Method, apparatus and system for identifying the midsagittal plane of the head

Abstract

The present application relates to a method, apparatus and system for identifying the median sagittal plane of the head. The identification method selects no less than two core tissues and / or left-right symmetric tissues from the 3D medical image of the head, with the core tissue having the point on the median sagittal plane of the head as the centroid point, and the left-right symmetric tissue having the point on the median sagittal plane of the head as the symmetry point; based on the first learning network, segment the selected core tissues and / or left-right symmetric tissues and generate corresponding 3D segmentation masks; determine the centroid points of the core tissues in each image layer based on the 3D segmentation masks of the core tissues, and determine the symmetry points of the left-right symmetric tissues in each image layer based on the 3D segmentation masks of the left-right symmetric tissues; based on the centroid point / symmetry point cloud sets of each tissue, use the surface fitting algorithm to perform multiple rounds of surface fitting to obtain the final median sagittal plane of the head. The present application can achieve accurate and robust identification of the median sagittal plane of the head without relying too much on image quality, single tissue segmentation quality, and symmetry calculations that may produce accuracy errors.

Description

Technical Field

[0001] This application belongs to the field of medical data processing, and particularly relates to a method, device, and system for identifying the midsagittal plane of the head. Background Art

[0002] The midsagittal plane (MSP, also known as the median sagittal plane) is an ideal geometric plane that symmetrically divides the brain into two equal left and right parts through the human body. It is crucial for determining the location, size, and extent of lesions. The detection of the midsagittal plane in medical images such as head CT has important application value in clinical medicine. In the diagnosis of brain tumors, the midsagittal plane can help doctors accurately locate the tumor and evaluate its impact on the surrounding brain tissue. In addition, the midsagittal plane is also very important for the evaluation of hydrocephalus, traumatic brain injury, cerebrovascular diseases, etc., and can provide important information about changes in the ventricular system and cerebral sulci and gyri. In emergency situations such as stroke, the detection of the midsagittal plane can quickly identify whether hemorrhagic or ischemic lesions lead to more serious secondary changes such as brain herniation, which is crucial for treatment decisions and also plays an indispensable role in the diagnosis and treatment planning of nervous system diseases.

[0003] Among the existing methods for extracting the midsagittal plane, one type is to use binary processing and dilation algorithms to remove the skull from cranial medical images and extract the three-dimensional mask of the brain parenchyma, use the region growing algorithm to fill the holes in the brain parenchyma and dilate to the original skull boundary, then use the middle slice of the brain as a reference for planar ellipse fitting, and finally determine the midsagittal plane according to the major axis coordinates and rotation angle of the ellipse. This method has high requirements for image quality and depends on the segmentation quality of a single brain parenchyma tissue. Moreover, the region growing algorithm has low robustness for segmentation, especially for data in pathological changes such as tumors and hemorrhages. Another type of midsagittal plane recognition method needs to first use SVM (Support Vector Machine) and Hough transform to detect the position of the cerebral longitudinal fissure to determine the initial midsagittal plane, and then select multiple pending planes around the initial midsagittal plane, and determine the final midsagittal plane based on the symmetry parameters and gray level parameters of these pending planes. This type of method also highly depends on image quality and the recognition accuracy of the position of the cerebral longitudinal fissure. Threshold segmentation of algorithms such as SVM and parameter selection of Hough transform are sensitive and may need to be adjusted for different cases, which will also affect the data effect of complex cases.

[0004] It can be seen that in the prior art, there has not been found a method for identifying the midsagittal plane of the head that has less dependence on image quality and the segmentation accuracy of a single tissue, does not depend on the recognition accuracy of the symmetry axis, and has high recognition accuracy for different data and strong generalization ability. Summary of the Invention

[0005] This application is provided to solve the above-mentioned defects in the prior art. There is a need for a method, apparatus, and system for identifying the midsagittal plane of the head, which can achieve accurate and robust identification of the midsagittal plane of the head without relying too much on image quality and the quality of single tissue segmentation, and at the same time avoid the accuracy errors that may occur due to relying on symmetry calculations such as the symmetry axis.

[0006] According to the first aspect of the present application, a method for identifying the midsagittal plane of the head is provided. The identification method includes, by a processor: obtaining a 3D medical image of the head; performing preprocessing of normalization and size change on the obtained 3D medical image of the head; selecting no less than two core tissues and / or left-right symmetric tissues as segmentation objects from the preprocessed 3D medical image of the head, wherein the core tissue uses the point on the midsagittal plane of the head as the centroid point, and the left-right symmetric tissue uses the point on the midsagittal plane of the head as the symmetric point; based on a first learning network, performing segmentation processing on the selected core tissue / left-right symmetric tissue to generate 3D segmentation masks of each core tissue / left-right symmetric tissue; determining the centroid points of each core tissue in each image layer based on the 3D segmentation masks of each core tissue, determining the symmetric points of each left-right symmetric tissue in each image layer based on the 3D segmentation masks of each left-right symmetric tissue, and taking the set of centroid points of each core tissue / symmetric points of each left-right symmetric tissue as the point cloud set of the corresponding tissue; based on the point cloud sets of each tissue, using a surface fitting algorithm to perform multiple rounds of surface fitting, so as to screen out the midsagittal plane of the head obtained by fitting based on partial centroid points / symmetric points.

[0007] According to the second aspect of the present application, an apparatus for identifying the midsagittal plane of the head is provided, including an interface and a processor, wherein the interface is configured to receive a 3D medical image of the head; the processor is configured to execute the method for identifying the midsagittal plane of the head as described in each embodiment of the present application.

[0008] According to the third aspect of the present application, a system for identifying the midsagittal plane of the head is provided. The identification system includes a 3D medical image imaging device of the head and the apparatus for identifying the midsagittal plane of the head as described in each embodiment of the present application.

[0009] The head midsagittal plane recognition methods, devices, and systems provided by various embodiments of the present application, on the basis of fully exploring the essential features of the midsagittal plane, utilize the important feature that the centroid points of the core tissues and the symmetry points of the left and right symmetric tissues should theoretically be on the midsagittal plane or very close to the midsagittal plane. First, part or all of these two types of tissues are segmented, and then the centroid points / symmetry points of these tissues are determined. A large number of centroid points / symmetry points contain rich information associated with the head midsagittal plane. Using these points, through multiple rounds of surface fitting, the head midsagittal plane is fitted by screening out the subset of the best centroid points / symmetry points. The above recognition method does not rely on the segmentation of a single tissue, nor does it require pre-identifying the axis of symmetry / symmetry plane as a reference. Therefore, it overcomes the deficiencies of the prior art that overly rely on the image quality and the segmentation quality of a single tissue. Moreover, even if some tissues have pathological changes resulting in a large deviation of the corresponding centroid points / symmetry points from the best head midsagittal plane, the influence of these deviations can be minimized as much as possible through multiple rounds of surface fitting. Therefore, the recognition method of the present application can have high recognition accuracy and strong robustness for different recognition objects.

[0010] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features, and advantages of the present application more obvious and understandable, the following specifically illustrates the specific embodiments of the present application.

[0011] It should be understood that the foregoing general description and subsequent detailed description are exemplary and explanatory only and are not restrictive of the claimed invention. Brief Description of the Drawings

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0013] Figure 1 A flowchart showing the recognition method of the head midsagittal plane according to an embodiment of the present application.

[0014] Figure 2 A schematic diagram showing the specific process of fitting the head midsagittal plane based on the point cloud sets of each tissue according to an embodiment of the present application.

[0015] Figure 3 A partial composition schematic diagram of the head midsagittal plane recognition device according to an embodiment of the present application.

[0016] Figure 4 Schematic diagram showing the composition of a system for identifying the midsagittal plane of the head according to an embodiment of the present application. Detailed implementation manners

[0017] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions of the embodiments of the present application with reference to the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present application without creative efforts fall within the scope of protection of the present application.

[0018] Unless otherwise defined, the technical terms or scientific terms used in the present application shall have the ordinary meanings understood by those of ordinary skill in the art to which the present application pertains. Words such as "including" or "comprising" and the like mean that the elements or items appearing before the word cover the elements or items listed after the word and their equivalents, without excluding other elements or items.

[0019] The "first", "second" and similar words used in the present application do not denote any order, quantity or importance, but are only used for distinction. Words such as "including" or "comprising" and the like mean that the elements before the word cover the elements listed after the word, and do not exclude the possibility of also covering other elements. The execution order of each step in the methods described in the present application in combination with the accompanying drawings is not limited. As long as the logical relationship between the steps is not affected, several steps can be integrated into a single step, a single step can be decomposed into multiple steps, or the execution order of each step can be adjusted according to specific requirements.

[0020] It should also be understood that the term "and / or" in the present application is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in the present application generally represents an "or" relationship between the associated objects before and after.

[0021] In order to keep the following description of the embodiments of the present application clear and concise, the detailed descriptions of known functions and known components are omitted in the present application.

[0022] According to an embodiment of the present application, a method for identifying the midsagittal plane of the head is provided. Figure 1 Schematic flowchart showing the method for identifying the midsagittal plane of the head according to an embodiment of the present application.

[0023] As Figure 1As shown, first, in step 101, the processor can obtain the head 3D medical image. In some embodiments, the head 3D medical image can be, for example, a plain head CT image, a contrast-enhanced head CT scan image, or a head magnetic resonance image (MRI), etc., and the above-mentioned head 3D medical image contains multiple image layers.

[0024] In step 102, it is usually necessary to perform preprocessing of normalization and size change on the obtained head 3D medical image. In order to make the original medical images with different specifications better adapt to various subsequent processes such as tissue segmentation and centroid point / symmetry point recognition, it is necessary to make the image grayscale and image size (including image layer spacing, etc.) of the obtained head 3D medical image within a certain range through preprocessing of normalization and size change. Among them, the algorithms for normalization processing can be, for example, Z-Score normalization method, minimum-maximum normalization method, and mean normalization method, etc. This application does not limit the specific method. Only as an example, for example, the image size can be changed to a unified 512*512, and the layer spacing of the image can be changed to no more than 1.5mm through methods such as linear interpolation, etc. The specific size can be determined according to the requirements of the subsequent processing flow for the input image, and this application does not limit this.

[0025] In step 103, select no less than two core tissues and / or left-right symmetric tissues from the preprocessed head 3D medical image as the segmentation objects, where the centroid point of the core tissue is the point on the mid-sagittal plane of the head, and the symmetry point of the left-right symmetric tissue is the point on the mid-sagittal plane of the head, and further, based on the first learning network, perform segmentation processing on the selected core tissue / left-right symmetric tissue to generate 3D segmentation masks for each core tissue / left-right symmetric tissue.

[0026] Among them, the core tissues can include, for example, brain parenchyma, brainstem, crista galli, center of sella turcica, bottom of hypothalamus, internal occipital protuberance, nasal bone, superior sagittal sinus, inferior sagittal sinus, etc.; the left-right symmetric tissues can include, for example, left and right eyeballs, left and right lenses, left and right optic nerves, left and right maxillary sinuses, optic chiasm, left and right intracranial segments of internal carotid arteries, etc. Among them, the intracranial part of the internal carotid artery mainly includes segments C2 - C7. It should be noted that there may be an inclusion relationship among the above-mentioned core tissues / left-right symmetric tissues. For example, clinically, the brain parenchyma contains the brainstem, and the brainstem is located in the lower half of the brain in fewer layers. However, from the image, usually, there is a tissue boundary in the brainstem area and it can be separately segmented. Therefore, in the case where it is inconvenient to perform complete segmentation of the brain parenchyma, the brainstem can be separately segmented as a core tissue alone, and it can also be used for the fitting of the mid-sagittal plane.

[0027] In some embodiments, at least two core tissues can be selected as segmentation objects, or at least two left-right symmetric tissues can be selected as segmentation objects. It is also possible to take both types of tissues into account and ensure that the total number of tissues used as segmentation objects is not less than two. The specific number of tissues is not limited. When the computing conditions can support and there is no too high requirement for real-time recognition, a larger number of tissues can bring higher recognition accuracy of the median sagittal plane of the head. In other embodiments, when selecting the tissues to be segmented, the centroid points of the core tissues or the symmetry points of the left-right symmetric tissues should preferably have different depths in the front-back direction of the median sagittal plane of the head, so that during surface fitting, these centroid points / symmetry points can be relatively dispersed and not concentrated at positions with approximately the same depth, which may affect the accuracy of surface fitting.

[0028] Preferably, when the image quality of the 3D medical image of the head permits, the brain parenchyma should be used as a segmentation object. From the perspective of anatomical structure, there are no other left-right symmetric tissues above the eyes. Therefore, including the brain parenchyma in the segmentation object can provide more available surface fitting point information for the part above the eyes during the subsequent median sagittal plane fitting process, and better ensure the accuracy of median sagittal plane fitting.

[0029] In step 104, based on the 3D segmentation masks of each core tissue, the centroid points of each core tissue in each image layer are determined, and based on the 3D segmentation masks of each left-right symmetric tissue, the symmetry points of each left-right symmetric tissue in each image layer are determined. The set of centroid points of each core tissue / symmetry points of each left-right symmetric tissue is used as the point cloud set of the corresponding tissue.

[0030] In some embodiments, for example, existing tissue segmentation algorithms such as deep learning techniques based on convolutional neural networks, algorithm frameworks such as Unet or ResNet can be used to segment each tissue to obtain the centroid points or symmetry points of each tissue in different image layers. This application does not limit this.

[0031] In step 105, based on the point cloud sets of each tissue, a multi-round surface fitting is performed using a surface fitting algorithm to screen out the median sagittal plane of the head obtained by fitting based on some centroid points / symmetry points. Among them, the surface fitting algorithm can, for example, use the least squares method, or variant algorithms such as weighted least squares method, total least squares method, robust least squares method, or other applicable polynomial fitting, spline function fitting, neural network fitting methods, etc. This application does not limit this. In each round of surface fitting, for example, subsets of different centroid points / symmetry points can be selected from the point cloud sets of each tissue to obtain different fitting surfaces, so as to select the optimal surface as the final median sagittal plane of the head.

[0032] As described above, the centroid points of each core tissue and the symmetry points of each left-right symmetric tissue should, from a prior perspective, be located on the median plane of the head or very close to it. Therefore, the set of these points contains rich enough information about the median plane of the head. By using the surface fitting algorithm and performing multiple rounds of fitting operations, defects in various aspects such as unclear images, diseased tissues with non-standard shapes, and further accuracy losses caused by segmentation can be eliminated to a certain extent, so that the median plane in head medical images can always be identified robustly and with high accuracy.

[0033] Figure 2 FIG. shows a schematic diagram of the specific process of fitting the median plane of the head based on the point cloud sets of each tissue according to an embodiment of the present application.

[0034] As Figure 2 shown, in step 201, based on the point cloud sets of each tissue, multiple rounds of surface fitting are performed using the surface fitting algorithm, and steps 2011 - 2013 are sequentially executed during each round of surface fitting.

[0035] In step 2011, a subset of centroid points / symmetry points is randomly sampled from the point cloud sets of each tissue according to the first sampling ratio of this round.

[0036] In the embodiment of the present application, the sampling method of the subset of centroid points / symmetry points can be selected in different ways as needed. Only as an example, the first sampling ratio of the first round of surface fitting can be determined as follows: Based on the number of centroid points / symmetry points required for each round of surface fitting, the number of centroid points / symmetry points to be sampled is evenly distributed to each tissue. That is, when the total number of tissues is n and the first sampling ratio is represented as a vector containing n elements, the first sampling ratio R1 can be expressed in the form of the following formula (1):

[0037] R1 = (r 1 , r 2 , …, r n ) = (1 / n, 1 / n, …, 1 / n) (1)

[0038] where r 1 , r 2 , …, r n respectively represent the first ratios of the number of centroid points / symmetry points to be sampled in the point cloud set of the corresponding tissue to the number of centroid points / symmetry points required for this round of surface fitting. Then, random sampling can be performed in the corresponding point cloud set according to this first ratio to obtain the subset of the centroid points / symmetry points.

[0039] In some other embodiments, the first sampling ratio of the first-round surface fitting can also be determined in the following manner: First, according to the number of centroid points / symmetric points included in the point cloud sets of each tissue, determine the first ratio between it and the total number of centroid points / symmetric points of all tissues, and use each first ratio as the proportional value of each corresponding tissue in the first sampling ratio R1. That is, calculate the first sampling ratio R1 according to the following formula (2):

[0040] R1 = (r 1 , r 2 , …, r n ) = (c 1 / c, c 2 / c, …, c n / c) (2)

[0041] Among them, c 1 , c 2 , …, c n respectively represent the number of centroid points / symmetric points included in the point cloud sets of each tissue, and c represents the total number of centroid points / symmetric points of all tissues. According to the method of the above formula (2), not only can it be ensured that the subset of centroid points / symmetric points for surface fitting contains points of each segmented tissue, and moreover, the number of sampling points of each tissue is proportional to the number of points in its point cloud set. Usually, since the number of points in the point cloud set is associated with the number of image layers, therefore, a larger number of points in the point cloud set often means that the tissue is more widely distributed in the longitudinal range. Therefore, correspondingly sampling more points in this point cloud set will be conducive to a more extensive and uniform coverage of the sampling points in the subset of centroid points / symmetric points for different image layers, thereby helping to improve the accuracy of surface fitting.

[0042] In some other embodiments, especially when the segmentation quality of each tissue in the 3D medical image of the head is uneven due to shooting or other reasons, the first sampling ratio of the first-round surface fitting can also be determined in the following manner: Based on the segmentation quality of each tissue, assign a corresponding proportional weight to it to weight the first ratio for sampling of the corresponding tissue. That is, calculate the first sampling ratio R1 according to the following formula (3):

[0043] R1 = (w 1 *r 1 , w 2 *r 2 , …, w n *r n ) (3)

[0044] Among them, r 1 , r 2 , …, r nRepresents the first ratio corresponding to tissue sampling, which can be the uniform 1 / n, 1 / n, …, 1 / n obtained by the method of Equation (1), or c 1 / c, c 2 / c, …,c n / c, or it can be the first ratio determined by other methods, and the present application does not limit this. w 1 , w 2 , …, w n Represents the proportional weight of each tissue. According to the above Equation (3), the segmentation quality of a specific tissue can be further considered, and the number of sampling points in the tissue with higher segmentation quality is relatively more, thereby ensuring higher accuracy of surface fitting.

[0045] In some other embodiments, the intermediate alternative surfaces obtained by each round of fitting are presented to doctor users in an intuitive manner. For example, on the 3D medical image of the head, they can be superimposed and displayed together with the segmentation of each tissue, so as to facilitate doctors' comparison and observation. In particular, when a doctor user, for example, judges based on their experience that there is an obvious bias in the intermediate alternative surface obtained by the previous round of fitting, they can manually intervene in parameters such as the first sampling ratio used in this round of surface fitting. Only as an example, for instance, when the user observes and determines that the centroid point / symmetry point of a certain core tissue or left-right symmetric tissue in the 3D medical image of the head is significantly not on the mid-sagittal plane due to congenital malformation or a large range of lesions, they can manually modify the first sampling ratio to be used in this round of surface fitting. For example, the sampling ratio value (i.e., the first ratio) of the malformed / lesioned tissue can be set to a very small value or even set to 0 to avoid the points of this tissue having an adverse impact on surface fitting. In response to the user's interactive operation on the first sampling ratio of this round, adjust the first sampling ratio of this round, and based on the adjusted first sampling ratio, perform the intermediate alternative surface fitting of this round. In this way, during the process of multiple rounds of surface fitting, the user's experience can be taken into account, which can not only make the accuracy of head mid-sagittal plane fitting higher, but also, in some cases, accelerate the convergence speed of the entire recognition process.

[0046] In step 2012, based on the subset of centroid points / symmetry points, use the surface fitting algorithm to fit and obtain the intermediate alternative surface of this round.

[0047] In some embodiments, in each round, multiple random samplings can be performed in each tissue at the same first sampling ratio to obtain multiple subsets of centroid points / symmetry points multiple times, and surface fitting can be performed more than once, and multiple intermediate alternative surfaces of the same round can be obtained.

[0048] In step 2013, based on the matching degree threshold range, verify the matching degree of other centroid points / symmetry points outside the subset of centroid points / symmetry points with the intermediate alternative surface, and evaluate the fitting quality of the intermediate alternative surface.

[0049] In some embodiments, the matching degree threshold can be set as needed considering the iterative convergence speed, and there is no specific limitation in this application. For example, the specific calculation method of the fitting quality can be to substitute each centroid point / symmetry point outside the subset of centroid points / symmetry points into the analytical expression of the intermediate alternative surface in this round, and take the points whose calculated values do not exceed the matching degree threshold as the inliers of the intermediate alternative surface, and take the points that exceed the matching degree threshold range as the outliers of the intermediate alternative surface. Then, based on the number of inliers or outliers, or the proportion of inliers or outliers in the total number of points, evaluate the fitting quality of the intermediate alternative surface. For example, the larger the number or proportion of inliers, the higher the fitting quality, which will not be listed one by one here.

[0050] Next, in step 202, select the intermediate alternative surface with the highest fitting quality from the intermediate alternative surfaces obtained by fitting in each round of step 201 as the final head mid-sagittal plane.

[0051] In some embodiments, the first sampling ratio in this round is dynamically adjusted according to the verification result of the matching degree between other centroid points / symmetry points outside the subset of centroid points / symmetry points in the previous round and the intermediate alternative surface. Specifically, it can be carried out according to the following steps:

[0052] First, when performing the verification in each round, take the centroid points / symmetry points whose distances between other centroid points / symmetry points outside the subset of centroid points / symmetry points and the intermediate alternative surface in this round exceed the matching degree threshold range as the outliers of the intermediate alternative surface in this round;

[0053] Then, based on the core tissue / symmetrical tissue on the left and right to which each outlier belongs during the verification in the previous round, adjust the first sampling ratio in this round, and make the core tissue / symmetrical tissue with a larger proportion of outliers during the verification in the previous round have a greater degree of reduction in the sampling ratio value in this round.

[0054] According to the above method, it is possible to reasonably identify tissues with low quality due to potential reasons such as segmentation quality or the presence of lesions through multiple rounds of surface fitting. And considering that the lesions in these tissues may be local and the centroid points / symmetry points in them are not necessarily completely unavailable, therefore, in the embodiments of this application, the method of reducing the sampling ratio of this tissue is adopted instead of discarding all centroid points / symmetry points of this tissue. In this way, when the resources of centroid points / symmetry points of tissues are scarce, the accurate fitting of the head mid-sagittal plane can still be achieved.

[0055] An apparatus for identifying the midsagittal plane of the head is also provided according to an embodiment of the present application. Figure 3 A partial schematic diagram of the apparatus for identifying the midsagittal plane of the head according to an embodiment of the present application is shown. As Figure 3 shown, the identification apparatus 300 may at least include an interface 301 and a processor 302. Among them, the interface 301 may be configured to receive a 3D medical image of the head. For example, the 3D medical image of the head may be a non-contrast head CT image, a contrast-enhanced head CT scan image, or a head magnetic resonance image (MRI), etc. The present application does not limit this.

[0056] In some embodiments, the processor 302 may be configured to execute the steps in the method for identifying the midsagittal plane of the head described in various embodiments of the present application, and finally output the identified midsagittal plane of the head. The processor 302 may be a processing device including more than one general-purpose processing device, such as a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), etc. More specifically, the processor may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor running other instruction sets, or a processor running a combination of instruction sets. The processor may also be more than one dedicated processing device, such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a system on chip (SoC), etc.

[0057] In some embodiments, the identification apparatus 300 may further include a display terminal 303, which may be configured to display the 3D medical image of the head obtained by the processor 302. In other embodiments, the display terminal 303 may further be configured to display the 3D medical image of the head, the segmentation of each core tissue and the left-right symmetric tissues, and the midsagittal plane of the head identified by the processor 302. The present application does not specifically limit the presentation manner of each image in the display terminal 303.

[0058] In some embodiments, the display terminal 303 may include a liquid crystal display (LCD), a light emitting diode display (LED), a plasma display, or any other type of display. In addition, a graphical user interface (GUI) may also be presented on the display for user input and image / data display, etc., which will not be elaborated here.

[0059] A system for identifying the midsagittal plane of the head is also provided according to an embodiment of the present application. Figure 4 A schematic diagram of the composition of the system for identifying the midsagittal plane of the head according to an embodiment of the present application is shown.

[0060] As Figure 4As shown, the recognition system 400 includes a 3D medical image imaging device 401 for the head and a recognition device 402 for the midsagittal plane of the head as described in various embodiments of the present application. The 3D medical image imaging device 401 for the head may include, for example, a CT scanner with plain CT scanning and enhanced CT functions, a nuclear magnetic resonance imaging device, etc., which are not listed one by one here. The recognition device 402 for the midsagittal plane of the head has been described in detail in various embodiments of the present application and will not be elaborated here.

[0061] In some embodiments, the recognition system 400 may further include a non-transitory computer-readable storage medium (not shown), which can be used to store computer-executable instructions. When the computer-executable instructions are executed by a processor, various operations of the recognition method for the midsagittal plane of the head described in various embodiments of the present application are implemented. In addition, this storage medium can also be used to store the 3D medical image of the head received from the 3D medical image imaging device 401 for the head, and / or, a point cloud set of centroid points / symmetry points obtained by segmenting and recognizing each core tissue / symmetric tissue on the left and right, etc., which are not listed one by one here.

[0062] In some embodiments, the above non-transitory computer-readable storage medium may be, for example, a read-only memory (ROM), a random access memory (RAM), a phase change random access memory (PRAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), an electrically erasable programmable read-only memory (EEPROM), other types of random access memory (RAM), a flash drive or other forms of flash memory, a cache, a register, a static memory, a compact disc read-only memory (CD-ROM), a digital versatile disc (DVD) or other optical memories, a cassette tape or other magnetic storage devices, or any other possible non-transitory medium used to store information or instructions that can be accessed by a computer device.

[0063] The method, apparatus and system for identifying the midsagittal plane of the head according to the embodiments of the present application make full use of the prior information that the centroid points of the core tissues and the symmetry points of the left and right symmetric tissues should theoretically be located on the midsagittal plane, and utilize the centroid point / symmetry point information of multiple core tissues / left and right symmetric tissues. This not only makes the surface fitting have richer and more diverse spatial information, but also effectively avoids over-reliance on the quality of the head 3D image and the quality of single tissue segmentation. Moreover, during the iterative process of multi-round surface fitting, it is possible to screen out some centroid points / symmetry points with poor quality caused by congenital malformations or pathological changes, etc., thereby further reducing the influence of various adverse factors and ensuring that the optimal head midsagittal plane is always fitted with the best set of centroid points / symmetry points. In addition, during the process of head midsagittal plane fitting, the doctor user can also promote the efficient and accurate identification of the head midsagittal plane according to his experience through interaction. Therefore, the present application can provide efficient, high-precision and robust head midsagittal plane identification with good generalization ability for different identification objects, can provide strong support for related disease and pathological diagnosis, and has broad application prospects.

[0064] In addition, although exemplary embodiments have been described herein, the scope includes any and all embodiments based on the present application having equivalent elements, modifications, omissions, combinations (e.g., schemes of cross-combination of various embodiments), adaptations or alterations. The elements in the claims will be broadly interpreted based on the language adopted in the claims and are not limited to the examples described in this specification or during the implementation of the present application, and the examples will be interpreted as non-exclusive. Therefore, this specification and the examples are only to be considered as examples, and the true scope and spirit are indicated by the full scope of the claims and their equivalents.

[0065] The above description is intended to be illustrative rather than restrictive. For example, the above examples (or one or more of their schemes) can be used in combination with each other. For example, those of ordinary skill in the art can use other embodiments when reading the above description. In addition, in the above specific implementation manners, various features can be grouped together to simplify the present application. This should not be construed as an intention that the disclosed features not claimed are necessary for any claim. On the contrary, the subject matter of the present application may be less than all the features of a specific disclosed embodiment. Thus, the claims are incorporated herein as examples or embodiments into the specific implementation manners, where each claim is independently regarded as a separate embodiment, and considering these embodiments, they can be combined with each other in various combinations or permutations. The scope of the present application should be determined with reference to the full scope of the claims and the equivalent forms empowered by these claims.

[0066] The above embodiments are only exemplary embodiments of the present application and are not used to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions within the essence and protection scope of the present application, and such modifications or equivalent substitutions should also be regarded as falling within the protection scope of the present application.

Claims

1. A method for identifying the midsagittal plane of the head, characterized in that: The identification method comprises, by a processor: Obtain 3D medical images of the head; Normalizing and resizing the acquired 3D head medical image; Selecting at least two core tissues and / or bilaterally symmetrical tissues from the preprocessed head 3D medical image as segmentation objects, wherein the core tissues have points on the midsagittal plane of the head as centroid points, and the bilaterally symmetrical tissues have points on the midsagittal plane of the head as symmetry points; performing segmentation processing on the selected core tissues / laterally symmetrical tissues based on the first learning network to generate 3D segmentation masks for the respective core tissues / laterally symmetrical tissues; Determine the centroid point of each core tissue in each image layer based on the 3D segmentation mask of each core tissue, determine the symmetric point of each bilaterally symmetric tissue in each image layer based on the 3D segmentation mask of each bilaterally symmetric tissue, and use the set of the centroid point of each core tissue / the symmetric point of the bilaterally symmetric tissue as the point cloud set of the corresponding tissue; Based on the point cloud set of each tissue, multiple rounds of surface fitting are performed using a surface fitting algorithm, so as to screen out the midsagittal plane of the head obtained based on the fitting of some centroid points / symmetric points.

2. The identification method according to claim 1, characterized in that: The point cloud set based on each tissue is subjected to multiple rounds of surface fitting using a surface fitting algorithm, thereby screening out the midsagittal plane of the head obtained based on partial centroid point / symmetric point fitting, further comprising: In each round of surface fitting: randomly sample the point cloud set of each tissue according to the first sampling ratio of this round to obtain a subset of centroid points / symmetric points; based on the subset of centroid points / symmetric points, use the surface fitting algorithm to fit to obtain the intermediate candidate surface of this round; based on the matching degree threshold range, verify the matching degree of other centroid points / symmetric points outside the subset of centroid points / symmetric points with the intermediate candidate surface of this round, and evaluate the fitting quality of the intermediate candidate surface of this round; The intermediate candidate surface with the highest fitting quality is selected as the final midsagittal plane of the head.

3. The identification method according to claim 2, characterized in that: The first sampling ratio of the first round of surface fitting is determined as follows: A first ratio of the number of centroid points / symmetric points in the point cloud set of each tissue to the total number of centroid points / symmetric points of all tissues is used as the sampling ratio value of the corresponding tissue in the first sampling ratio.

4. The identification method according to claim 2, characterized in that: The first sampling ratio of the first round of surface fitting is determined as follows: Calculate a first ratio of the number of centroid points / symmetric points in the point cloud set of each tissue to the total number of centroid points / symmetric points of all tissues, assign a corresponding proportional weight to each tissue based on the segmentation quality of each tissue, and determine the sampling proportion value of the corresponding tissue in the first sampling proportion based on the first ratio and the proportional weight.

5. The identification method according to claim 3 or 4, characterized in that: The first sampling ratio of this round is dynamically adjusted according to the verification result of the matching degree between other centroid points / symmetric points other than the subset of centroid points / symmetric points in the previous round and the intermediate candidate surface, including: When performing each round of verification, the centroid points / symmetric points other than the subset of centroid points / symmetric points and the centroid points / symmetric points whose distances from the intermediate candidate curved surface of the round exceed the matching degree threshold range are taken as the outer points of the intermediate candidate curved surface of the round; Based on the core tissue / left-right symmetrical tissue to which each external point belongs in the previous round of verification, the first sampling ratio of this round is adjusted, and the core tissue / left-right symmetrical tissue with a larger proportion of external points in the previous round of verification has a greater degree of reduction in the sampling ratio value in this round.

6. The identification method according to claim 3 or 4, characterized in that: The identification method further comprises: After each round of fitting, the intermediate candidate surface is obtained and presented to the user; In response to a user's interactive operation on the first sampling ratio of the current round, adjusting the first sampling ratio of the current round; Based on the adjusted first sampling ratio, the intermediate candidate surface fitting of this round is performed.

7. The identification method according to any one of claims 1 to 4, characterized in that: The centroid point of the selected core tissue / the symmetry point of the left and right symmetrical tissues have different depths in the front-to-back direction of the midsagittal plane of the head.

8. The identification method according to any one of claims 1 to 5, characterized in that: The surface fitting algorithm includes the least squares method.

9. A device for identifying the midsagittal plane of the head, characterized in that: include: An interface configured to receive a 3D medical image of the head; and A processor configured to execute the method for identifying the midsagittal plane of the head as described in any one of claims 1-8.

10. A system for identifying the midsagittal plane of the head, characterized in that: include: A head 3D medical imaging device and a head midsagittal plane recognition device as claimed in claim 9.

Images