the image outputting operation. For example, the selective information providing operation may be performed together with the image acquiring operation, the operation of acquiring medical information from the image, the operation of acquiring selective information from the acquired information, the operation of outputting the selective information, and / or the operation of providing a comment based on the selective information, as described herein. Each of the above operations will be described in detail below.

[0470] The configuration and operation of the image analysis device 2000 according to one embodiment of the present application have been described above. The image analysis method according to this embodiment will now be described in more detail.

[0471] In the following description, the image analysis method according to an embodiment of the present application will be described as being executed by the image analysis device 2000, the learning device 2200, the correction parameter acquisition device 2400, or the output device 2600. However, this is merely for the convenience of description, and the image analysis method according to an embodiment of the present application is not limited to the image analysis device 2000, the learning device 2200, the correction parameter acquisition device 2400, or the output device 2600. In other words, the image analysis method described below is not necessarily executed only by the image analysis device 2000, the learning device 2200, the correction parameter acquisition device 2400, or the output device 2600, but can also be executed by other systems or devices having functions similar to the image analysis device 2000, the learning device 2200, the correction parameter acquisition device 2400, or the output device 2600.

[0472] Various information is obtained from medical images to determine a patient's health condition, but to obtain accurate information, analysis must be based on medical images that meet a certain level of quality and meet the necessary conditions.

[0473] More specifically, medical images acquired by an image acquisition device may have a quality (or characteristics) that is not suitable for image analysis. For example, the image may have formal or substantial defects that make it inappropriate for medical data acquisition. For example, medical images may not have a certain level of quality due to various defects in the image caused by image brightness, image resolution, the part of the body captured, the direction in which the image was captured, the angle at which the image was captured, or other problems that may occur during capture. Or, the format of the image, such as the format and size of the image file, may not meet required conditions.

[0474] When an analysis is performed based on a medical image that does not have a certain level of quality, the analysis results may also not have a certain level of reliability. Therefore, in order to derive an analysis result that can have a higher level of reliability, an image quality judgment must be performed to determine whether the image acquired by the image acquisition device has a certain level of quality.

[0475] In the past, inspectors would directly judge each medical image captured via imaging equipment to determine whether there were any defects. However, there were limitations in that it was difficult to make a consistent, highly accurate quality judgment, as the results could vary depending on the inspector's perspective, experience, and physical condition.

[0476] According to one embodiment, artificial intelligence can be utilized to overcome the above limitations and improve the quality assessment of captured medical images. That is, an image analysis device according to one embodiment can utilize artificial intelligence to perform high-level medical image quality assessment.

[0477] FIG. 18 is a diagram illustrating an image quality determination process according to one embodiment.

[0478] 18, an image quality assessment process according to one embodiment may include a medical image acquisition step S3000, a medical image quality assessment step S3300, and a medical image quality-related information output step S3500. In this case, the medical image quality-related information output step S3500 may be omitted. Meanwhile, the image quality assessment process may further include a medical image pre-processing step and / or a medical image segmentation step.

[0479] FIG. 19 is a diagram illustrating an image analysis device 3000 that performs an image quality assessment process.

[0480] 19, an image analysis device 3000 can execute an image quality assessment process. The image analysis device 3000 can include one or more modules for executing image quality assessment. For example, the image analysis device 3000 can include at least one of a pre-processing execution module 3300, an image segmentation module 3500, an image quality assessment module 3700, and an image quality related information output module 3900.

[0481] Each step of the image quality determination process will be described below with more specific examples. The image quality determination process described below may be performed by the image analysis device 3000 described above or by any device or system described throughout this specification.

[0482] An image acquisition device according to an embodiment may acquire medical images. Exemplarily, medical images include, but are not limited to, images from computed tomography (CT), magnetic resonance imaging (MRI), and X-ray. In this case, the MRI images may include various types of images that can be acquired by an MRI imaging device, such as T1-weighted images, T2-weighted images, and FLAIR images. In addition, the MRI images may include images acquired in various types of planes that can be acquired by an MRI imaging device, such as axial, sagittal, or coronal planes.

[0483] According to some embodiments, an image capture device may capture medical images in which at least a portion of the area of ​​interest is not fully captured, e.g., the image capture device may capture medical images in which at least a portion of the area of ​​interest is omitted or improperly captured.

[0484] For example, an image acquisition device may acquire medical images having a quality below a certain level. An image acquisition device may acquire medical images from which it is difficult to expect extraction of medical information with a certain level of reliability or above. As another example, an image acquisition device may acquire medical images having an abnormal file structure or missing patient information. As yet another example, an image acquisition device may acquire medical images containing at least one noise. Here, noise refers to various types of defects or conditions that affect information acquisition based on the image.

[0485] An image quality assessment process according to one embodiment may include pre-processing a medical image, for example, pre-processing a medical image acquired by an image acquisition device to make it suitable for performing an image quality assessment.

[0486] Illustratively, the image preprocessing step may include performing various preprocessing steps, such as correcting the brightness, size, ratio, orientation, or resolution of the image to facilitate detection of artifacts contained in the medical image. As another example, the image preprocessing step may include performing various preprocessing steps, such as correcting the brightness, size, ratio, orientation, or resolution of the image to facilitate acquisition of information about anatomical structures contained in the medical image.

[0487] In addition, the image pre-processing step may include performing various pre-processing processes for determining image quality, and since details regarding this have been described above, duplicate content will be omitted.

[0488] An image quality assessment process according to one embodiment may include a step of segmenting a medical image. For example, an image quality assessment process according to one embodiment may include a step of segmenting a medical image acquired by an image acquisition device for image quality assessment.

[0489] Illustratively, the image segmentation step can include performing medical image segmentation to obtain information related to artifacts contained within the medical image. As a more specific example, the image segmentation step can include performing segmentation of regions corresponding to artifacts contained within the medical image.

[0490] Segmenting the regions corresponding to the artifacts may be performed using a neural network model trained to obtain regions corresponding to artifacts contained in the medical images. Segmenting the artifact regions may be performed using a neural network model trained using training data including one or more medical images in which artifact regions have been labeled.

[0491] As another example, the image segmentation step may include segmenting the medical image such that an image quality determination is based at least in part on regions of a human body contained within the medical image. As a more specific example, the image segmentation step may include segmenting the medical image to obtain information about anatomical or functional structures of a human body that may serve as a basis for the image quality determination. The image segmentation step may include segmenting regions corresponding to structures of a human body contained within the medical image. The image segmentation step may include obtaining regions corresponding to structures utilized in the image quality determination.

[0492] The segmentation of regions corresponding to human body structures may be performed using a neural network model. The segmentation of regions corresponding to structures may be performed using a neural network model trained to segment regions included in a medical image. The image segmentation step may include performing segmentation of the medical image using a pre-trained neural network model to obtain at least one segmented region. In this case, the at least one segmented region may correspond to another anatomical or functional structure. Furthermore, the at least one segmented region may include a region corresponding to a human body structure used for quality assessment. The content related to image segmentation described throughout this specification may also be applied to the segmentation of regions corresponding to anatomical or functional structures.

[0493] FIG. 20 is a diagram for explaining the image quality determination module 3700. As shown in FIG.

[0494] Referring to FIG. 20, the image quality determination module 3700 may include at least one of a first image quality determination unit 3710, a second image quality determination unit 3730, a third image quality determination unit 3750, or a fourth image quality determination unit 3770.

[0495] Exemplarily, the first image quality determination unit 3710 may determine the quality of the image based on metadata information, the second image quality determination unit 3730 may determine the quality of the image based on noise information, the third image quality determination unit 3750 may determine the quality of the image based on anatomical segmentation information, and the fourth image quality determination unit 3770 may determine the quality of the image based on composite information, for example, the relationship between noise information and anatomical segmentation information. The first to fourth image quality determination units 3710 to 3770 will be described in detail below.

[0496] According to an embodiment, the neural network model for performing image quality judgment may be trained and executed differently depending on the type of medical image acquired by the image acquisition device. For example, if the type of medical image acquired by the image acquisition device is a CT image, the image quality judgment model of the image analysis device 3000 may be trained and executed based on the CT image. As another example, if the type of medical image acquired by the image acquisition device is an MRI, the image quality judgment model of the image analysis device 3000 may be trained and executed based on the MRI image.

[0497] According to an embodiment, the first image quality determining unit 3710 may determine whether image analysis is successful based on metadata information. The metadata information may include information about the medical image acquired from the image acquisition device. More specifically, the metadata information may include at least one of file structure information of the medical image or patient information entered in the medical image.

[0498] FIG. 21 is a diagram for explaining the first image quality determination process.

[0499] Referring to FIG. 21, the first image quality judgment process may include a medical image acquisition step S3711, a non-image information acquisition step S3713, a non-image information normality judgment step S3715, and a non-image related information output step S3717.

[0500] The step of acquiring non-image information S3713 may include acquiring medical image information from the medical image, where the non-image information may include file structure information of the medical image or patient information entered in the medical image.

[0501] In this case, the file structure information of the medical image may include, but is not limited to, information about the file type, format, or file size of the medical image, and the patient information may include, but is not limited to, information related to personal information such as the patient's name and age, information related to the time when the patient's medical image was taken, or information about the patient's health condition.

[0502] The step S3715 of determining whether non-image information is normal may include performing an image quality determination based on non-image information extracted from the medical image. The step S3715 of determining whether non-image information is normal may include performing an image quality determination based on at least one of file structure information and patient information from the medical image.

[0503] According to one embodiment, the step S3715 for determining whether non-image information is normal may determine whether the file structure of the medical image is abnormal, which may refer to, but is not limited to, the file type, format, or file size of the medical image.

[0504] For example, since the file format of the medical image needs to be a file format suitable for image analysis performed by the image analysis device 3000, the step S3715 of determining whether the non-image information is normal may include determining whether the file format of the medical image is a file format suitable for image analysis performed by the image analysis device 3000. Thus, the step S3715 of determining whether the non-image information is normal may include obtaining information about whether the image analysis is performed normally based on file structure information of the medical image.

[0505] According to another embodiment, the step S3715 of determining whether non-image information is normal may include determining whether patient information is missing from the medical image acquired from the image acquisition device, where the patient information may include, but is not limited to, personal information such as the patient's name and age, or information regarding the patient's health condition.

[0506] For example, if the medical image does not contain patient information, it is unclear which patient the image analysis result by the image analyzer 3000 relates to, so the step S3715 of determining whether the non-image information is normal may include determining whether the medical image contains patient information. Thus, the step S3715 of determining whether the non-image information is normal may include obtaining information regarding whether patient information has been entered into the medical image.

[0507] FIG. 22 is a diagram illustrating the first image quality determining unit 3710. As shown in FIG.

[0508] Referring to FIG. 22, the first image quality determining unit 3710 may include at least one of a non-image information acquiring unit 3711 or a non-image information normality determining unit 3717 .

[0509] The non-image information acquisition unit 3711 may include a file structure information acquisition unit 3713 or a patient information acquisition unit 3715. The non-image information acquisition unit 3711 may acquire non-image information based on an acquired medical image. In this case, the acquired medical image may include raw data before a preprocessing operation or an image on which a preprocessing operation has been performed.

[0510] The non-image information normality determination unit 3717 may determine whether the non-image information is normal based on the non-image information acquired by the non-image information acquisition unit 3711. The non-image information normality determination unit 3717 may output information regarding the result of determining whether the non-image information is normal. Exemplarily, the non-image information normality determination unit 3717 may output information regarding whether image analysis is performed normally based on file structure information of the medical image or information regarding whether patient information is entered in the medical image.

[0511] According to an embodiment, the second image quality determining unit 3730 may perform a function of extracting noise information contained in an image. Here, the noise information may refer to various types of defects or conditions that affect information acquisition based on the image. For example, the noise information may include information related to image resolution, information related to image brightness, or information about artifacts occurring in the image. In addition, the noise information may refer to various image defects created as a result of inappropriate image sampling.

[0512] FIG. 23 is a diagram for explaining the second image quality determination process.

[0513] Referring to FIG. 23, the second image quality judgment process may include a medical image acquisition step S3731, a medical image artifact occurrence determination step S3733, an artifact information acquisition step S3735 from the medical image, and an artifact information output step S3737.

[0514] The step S3733 of determining whether artifacts are present in the medical image may include determining whether artifacts are present in the medical image. Here, artifacts may include noise associated with pixels that do not faithfully represent anatomical structures in the medical image. Exemplarily, artifacts may include, but are not limited to, motion artifacts, bias artifacts, zipper artifacts, ghost artifacts, spike artifacts, and the like, and may include various types of artifacts that are already known.

[0515] The step S3735 of acquiring artifact information from the medical image may include acquiring information about artifacts occurring in the medical image, where the artifact information may include information related to the artifact, such as information about whether or not an artifact has occurred, information about the location of the artifact, information about the type of artifact that has occurred, information about the extent of the artifact, etc.

[0516] According to one embodiment, step S3733 of determining whether artifacts have occurred in the medical image or step S3735 of acquiring artifact information from the medical image may include being performed using a trained neural network model. Step S3733 of determining whether artifacts have occurred in the medical image may include determining whether artifacts have occurred in the medical image using a trained neural network model. Although step S3733 of determining whether artifacts have occurred in the medical image and step S3735 of acquiring artifact information from the medical image are depicted separately in the drawings, each step may include acquiring whether artifacts have occurred and information related to the artifacts using a single neural network model.

[0517] The neural network model can determine whether an artifact occurs based on the medical image. The neural network model can determine whether an artifact exists in a specific region of the acquired medical image based on the input medical image.

[0518] The neural network model can also acquire artifact information based on the medical image. Based on the input medical image, the neural network model can acquire information regarding the presence or absence of artifacts, the location of artifacts, the type of artifacts that have occurred, and the degree of artifact occurrence.

[0519] The artifact information may be acquired based on a medical image acquired by an image acquisition device. Exemplarily, the artifact information may be acquired using a neural network model trained to acquire the artifact information based on the medical image.

[0520] To obtain artifact information, a sorter algorithm can be used to classify or predict an input image with respect to one or more labels. Various types of algorithms can be used for classification or prediction, such as k-nearest neighbor, support vector machine, artificial neural network, decision tree, self-organizing map, logistic regression, etc.

[0521] Artificial neural networks can be sorters, hybrid classifiers, ensemble classifiers, linear regression neural networks, etc. Artificial neural networks can include convolutional neural networks (CNNs). Artificial neural networks can be supervised learning, unsupervised learning, or reinforcement learning models.

[0522] FIG. 25 is a diagram illustrating a neural network model for acquiring artifact information according to an embodiment. Referring to FIG. 25, the neural network model according to an embodiment may be prepared in a sorter form. According to an embodiment, the neural network model may include an input layer (IL), a pooling layer (PL), a convolutional neural network layer (CL), a fully connected layer (FCL), a hidden layer (HL), and an output layer (OL), and may acquire a feature vector based on an input image. The neural network model may be prepared in a sorter form that classifies the input image into one or more labels. Alternatively, the neural network model may be prepared in the form of a regression model. The neural network model may be prepared as a regression model that acquires a linear output value for specific artifact information based on the input image.

[0523] The neural network model may receive input medical images acquired by photographing a specific region of the human body and obtain output information. The output information may indicate whether the input image contains a target object. For example, the output layer of the neural network model may include output nodes to which a probability function is assigned. The output layer of the neural network model may include output nodes to which a probability function indicating whether the target image contains the target object is assigned. The output layer may include output nodes to which, for one or more target objects, one or more probability functions indicating whether the input image contains each target object are assigned.

[0524] The neural network model can be trained to acquire artifact information. The neural network model can be trained to acquire artifact information based on training data including one or more medical images labeled with artifact information. The neural network model can be trained to acquire artifact regions based on training data including one or more medical images labeled with artifact regions.

[0525] The artifact training data may include a plurality of medical images. The artifact training data may include medical images captured using various methods, such as CT, MRI, or X-ray images. The artifact training data may include a plurality of medical images containing artifacts of various types, ranges, sizes, shapes, or locations.

[0526] The artifact training data may include medical images to which artifact labels indicating whether or not artifacts occur are assigned. The artifact training data may include medical images acquired by photographing various parts of the human body and to which artifact labels indicating whether or not artifacts occur are assigned. Here, the artifact labels may be assigned differently depending on the location of the artifact, the degree of the artifact, the form of the artifact, or the type of the artifact.

[0527] The artifact training data may include medical images masked for artifact occurrence regions. The artifact training data may include medical images masked (or labeled) for one or more artifact occurrence regions. The artifact training data may include medical images masked (or labeled) differently for multiple types of artifact occurrence regions. For example, the artifact training data may include medical images displayed in a first color for occurrence of a first type of artifact and in a second color for occurrence of a second type of artifact.

[0528] The neural network model can be trained using the artifact training data described above. The neural network model can be trained using supervised, unsupervised, or reinforcement learning to acquire artifact information based on medical images using the artifact training data. The neural network model can be trained using backpropagation.

[0529] The neural network model can be trained to classify medical images according to whether they contain artifacts using artifact training data including medical images labeled for the presence or absence of artifacts. The neural network model can be trained to classify medical images according to the type of artifact contained in the medical image. The neural network model can be trained to acquire whether a target image contains each type of artifact for multiple artifacts through artifact training data including medical images labeled for the presence or absence of multiple types of artifacts.

[0530] The neural network model can be trained to acquire artifact region information through artifact training data. The neural network model can be trained to detect artifact-occurring regions from a target image using artifact training data including medical images in which artifact-occurring regions are masked for one or more types of artifacts. The neural network model can be trained to acquire artifact region information for each of a plurality of artifacts included in a medical image, for the region and / or type in which each artifact is distributed, through artifact training data including medical images labeled with a plurality of types of artifact-occurring regions.

[0531] The neural network model can also be trained using multiple artifact training data: first artifact training data including medical images with a first type of artifact and second artifact training data including medical images with a second type of artifact.

[0532] Alternatively, multiple neural network models may be trained and utilized, with a first neural network model trained based on first artifact training data and a second neural network model trained based on second artifact training data, which may differ at least in part from the first artifact training data.

[0533] 24 is a diagram for exemplarily explaining that the second image quality assessment process determines the location of artifacts in a medical image using an artificial neural network model. Referring to FIG. 24, the neural network model can acquire artifact region information regarding the location of artifacts using a medical image acquired through an image acquisition device as an input.

[0534] Referring to FIG. 24, the step S3735 of obtaining artifact information from the medical image may include obtaining artifact region information within the medical image.

[0535] The artifact region information may be provided in the form of a display, for example, a saliency map in the form of a heat map, to indicate areas in the medical image where artifacts are likely to be located. The artifact region information may be a feature map obtained from a neural network model that determines the presence or absence of artifacts or a neural network model that obtains artifact information.

[0536] The artifact region may be a region on the feature map obtained from the trained neural network model and associated with a target artifact present in the target medical image, where the correlation with the target artifact is equal to or greater than a reference value based on the feature map. The artifact region may include a first region and a second region located within the first region. The first region may be a region on the feature map where the correlation with the target artifact is equal to or greater than a first reference value. The second region may be a region on the feature map where the correlation with the target artifact is equal to or greater than a second reference value that is greater than the first reference value.

[0537] 24(a) to (c), step S3735 of acquiring artifact information from a medical image may include acquiring a first original image (a), a second original image (b), and a third original image (c) including artifacts. In this case, the first original image (a) to the third original image (c) may be images captured on different planes. The first original image (a) to the third original image (c) may be images captured at the same time point for the same subject.

[0538] 24(d) to (f), step S3735 of acquiring artifact information from a medical image may include acquiring a first artifact image (d), a second artifact image (e), and a third artifact image (f) representing the location of an artifact. Here, the first artifact image (d) may be an image in which artifact area information acquired based on the first original image (a) is displayed on the first original image (a). The second artifact image (e) may be an image in which artifact area information acquired based on the second original image (b) is displayed on the second original image (b). The third artifact image (f) may be an image in which artifact area information acquired based on the third original image (c) is displayed on the third original image (c).

[0539] 24(d) to 24(f), the first to third artifact images may be images in which the first to fourth regions A1 to A4 are displayed in descending order of the likelihood that an artifact will be located in the medical image. Here, the first to fourth regions A1 to A4 may correspond to the region in which the artifact in the image is located. The first to fourth regions A1 to A4 may include at least a portion of the region in which the artifact in the image is located. The first region A1 may be included in the second region A2. The second region A2 may be included in the third region A3. The third region A3 may be included in the fourth region A4.

[0540] Although not shown in the drawings, the artifact region information may include a bounding box for displaying the region where the artifact is located in the medical image. The artifact region information may include coordinate information, pixel information, etc., where the artifact is located in the medical image. Exemplarily, the second image quality determiner 3730 may use an artificial neural network model, such as a backpropagation-based method (BBM), an activation-based method (ABM), or a pertubation-based method (PBM), to acquire the artifact occurrence location information. Examples of BBM methods include layer-wise relevance propagation (LRP), DeepLIFT, SmoothGrad, and VarGrad. Examples of ABM methods include class activation map (CAM) and gradient-class activation map (Grad-CAM). Examples of PBM methods include local interpretable model-agnostic explanation (LIME).

[0541] According to an embodiment, the third image quality determining unit 3750 may perform quality determination based on segmentation information obtained by anatomically segmenting a medical image acquired from an image acquisition device, by considering whether the segmentation information satisfies a quantitative criterion. More specifically, the third image quality determining unit 3750 may determine whether at least a portion of the anatomical structures included in the medical image satisfies a predetermined quantitative criterion.

[0542] The third image quality determination unit 3750 may determine whether a value associated with a predetermined region included in the image and corresponding to one part of the human body satisfies a criterion. The third image quality determination unit 3750 may determine whether at least a portion of the anatomical structure of the human body captured through the image capturing device is completely captured.

[0543] For example, when a specific part of the human body is photographed using an image capturing device and then analyzed, disease can be diagnosed based on morphological indices of at least some of the photographed regions of the human body. In this case, if each region of the human body is not photographed completely, the disease diagnosis result cannot have a certain level of reliability. Therefore, since each region of the human body should be photographed completely, the third image quality determining unit 3750 can perform a function of determining whether each region of the human body has been photographed completely by considering whether morphological indices based on the anatomical regions of the human body included in the medical image satisfy a generally required standard value.

[0544] As a more specific example, when the image analyzer 3000 analyzes an MRI image of the brain, it can calculate the ratio of volume values ​​for each anatomical region of the brain to determine whether a disease exists. If the volume value of a specific region of the brain does not meet a generally required standard, the disease diagnosis result cannot have a certain level of reliability. If a specific region of the brain is not fully imaged, the disease diagnosis result may be inaccurate if it is based on the volume value of the region. Therefore, the third image quality determiner 3750 can determine whether each region of the brain has been fully imaged by considering whether the morphological indicators for each region of the brain included in the MRI image meet a generally required standard.

[0545] FIG. 26 is a diagram for explaining the third image quality determination process.

[0546] Referring to Figure 26, the third image quality judgment process may include an image segmentation information acquisition step S3751, a quality judgment target area acquisition step S3753, a morphological index acquisition step S3755 of the target area, and a step S3757 of comparing the morphological index of the target area with a reference value.

[0547] The image segmentation information obtaining step S3751 may include obtaining segmentation information obtained based on the medical image segmented by the image segmentation module 3500.

[0548] The segmentation information may include information about the anatomical structure of the human body obtained by segmenting the medical image. The segmentation information may include morphological indicators of at least a portion of the anatomical structure of the human body obtained by segmenting the medical image. The segmentation information may also include morphological values ​​of at least a portion of the anatomical structure of the human body obtained by segmenting the medical image.

[0549] The morphological index may be obtained based on a morphological value. The morphological index may be obtained based on a plurality of morphological values, for example, a first morphological value and a second morphological value. For example, the morphological index may be obtained based on a ratio between a first morphological value and a second morphological value. In this case, the first morphological value may be the cerebellum, and the second morphological value may be an anatomical region within the skull. The morphological index obtained by the ratio between the first morphological value and the second morphological value may represent an ICV value. The morphological value may be information regarding the area, volume, position, or shape of at least a portion of an anatomical structure of the human body.

[0550] The quality assessment target region acquisition step S3753 may include identifying a target region for performing a third image quality assessment based on the segmentation information. Here, the target region may include at least a portion of the anatomical structure of the human body acquired through medical image segmentation. The target region may also include at least a portion of the anatomical structure of the human body that may affect the reliability of the disease assessment result.

[0551] For example, the target region may include the outermost region of the human anatomical structure acquired through medical image segmentation. The target region may include a region of the human anatomical structure acquired through medical image segmentation that serves as a basis for disease diagnosis. The target region may include any one of the leftmost, rightmost, uppermost, or lowermost regions of the human anatomical structure acquired through medical image segmentation. When image quality assessment is performed based on a medical image of the brain, the target region may include at least a portion of the internal skull region of the brain acquired through medical image segmentation.

[0552] The step S3755 of acquiring morphological indicators of the region of interest may include acquiring morphological indicators for the identified region of interest. Illustratively, the step S3755 of acquiring morphological indicators of the region of interest may include acquiring morphological indicators of the region of interest based on at least a portion of a plurality of medical images acquired by the image acquisition device. The step S3755 of acquiring morphological indicators of the region of interest may include acquiring morphological values ​​of the region of interest based on at least a portion of the plurality of medical images acquired by the image acquisition device.

[0553] The step S3757 of comparing the morphological indices of the target region with a reference value may include comparing the morphological indices or morphological values ​​of the target region with a predetermined reference value. Here, the predetermined reference value refers to an average value of the morphological values ​​of the target region for a large number of ordinary people. The predetermined reference value may refer to an index or measurement value related to the target region when medical information with a certain level of reliability or higher can be obtained when diagnosing a disease based on the morphological indices of the target region.

[0554] FIG. 27 is a diagram illustrating an embodiment of the third image quality determination process.

[0555] 27(a) to 27(c), a third image quality assessment process according to an embodiment may include performing a quality assessment based on a medical image of the brain. According to an embodiment, the third image quality assessment process may include performing a quality assessment based on segmentation information obtained by anatomically segmenting the medical image of the brain, regarding whether the segmentation information meets a quantitative criterion.

[0556] The image segmentation information acquisition step S3751 may include receiving segmentation information obtained by anatomically segmenting an MRI image of the brain. The quality assessment target region acquisition step S3753 may include identifying a quality assessment target region among the anatomically segmented regions of the brain, for example, a predetermined region corresponding to the cerebellum among the anatomically segmented regions of the brain.

[0557] For example, if the quality assessment target region is a predetermined region corresponding to the cerebellum, the segmentation information received in image segmentation information acquisition step S3751 may be information acquired based on an MRI image in which a predetermined region corresponding to one part of the brain is not completely imaged. Here, the criterion for whether the cerebellum is completely imaged may be the same as the criterion for defining an internal skull region based on the morphological value of the cerebellum in the part describing ICV in this specification.

[0558] Referring to (a) of Figure 27, the image segmentation information acquisition step S3751 may include receiving segmentation information acquired based on a first target area image in which at least a portion of a predetermined area corresponding to the cerebellum among each area of ​​the brain is not imaged.

[0559] Referring to (b) of Figure 27, the image segmentation information acquisition step S3751 may include receiving segmentation information acquired based on a second target area image in which all of the specified areas corresponding to the cerebellum among the various areas of the brain are photographed.

[0560] Referring to (c) of Figure 27, the image segmentation information acquisition step S3751 may include receiving segmentation information acquired based on a third target area image in which all of the specific areas corresponding to the cerebellum among the brain areas have been imaged, but at least some of the areas corresponding to other anatomical structures other than the cerebellum have not been completely imaged.

[0561] The step S3755 of acquiring morphological indicators of the target region may include acquiring morphological indicators of the quality assessment target region, for example, a volume value of the cerebellum identified in the quality assessment target region. The step S3755 of acquiring morphological indicators of the target region may include comparing the morphological indicators of the target region with a reference value. The step S3755 of acquiring morphological indicators of the target region may include comparing the volume value of the cerebellum identified in the target region with a predetermined reference value to determine whether or not the volume value satisfies a quantitative criterion.

[0562] For example, the predetermined reference value may refer to a morphological index of the cerebellum acquired based on a second target region image in which a predetermined region corresponding to the cerebellum is completely captured, as shown in (b) of Figure 27. The predetermined reference value may be acquired based on a morphological index of the cerebellum acquired based on a plurality of images in which a predetermined region corresponding to the cerebellum is completely captured. The predetermined reference value may be an average value of the volume values ​​of the cerebellum acquired based on the plurality of images.

[0563] In this case, step S3757 of comparing the morphological index of the target region with a reference value may include comparing the morphological index of the cerebellum acquired based on the first target region image in which at least a portion of a predetermined region corresponding to the cerebellum is not imaged as in Figure 27(a) with the morphological index of the cerebellum acquired based on the second target region image of Figure 27(b), i.e., with a predetermined reference value, to determine whether the quantitative criterion is satisfied. Step S3757 of comparing the morphological index of the target region with the reference value may include performing an image quality determination based on whether the volume value of the cerebellum acquired based on the first target region image of Figure 27(a) shows a difference of more than a certain range from the predetermined reference value.

[0564] Referring to (c) of Figure 27, in the case of a third target area image in which the area corresponding to the cerebellum identified in the target area is completely captured, but at least a portion of the area corresponding to other anatomical structures other than the cerebellum is not completely captured, the third image quality assessment process may include again identifying the quality assessment target area and then performing image quality assessment.

[0565] Referring to (c) of Figure 27, when image quality is assessed based on the region corresponding to the cerebellum, it is possible to assess the inferior leakage of the image, but it may be difficult to assess the superior leakage of the image. To avoid this problem, other indices than the region corresponding to the cerebellum can be used as an index for assessing image quality. For example, the volume (or ICV) of the region corresponding to the skull or cerebrum can be used as an index for assessing quality.

[0566] Alternatively, if a morphological index obtained based on the region corresponding to the cerebellum (e.g., the ICV value of the cerebellum) is equal to or greater than a reference value, it can be determined that imaging of other regions excluding the cerebellum has been incomplete.

[0567] According to an embodiment, the fourth image quality determination unit 3770 may perform image quality determination based on a relationship between segmentation information obtained by anatomically segmenting a medical image acquired from an image acquisition device and artifact information extracted from the medical image acquired from the image acquisition device. For example, the fourth image quality determination unit 3770 may determine whether an artifact occurring in the medical image overlaps with at least a portion of an anatomical structure included in the medical image.

[0568] For example, when the image analyzer 3000 captures an image of a specific part of the human body using an image capture device and analyzes the image, it can diagnose illness based on segmentation information obtained by anatomically segmenting each region of the human body. If artifacts are present in the segmentation information that can serve as the basis for illness diagnosis, the result of illness diagnosis based on the medical image cannot have a certain level of reliability. Conversely, even if artifacts are present in the medical image, if the artifacts do not affect the segmentation information that can serve as the basis for illness diagnosis, for example, if the artifacts occur in an area that does not overlap with the segmentation information, the result of illness diagnosis based on the medical image can have a certain level of reliability. Therefore, even if artifacts are present in the medical image, the fourth image quality determiner 3770 can perform a function of determining whether the medical image can be properly analyzed by considering whether the artifacts affect the segmentation information that can serve as the basis for illness diagnosis.

[0569] As a more specific example, when the image analysis device 3000 analyzes an MRI image of the brain, it can diagnose a disease based on segmentation information obtained by anatomically segmenting each region of the brain. If artifacts occur in at least a portion of the segmented anatomical regions of the brain that can serve as the basis for disease diagnosis, the disease diagnosis result based on the medical image cannot have a certain level of reliability. Therefore, the fourth image quality assessment unit 3770 can perform a function of determining whether the medical image can be analyzed normally by considering whether artifacts occur in at least a portion of the anatomical structures of each region of the brain included in the MRI image.

[0570] FIG. 28 is a diagram for explaining the fourth image quality determination process.

[0571] Referring to FIG. 28, the fourth image quality judgment process may include an artifact occurrence location information acquisition step S3771, an image segmentation information acquisition step S3773, a region of interest acquisition step S3775, and a relationship determination step S3777 between the artifact occurrence location and the region of interest.

[0572] The artifact occurrence position information acquisition step S3771 may include acquiring artifact occurrence position information. At this time, the artifact occurrence position information may be acquired from the image analysis device 3000. The artifact occurrence position information may include artifact region information related to the artifact occurrence position, for example, a saliency map in the form of a heat map indicating an area where an artifact is likely to be located in the medical image, a bounding box for displaying an area where an artifact is located in the medical image, coordinate information or pixel information where the artifact is located in the medical image, etc., and a description thereof has been given above in relation to the second image quality determination unit 3730, so detailed description thereof will be omitted.

[0573] Artifacts, as described below, can include various types of defects or conditions that affect image-based information capture, such as noise, image resolution, image brightness, and various other image imperfections that result from improper video sampling.

[0574] The image segmentation information acquiring step S3773 may include acquiring segmentation information acquired based on the medical image segmented by the image segmentation module 3500. Here, the image segmentation information may include information acquired by anatomically segmenting the medical image. The image segmentation information may include segmentation information acquired by anatomically segmenting an MRI image of the brain, which has been described above in relation to the third image quality determining unit 3750, and therefore will not be described in detail.

[0575] The region of interest acquisition step S3775 may include acquiring information related to the region of interest based on segmentation information obtained by anatomically segmenting the medical image. The region of interest acquisition step S3775 may include acquiring information related to the region of interest from the segmentation information obtained by anatomically segmenting the medical image. The region of interest acquisition step S3775 may segment the medical image to acquire one or more regions and acquire a region corresponding to the region of interest.

[0576] The region of interest may include at least a portion of the anatomical structure of the human body acquired through medical image segmentation. The region of interest may be predefined as a region corresponding to a specific anatomical (or functional) element on the medical image. The region of interest may include the outermost region of the anatomical structure of the human body acquired through medical image segmentation from the center of the medical image. The region of interest may include a region of the anatomical structure of the human body acquired through medical image segmentation that is the basis for disease diagnosis. When image quality assessment is performed based on an MRI image of the brain, the region of interest may include at least a portion of the skull region and the internal skull region of the brain acquired through medical image segmentation. The region of interest may be an element related to the diagnosis of the target disease.

[0577] The step S3777 of determining the relationship between the artifact occurrence location and the region of interest may include determining the relationship between the artifact occurrence location and the region of interest. The step S3777 of determining the relationship between the artifact occurrence location and the region of interest may include performing a quality determination based on information about artifacts included in the medical image and information about the region of interest among the anatomical structures of the human body obtained by segmenting the medical image. The step S3777 of determining the relationship between the artifact occurrence location and the region of interest may perform an image quality determination based on whether the degree of overlap between the artifact occurring in the medical image and the region of interest affects the reliability of the disease diagnosis result below a certain level.

[0578] The step S3777 of determining the relationship between the location of artifact occurrence and the region of interest may include performing a quality determination based on whether at least a portion of the artifact occurring in the medical image is located on the region of interest or whether at least a portion of the artifact occurring in the medical image overlaps with the region of interest.

[0579] The step S3777 of determining a relationship between the artifact occurrence position and the region of interest may include determining the relationship between the artifact occurrence position and the region of interest based on the relationship between a reference line (artifact boundary) associated with the artifact position and the region of interest. Here, the artifact boundary may be the outline of the artifact region. The artifact boundary may be obtained from a neural network model that determines the presence or absence of an artifact and may be obtained based on a saliency map (e.g., CAM) associated with the artifact position estimation.

[0580] The step S3777 of determining the relationship between the artifact occurrence location and the region of interest may include obtaining the outline of the artifact region and the outline of the region of interest on the medical image, and determining whether the artifact and the region of interest overlap based on whether the outline of the artifact region and the outline of the region of interest overlap on the medical image. In this case, the image analysis device 3000 may determine that the quality of the medical image is normal if the number of intersections between the outline of the artifact region and the outline of the region of interest on the medical image is less than two. Alternatively, the image analysis device 3000 may determine that the quality of the target medical image is abnormal if the number of intersections is two or more.

[0581] The step S3777 of determining the relationship between the artifact occurrence location and the region of interest may include obtaining the number of common pixels contained in both the artifact region and the region of interest on the medical image and determining the degree of overlap between the artifact and the region of interest using the number of common pixels. In this case, the image analysis device 3000 may determine that the quality of the medical image is abnormal if the number of common pixels exceeds a predetermined reference value. Alternatively, the image analysis device 3000 may determine that the quality of the medical image is normal if the number of common pixels is equal to or less than the predetermined reference value.

[0582] The step S3777 of determining the relationship between the artifact occurrence position and the region of interest may include making a determination as to whether an artifact has occurred in at least a portion of the region of interest, for example, the skull region or the region inside the skull. The step S3777 of determining the relationship between the artifact occurrence position and the region of interest may include making a determination as to whether an artifact overlaps at least a portion of the region of interest, for example, the skull region and the region inside the skull.

[0583] 29 is a diagram illustrating an image in which artifacts occur. The image quality determination process along the range in which artifacts occur will be described below with reference to FIGS.

[0584] Referring to (a) of Figure 29, an image including artifacts overlapping the skull region and the internal region of the skull can be acquired. Referring to (b) of Figure 29, an image including artifacts overlapping a portion of the skull region can be acquired. Referring to (a) and (b) of Figure 29, the region of interest can be the internal region of the skull or a region corresponding to the skull.

[0585] Referring to (a) of Figure 29, if the target medical image includes artifacts that overlap with the skull region and the internal region of the skull, and the region of interest is the internal region of the skull or the skull region, determining the relationship between the location where the artifact occurs and the region of interest (S3777) may include determining that the region of interest and the location where the artifact occurs overlap.

[0586] Referring to (b) of Figure 29, if the target medical image contains an artifact that overlaps with the skull region and the region of interest is an internal skull region, determining the relationship between the artifact occurrence location and the region of interest (S3777) may include determining that the region of interest and the artifact occurrence location do not overlap.

[0587] Referring to (b) of Figure 29, if the target medical image includes an artifact that overlaps with the skull region and the region of interest is the skull region, determining the relationship between the artifact occurrence location and the region of interest (S3777) may include determining that the region of interest and the artifact occurrence location overlap.

[0588] The artifacts illustrated in Figure 29 refer to areas determined as artifact areas. Areas determined as artifact areas can be obtained through a segmentation neural network model that obtains artifact areas or a sorter model that obtains artifact information, and the related content will be described in more detail below.

[0589] The fourth image quality determination can include obtaining a plurality of reference regions associated with the artifact and determining the quality of the image based on the overlap of each reference region with the region of interest.

[0590] According to one embodiment, the step S3777 of determining a relationship between an artifact occurrence location and a region of interest may include performing an image quality determination based on artifact information related to the location of the artifact and an area in which the region of interest is distributed. The step S3777 of determining a relationship between an artifact occurrence location and a region of interest may include performing an image quality determination based on information regarding the probability that an artifact is located in a specific area in the medical image and region of interest information. The step S3777 of determining a relationship between an artifact occurrence location and a region of interest may include determining whether the degree of overlap between the artifact occurring in the medical image and the region of interest exceeds a predetermined reference value.

[0591] The information regarding the probability that an artifact is located in a specific region within the medical image may include artifact region information. Here, the artifact region information may include a first artifact region through an n-th artifact region in order of the likelihood that an artifact is located within the medical image. The artifact regions may include a first region and a second region located within the first region, as described above in the second image quality assessment process. The first region may be a region on the feature map whose association with the target artifact is equal to or greater than a first reference value, and the second region may be a region on the feature map whose association with the target artifact is equal to or greater than a second reference value that is greater than the first reference value.

[0592] FIG. 30 is a diagram for explaining a fourth image quality assessment process according to another embodiment, in which it is determined whether an artifact has occurred overlapping with a region of interest.

[0593] 30(b), the artifact regions may include a first artifact region AR1, a second artifact region AR2, a third artifact region AR3, or a fourth artifact region AR4, in order of likelihood that an artifact will be located in the medical image. The first artifact region AR1 through the fourth artifact region AR4 may include at least a portion of the region in which the artifact in the image is located. The artifact regions may be the first artifact region AR1, the second artifact region AR2, the third artifact region AR3, and the fourth artifact region AR4, in order of likelihood that the artifact will be highly associated with the artifact.

[0594] Alternatively, the artifact region may include a region on the feature map whose association with the target artifact is equal to or greater than a first reference value (e.g., the third artifact region AR3 and the fourth artifact region AR4).The artifact region may include a region on the feature map whose association with the target artifact is equal to or greater than a second reference value that is greater than the first reference value (e.g., the first artifact region AR1 and the second artifact region AR2).

[0595] 30(b), the degree of overlap between the artifact and the region of interest can be determined using the number of common pixels included in both the artifact region and the region of interest. For example, if the number of common pixels included in both the third artifact region A3 and the fourth artifact region A4 and the region of interest exceeds a predetermined reference value, the image analysis device 3000 can determine that the quality of the medical image is abnormal. As another example, if the number of common pixels included in both the first artifact region A1 and the second artifact region A2 and the region of interest is equal to or less than a predetermined reference value, the image analysis device 3000 can determine that the quality of the medical image is normal.

[0596] Referring to (c) of FIG. 30, the outer boundary of the artifact region may be used as the artifact boundary. The artifact boundary may be determined based on a feature map associated with the artifact region, such as that shown in (a) of FIG. 30. The artifact boundary may be determined as the boundary of any one of the first artifact region AR1 through the fourth artifact region AR4, as shown in (b) of FIG. 30. The artifact boundary may be determined as the average of the boundaries of the first artifact region AR1 through the fourth artifact region AR4.

[0597] The artifact boundary may be the boundary of a first region (e.g., a region on the feature map where the correlation with the target artifact is equal to or greater than a first reference value). The artifact boundary may be the boundary of a second region (e.g., a region on the feature map where the correlation with the target artifact is equal to or greater than a second reference value that is greater than the first reference value).

[0598] Determining the relationship between the artifact occurrence location and the region of interest based on the relationship between the artifact boundary and the region of interest may include performing a quality determination based on whether or not the artifact boundary and the region of interest overlap.

[0599] If there are two or more intersections where the artifact boundary and the outline of the region of interest overlap on the medical image, as shown in (c) of Figure 30, the image analysis device 3000 can determine that the quality of the target medical image is abnormal.

[0600] If multiple artifacts exist in the target image, step S3777 of determining the relationship between the artifact occurrence location and the region of interest may include performing a quality determination based on whether or not any one of the multiple artifacts overlaps with the region of interest, or the degree of overlap.

[0601] The image output device may include an image quality related information output module 3900. The image quality related information output module 3900 may output the image quality determination result performed by the image quality determination module 3700 of the image analysis device 3000. In addition, the image quality related information output module 3900 may output information generated based on the image quality determination result performed by the image quality determination module 3700 of the image analysis device 3000.

[0602] The image output device or image analysis device may receive user input directing additional actions in response to the selection window output.

[0603] FIG. 31 is a diagram illustrating an image quality related information output module according to an embodiment.

[0604] 31, the image quality related information output module 3900 may include at least one of an error information output unit 3910 or a selection window output unit 3930. In this case, the error information output unit 3910 may output error information related to information acquired based on the image quality determination result, and the selection window output unit 3930 may output a selection window related to information acquired based on the image quality determination result.

[0605] The error information output unit 3910 and the selection window output unit 3930 may output information determined based on independent image quality assessment results. For example, the error information output unit 3910 may output error information determined based on a first quality assessment result, and the selection window output unit 3930 may output a selection window determined based on a second quality assessment result.

[0606] Alternatively, the error information output unit 3910 and the selection window output unit 3930 may output information determined based on the same image quality assessment result. The error information output unit 3910 may output error information determined based on the first quality assessment result, and the selection window output unit 3930 may output a selection window determined based on the first quality assessment result.

[0607] 31 illustrates the error information output unit 3910 and the selection window output unit 3930 as being separate units, but this is merely an example, and the error information output unit 3910 and the selection window output unit 3930 may be provided in a single physical or logical configuration. For example, the image quality related information output module 3900 may output error information generated based on the quality determination result and a selection window corresponding to the error information.

[0608] The selection window output unit 3930 may output a selection window for obtaining user instructions related to additionally executable actions based on the quality determination result.

[0609] The image quality related information output module 3900 can output at least one of the error information and the selection window in text or image format.

[0610] The image quality related information output module 3900 may output error information. In this case, the error information may include information acquired based on the image quality determination result performed by the image quality determination module 3700. The error information may include at least one of first error information associated with the first image quality determination result, second error information associated with the second image quality determination result, third error information associated with the third image quality determination result, and fourth error information associated with the fourth image quality determination result.

[0611] The image quality related information output module 3900 may output a first error information screen. The first error information may include information obtained based on the first image quality judgment result. The first error information may include information obtained based on metadata information of the medical image.

[0612] According to one embodiment, the image quality-related information output module 3900 may output a first error information screen if there is an abnormality in the file structure information of the medical image acquired from the image acquisition device, such as the file format, format, or file size of the medical image. In this case, the first error information screen may include a first notification window indicating that there is an abnormality in the file structure information of the medical image or information about the file structure of the medical image.

[0613] According to another embodiment, the image quality-related information output module 3900 may output a first error information screen if the medical image acquired from the image acquisition device is missing patient information, such as information related to personal information such as the patient's name and age, information related to the time the medical image was taken, or information related to the patient's health condition, etc. In this case, the first error information may include a first notification window informing the user that the patient information is missing from the medical image or the patient information that is missing or included in the medical image.

[0614] FIG. 32 is a diagram for explaining an example of the first error information screen.

[0615] Referring to (a) of Figure 32, the first error information screen may include at least one of a medical image MI based on the first image quality judgment, file structure information DI of the medical image, or patient information PI entered in the medical image.

[0616] In this case, the medical image MI may be a displayed image of the medical image used in the first image quality judgment. The medical image file structure information DI may include first file structure information regarding the type of the medical image, second file structure information regarding the format of the medical image, or third file structure information regarding the size of the medical image. In this case, although not shown in the drawings, the first error information screen may include a detailed information object, and the output screen may additionally display more detailed information about the medical image MI or the file structure information DI (e.g., patient's date of birth, patient's address, time of shooting, location of shooting, shooting device, etc.) depending on a user selection of the detailed information object.

[0617] 32(b), the first error information screen may include a first notification window CO including first information indicating that the first image quality judgment result indicates that the medical image is abnormal. Exemplarily, the first notification window CO may include an error code indicating that there is an abnormality in the file structure of the medical image or that patient information has been omitted from the medical image. In this case, the first error information screen may include a detailed information object, and the output screen may additionally display detailed content regarding the first information (e.g., an explanation of the error code, the cause of the error message, a manual for dealing with the error message, etc.) in response to a user selection of the detailed information object.

[0618] The image quality related information output module 3900 may output a second error information screen. The second error information may include information acquired based on the second image quality judgment result. The second error information may include information acquired based on noise information. The second error information may include information acquired based on various types of defects or conditions that affect image-based information acquisition.

[0619] According to an embodiment, the image quality related information output module 3900 may output a second error information screen when the medical image acquired from the image acquisition device contains noise. The image quality related information output module 3900 may output the second error information in text or image format.

[0620] For example, the image quality-related information output module 3900 may output a second error information screen when an artifact is included in the medical image acquired from the image acquisition device. The second error information screen may include an error message indicating that an artifact is included in the medical image. The second error information may include any one of information regarding the location of the artifact in the medical image, information regarding the degree of artifact occurrence, information regarding the range of artifact occurrence, or information regarding the type of artifact.

[0621] The second error information screen may include text or an image. For example, the second error information screen may include a display, such as an image in the form of a heat map, to indicate an area in the medical image where an artifact is likely to be located. As another example, the second error information screen may include an image representing a bounding box indicating the location of the artifact in the medical image. As yet another example, the second error information screen may include an image to indicate anatomical segmentation information in the medical image.

[0622] 33 and 34 are diagrams for explaining an example of the second error information screen.

[0623] 33(a), the second error information screen may include a medical image MI display area associated with the second image quality judgment, and an artifact information AF output area.

[0624] The medical image MI may be an image in which areas where artifacts are likely to be located are highlighted based on the second image quality judgment. The medical image MI may be an image in which areas determined to be highly associated with artifacts are highlighted in the form of a heat map. The medical image MI may be zoomed in or out by user operation to enable more detailed confirmation of the type, extent, shape, etc. of the artifact.

[0625] The artifact information AF may include at least one of information about the type of artifact, information about an area where the artifact is likely to be located, or information about the degree of occurrence of the artifact. The second error information screen includes a detailed information object, and the output screen may additionally display more detailed information about the medical image MI or the artifact information AF (e.g., information about the probability that an artifact is located in a specific area within the image, heat map information about the extent of artifact occurrence, boundary information, etc.) depending on the user's selection of the detailed information object.

[0626] 33(b), the second error information screen may include a second notification window CO containing second information regarding an artifact contained in the medical image. Illustratively, the second notification window CO may include a message displaying an error code indicating that an artifact has occurred in the medical image. In this case, the second error information screen may include a detailed information object, and the output screen may additionally display detailed content regarding the second information (e.g., an explanation of the error code, the cause of the error message, a manual for dealing with the error message, etc.) in response to a user selection of the detailed information object.

[0627] Referring to FIG. 34, artifact information AF may include first artifact information AF1 including information regarding the type of artifact, second artifact information AF2 including information regarding the area where the artifact is likely to be located, or third artifact information AF3 including information regarding the degree of occurrence of the artifact.

[0628] The first artifact information AF1 may include information about the type of artifact, information about whether the image in which the artifact has occurred can be corrected, etc. In this case, the information about whether the image in which the artifact has occurred can include information acquired based on any one of the type of artifact, the location of the artifact, the range of the artifact, or the degree of the artifact.

[0629] The second artifact information AF2 may include information about an area in the medical image where an artifact is likely to be located, information about an analyzable area among areas corresponding to anatomical structures in the medical image, etc. In this case, the information about the analyzable area may be obtained based on anatomical segmentation information and artifact occurrence location information. For example, the information about the diagnostic area may include information about an area corresponding to anatomical structures in the medical image that does not overlap with artifacts, or information about an area corresponding to anatomical structures in the medical image that has a low degree of overlap with artifacts, etc. In this case, diagnosable diseases may differ depending on the analyzable area.

[0630] The third artifact information AF3 may include information regarding the degree, level, range, etc. of artifact occurrence. In this case, the image analysis device may determine whether to proceed with image analysis based on the third artifact information AF3. For example, even if artifacts occur in the medical image, the image analysis device may determine that image analysis can be performed if the degree of artifact occurrence is determined to be lower than a predetermined reference value based on the third artifact information AF3.

[0631] The image quality related information output module 3900 may output a third error information screen. The third error information may include information acquired based on the third image quality judgment result. The third error information may also include information acquired based on segmentation information obtained by anatomically segmenting a medical image acquired from an image acquisition device.

[0632] According to one embodiment, the image quality-related information output module 3900 may output a third error information screen if at least some of the anatomical structures included in the medical image acquired from the image acquisition device do not satisfy a predetermined quantitative standard. For example, the image quality-related information output module 3900 may output the third error information screen if a value associated with a predetermined region included in the medical image and corresponding to a part of the human body does not satisfy the standard. As a more specific example, the image quality-related information output module 3900 may output the third error information screen if a volumetric value of a predetermined region included in an MRI image of the brain and corresponding to a part of the brain does not satisfy a predetermined standard. Furthermore, the image quality-related information output module 3900 may output the third error information screen in text or image format.

[0633] The third error information may include an error message indicating that at least a portion of an anatomical structure in the medical image has an abnormality or information about the anatomical structure having the abnormality. More specifically, the third error information may include information related to a target region for performing a quality assessment based on the image segmentation result. The third error information may include information related to at least a portion of an anatomical structure of the human body acquired through image segmentation that may affect the reliability of the disease assessment result.

[0634] The third error information may include information about morphological indices or values, such as area, volume, position, or shape, of at least a portion of a human anatomical structure obtained by segmenting a medical image.

[0635] The third error information may include information regarding a comparison result between the morphological index or morphological value of the target region and a predetermined reference value. As a more specific example, the third error information may include information regarding whether at least a portion of the anatomical structure included in the medical image satisfies a predetermined quantitative criterion.

[0636] The third error information screen may include text or an image. The third error information screen may include text that numerically indicates morphological indices or morphological values ​​of at least a portion of a human anatomical structure obtained by segmenting the medical image. The third error information screen may include an image in which a target region corresponding to the anatomical structure that is the basis for quality judgment is displayed on the original medical image.

[0637] 35 and 36 are diagrams for explaining an example of the third error information screen.

[0638] Referring to FIG. 35(a), the third error information screen may include medical image MI or region information RI associated with the third image quality judgment.

[0639] In this case, the medical image MI may include information related to the target region used in the third image quality assessment based on the image segmentation results. For example, when determining whether the lower part of the medical image is completely captured, the medical image may include an image displaying information related to the target region for which the quality assessment is to be performed, such as the cerebellum. The third error information screen may include buttons for zooming in or out on the medical image MI to view more detailed information about the region corresponding to the anatomical structure in the medical image. In addition, the region information RI may include any one of information related to the target region for which the image quality assessment is to be performed, information about a portion of the medical image that was not completely captured, or information about whether the target region meets a predetermined quantitative standard. In this case, the third error information screen may include a detailed information object, and the output screen may additionally display more detailed information about the medical image MI or the region information RI (e.g., information about the region of the human body that was the target of the quality assessment, information about the image region that was the target of the quality assessment, etc.) depending on the user's selection of the detailed information object.

[0640] 35(b), the third error information screen may include a third notification window CO containing third information related to a region corresponding to an anatomical structure in the medical image. Illustratively, the third notification window CO may contain error information related to a region corresponding to an anatomical structure in the medical image, such as a message containing an error code indicating that the segmentation information did not meet a predetermined quantitative criterion. In this case, the third error information screen may include a detailed information object, and the output screen may additionally display detailed content related to the third information (e.g., an explanation of the error code, the cause of the error message, a manual for dealing with the error message, etc.) in response to a user selection of the detailed information object.

[0641] Referring to FIG. 36, the region information RI may include first region information RI1 including information regarding a first target region included in a first portion of the medical image and whether the first target region meets a predetermined quantitative criterion, second region information RI2 including information regarding a second target region included in a second portion of the medical image and whether the second target region meets a predetermined quantitative criterion, or third region information RI3 including information regarding the location of anatomical structures within the medical image and whether correction is possible.

[0642] For example, the first region information RI1 may include information regarding whether a morphological index or morphological value relating to a target region located at the top of the medical image, for example, the frontal lobe, meets a predetermined quantitative criterion.

[0643] The second region information RI2 may include information regarding whether a morphological index or morphological value relating to a region of interest located at the bottom of the medical image, for example, the cerebellum, meets a predetermined quantitative criterion.

[0644] The third region information RI3 may include information regarding the position of anatomical structures in the medical image, for example, information regarding whether the anatomical structures in the medical image were photographed at an angle, and judgment information regarding whether the medical image can be corrected based on this.

[0645] The image quality related information output module 3900 may output a fourth error information screen. The fourth error information may include information obtained based on the fourth image quality determination result.

[0646] According to one embodiment, the image quality related information output module 3900 can output a fourth error information screen when an artifact occurring in the medical image overlaps with at least a portion of an anatomical structure contained in the medical image.

[0647] The fourth error information may include information obtained based on a relationship between segmentation information obtained by anatomically segmenting the medical image obtained from the image acquisition device and artifact information extracted from the medical image obtained from the image acquisition device. Exemplarily, the fourth error information may include information regarding whether artifacts have occurred in at least a portion of the anatomical structures of each brain region included in the MRI image.

[0648] The fourth error information may include information regarding whether an artifact occurring in the medical image overlaps with at least a portion of an anatomical structure included in the medical image. The fourth error information may also include information regarding the degree of overlap between the artifact occurring in the medical image and at least a portion of an anatomical structure included in the medical image. The fourth error information may also include information regarding the reliability of a disease diagnosis result obtained by analyzing the medical image based on the degree of overlap between the artifact occurring in the medical image and at least a portion of an anatomical structure included in the medical image.

[0649] The fourth error information may include information regarding the relationship between the probability that an artifact is located in a specific region within the medical image and the region of interest. The fourth error information may also include information regarding at least one of a first artifact region, which is a region within the medical image where an artifact is likely to be located, a second artifact region, which is a region within the medical image where an artifact is likely to be located, and a third artifact region, which is a region within the medical image where an artifact is unlikely to be located. The first to third artifact regions may be visually highlighted on the original medical image. The first to third artifact regions may be displayed superimposed on the original medical image in the form of a heat map.

[0650] The fourth error information screen may include text or an image. For example, the fourth error information screen may include an image displaying artifact region information on the anatomical structure of the human body obtained by segmenting the medical image, such as a saliency map in the form of a heat map indicating regions where artifacts are likely to be located in the medical image, a bounding box for displaying regions where artifacts are located in the medical image, or coordinate information or pixel information where the artifacts are located in the medical image.

[0651] 37 and 38 are diagrams for explaining an example of the fourth error information screen.

[0652] Referring to (a) of FIG. 37, the fourth error information screen may include at least one of the medical image MI associated with the fourth image quality judgment or information QI regarding the relationship between the artifact and the region of interest.

[0653] The medical image MI may be an image in which a region of interest is displayed within the image, or an image in which a region where an artifact is likely to be located is displayed within the image.

[0654] The medical image MI can display the extent of overlap between the acquired region of interest and the artifact based on the location information of the artifact. For example, the medical image MI can display information related to the overlapping area between the region of interest and the artifact, such as the area, shape, and number of pixels of the overlapping area. The medical image MI can display the region of interest and the ratio of the region of interest that overlaps with the artifact.

[0655] The medical image MI may be an image in which the relationship between the region of interest and the probability that an artifact is located in a particular region within the medical image is displayed.

[0656] For example, the medical image MI may be an image in which regions corresponding to artifacts are visually highlighted on the target medical image, or an image in which regions corresponding to artifacts are displayed in a heat map format on the target medical image, or an image in which a boundary line, for example, an outline, of the region corresponding to the artifact is displayed.

[0657] Alternatively, the medical image MI may be an image displayed such that the overlapping portion between the artifact and the region of interest is visually emphasized. The medical image MI may be an image in which a region corresponding to the artifact is displayed on the target medical image. The medical image MI may be an image in which a boundary, e.g., an outline, of a region corresponding to the artifact is displayed on the target medical image. The medical image MI may be an image in which a region corresponding to the artifact is displayed in a heat map format on the target medical image.

[0658] The information QI regarding the relationship between the artifact and the region of interest may include, but is not limited to, information regarding the type of region of interest, overlap information between the region of interest and the artifact, or quality judgment result information based on the overlap information.

[0659] In this case, the fourth error information screen includes a detailed information object, and the output screen can additionally display more detailed information QI about the relationship between the medical image MI or artifact and the region of interest (for example, information expressing the degree of overlap between the region of interest and the artifact in more detailed steps, information indicating the degree of overlap between the region of interest and the artifact in numerical form, detailed information about the part of the region of interest that overlaps with the artifact, etc.) depending on the user's selection of the detailed information object.

[0660] Referring to FIG. 37(b), the fourth error information screen may include a fourth notification window CO that includes fourth information regardin...

Claims

1. 1. A method of analyzing medical images comprising: an apparatus for acquiring medical images and performing morphological analysis on the medical images, the apparatus comprising: Obtaining a target medical image; obtaining a target scanning condition associated with the target medical image; obtaining a target morphological value associated with a target element based on voxel data of a target area corresponding to the target element included in the target medical image; determining a target correction parameter of one or more correction parameters based on the target scanning condition; and outputting a morphological index based on the determined target correction parameters and the target morphological values; Determining the target correction parameters comprises: If the target scanning condition corresponds to a first scanning condition, determining the target compensation parameters with first compensation parameters previously obtained to compensate for a first morphological value associated with the target feature, the first morphological value being obtained under the first scanning condition; if the target scanning condition corresponds to a second scanning condition, determining the target compensation parameters with previously obtained second compensation parameters for compensating for a second morphological value associated with the target feature, the second morphological value being acquired under the second scanning condition; The first morphological value, the second morphological value, the target morphological value, and the morphological indicators include at least one of a volume, a length, and a thickness of the target element. Medical image analysis methods.

2. Outputting the morphological index includes: calculating a first corrected morphological index based on the target morphological value and the first correction parameter when the target scanning condition corresponds to the first scanning condition, and calculating a second corrected morphological index different from the first corrected morphological index based on the target morphological value and the second correction parameter when the target scanning condition corresponds to the second scanning condition; outputting the first corrected morphological signature when the target scanning condition corresponds to the first scanning condition, and outputting the second corrected morphological signature when the target scanning condition corresponds to the second scanning condition. The method of claim 1 .

3. Further comprising obtaining a reference scan condition that serves as a reference for determining the target correction parameter. The method of claim 2 .

4. the first correction parameters include parameters for estimating a morphological value acquired under a third scanning condition using a morphological value acquired under the first scanning condition, and the second correction parameters include parameters for estimating a morphological value acquired under a third scanning condition using a morphological value acquired under the second scanning condition; calculating the first corrected morphological index includes obtaining the first corrected morphological value, which is an estimate for a morphological value of the target element obtained under the third scanning condition, based on the target morphological value and the first correction parameter; calculating the second corrected morphological index includes obtaining the second corrected morphological value, which is an estimate for a morphological value of the target element obtained under the third scanning condition, based on the target morphological value and the second correction parameter; The method of claim 3 .

5. the first correction parameters include a first parameter set associated with a first linear function for estimating morphological values ​​acquired under a third scanning condition using morphological values ​​acquired under the first scanning condition, and the second correction parameters include a second parameter set associated with a second linear function for estimating morphological values ​​acquired under a third scanning condition using morphological values ​​acquired under the second scanning condition; Calculating the first corrected morphological index includes obtaining the first corrected morphological value based on the first linear function including the target morphological value and the first set of parameters; calculating the second corrected morphological index includes obtaining the second corrected morphological index based on the target morphological value and the second linear function including the second set of parameters. The method of claim 2 .

6. the first scan condition, the second scan condition, and the target scan condition are associated with at least one of a magnetic field strength associated with a resolution of a medical image of a medical image acquisition device, a manufacturer of the medical image acquisition device, and a configuration of a magnetic field generated by the medical image acquisition device; The method of claim 1 .

7. the target medical image is acquired from a subject having a first characteristic; the first correction parameter or the second correction parameter is obtained from a first medical image and a second medical image obtained from a target object having the first characteristic, The first characteristic is related to the age and sex of the subject. The method of claim 1 .

8. Obtaining the target morphological value comprises: obtaining a target region from the target medical image that is associated with the target element by performing a segmentation of the target medical image into a plurality of regions corresponding to a plurality of elements including at least the target element; obtaining the target morphological value based on voxel data corresponding to the target region; the segmentation is performed using a neural network configured to obtain a plurality of regions corresponding to a plurality of features based on the target medical image; The method of claim 1 .

9. converting the target medical image acquired under the target scanning condition into a medical image corresponding to an image captured under a scanning condition other than the target scanning condition; Segmentation is performed based on the transformed medical image. The method of claim 7 .

10. where the first scanning condition, the second scanning condition, and the target scanning condition are related to magnetic field strength, the first correction parameters include a first set of parameters for correcting the first morphological value associated with the target element acquired under a first magnetic field strength, and the second correction parameters include a second set of parameters for correcting the second morphological value associated with the target element acquired under a second magnetic field strength; Determining the target correction parameters comprises: When the target scanning condition corresponds to the first magnetic field strength, the target correction parameters are determined by the first parameter set; determining the target correction parameters according to a second parameter set when the target scanning condition corresponds to the second magnetic field strength; The method of claim 1 .

11. An apparatus for analyzing medical images, comprising: an image acquisition unit for acquiring a target medical image; and a controller for providing medical image analysis information based on the target medical image. the controller is configured to obtain a target scan condition associated with the target medical image, obtain a target morphological value associated with the target element based on voxel data of a target area corresponding to a target element included in the target medical image, determine a target compensation parameter among one or more compensation parameters based on the target scan condition, and output a morphological index based on the determined target compensation parameter and the target morphological value; If the target scanning condition corresponds to a first scanning condition, determining the target compensation parameters with first compensation parameters previously obtained to compensate for a first morphological value associated with the target feature, the first morphological value being obtained under the first scanning condition; configured to determine the target compensation parameters by determining the target compensation parameters with second compensation parameters previously obtained to compensate for a second morphological value associated with the target feature, the second morphological value being acquired under the second scanning condition if the target scanning condition corresponds to a second scanning condition; The first morphological value, the second morphological value, and the target morphological value include at least one of a volume, a length, and a thickness of the target element. Medical image analysis equipment.

12. the controller is configured to calculate a first corrected morphological index based on the target morphological value and the first correction parameter when the target scanning condition corresponds to the first scanning condition, calculate a second corrected morphological index different from the first corrected morphological index based on the target morphological value and the second correction parameter when the target scanning condition corresponds to the second scanning condition, and output the morphological index by outputting the first corrected morphological index when the target scanning condition corresponds to the first scanning condition and outputting the second corrected morphological index when the target scanning condition corresponds to the second scanning condition. The medical image analysis apparatus according to claim 11.

13. the first correction parameters include parameters for estimating a morphological value acquired under a third scanning condition using a morphological value acquired under the first scanning condition, and the second correction parameters include parameters for estimating a morphological value acquired under a third scanning condition using a morphological value acquired under the second scanning condition; The controller, obtaining the first corrected morphological value, which is an estimate for the morphological value of the target element obtained under the third scanning condition, based on the target morphological value and the first correction parameter, and calculating the first corrected morphological index; and obtaining the second corrected morphological value, which is an estimate for the morphological value of the target element obtained under the third scanning condition, based on the target morphological value and the second correction parameter, to calculate the second corrected morphological index.

13. The medical image analysis apparatus of claim 12.

14. the first correction parameters include a first parameter set associated with a first linear function for estimating morphological values ​​acquired under a third scanning condition using morphological values ​​acquired under the first scanning condition, and the second correction parameters include a second parameter set associated with a second linear function for estimating morphological values ​​acquired under a third scanning condition using morphological values ​​acquired under the second scanning condition; The controller, obtaining the first corrected morphological value based on the first linear function including the target morphological value and the first set of parameters to calculate the first corrected morphological index; and configured to obtain the second corrected morphological value and calculate the second corrected morphological index based on the second linear function including the target morphological value and the second set of parameters.

14. The medical image analysis apparatus of claim 13.

15. the first scan condition and the second scan condition are associated with at least one of a magnetic field strength associated with a resolution of a medical image of a medical image acquisition device, a manufacturer of the medical image acquisition device, and a setting parameter associated with a form of a magnetic field generated by the medical image acquisition device; The medical image analysis apparatus according to claim 11.

16. the target medical image is acquired from a subject having a first characteristic; the target compensation parameters including at least one of the first compensation parameter and the second compensation parameter are obtained from a first medical image and a second medical image obtained from a subject having the first characteristic; The first characteristic is related to the age and sex of the subject. The medical image analysis apparatus according to claim 11.

17. the controller is configured to obtain a target region associated with the target feature from the target medical image by performing a segmentation of the target medical image into a plurality of regions corresponding to a plurality of features including at least the target feature, and to obtain the target morphological value based on voxel data corresponding to the target region; the segmentation is performed using a neural network configured to obtain a plurality of regions corresponding to a plurality of features based on the target medical image; The medical image analysis apparatus according to claim 11.

18. the controller is configured to convert the target medical image acquired under the target scanning condition into a medical image corresponding to an image taken under a scanning condition other than the target scanning condition, and perform the segmentation based on the converted medical image.

18. The medical image analysis apparatus of claim 17.

19. where the first scanning condition, the second scanning condition, and the target scanning condition are related to magnetic field strength, the first correction parameters include a first set of parameters for correcting the first morphological value associated with the target element acquired under a first magnetic field strength, and the second correction parameters include a second set of parameters for correcting the second morphological value associated with the target element acquired under a second magnetic field strength; The controller, When the target scanning condition corresponds to the first magnetic field strength, the target correction parameters are determined by the first parameter set; and determining the target correction parameters by determining the target correction parameters with a second parameter set when the target scanning condition corresponds to the second magnetic field strength. The medical image analysis apparatus according to claim 11.

Images