A 3D segmentation method for hematoma regions in cranial MRI scan images

Abstract

The present invention relates to the technical field of region segmentation, and particularly to a three-dimensional segmentation method for a hematoma region in a cranial MRI scan image. The method includes: acquiring a target 3D image corresponding to a patient to be detected through cranial MRI scanning; determining a T1 gradient performance factor and a T2 gradient performance factor for each voxel point in each preset direction according to the T1 and T2 weighted gray-scale change conditions of each voxel point in each preset direction; screening out the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point; determining the fuzzy possibility factor and the overall main gradient direction corresponding to each voxel point, and making a gradient perpendicular plane to which each voxel point belongs; determining the redness and swelling possibility factor corresponding to each voxel point; and segmenting the target hematoma region from the target 3D image according to the position, gray-scale value and redness and swelling possibility factor corresponding to the voxel point. The present invention realizes the segmentation of the hematoma region and improves the accuracy of the hematoma region segmentation.

Description

Technical Field

[0001] The present invention relates to the technical field of regional segmentation, and particularly relates to a three-dimensional segmentation method for a hematoma region in a cranial MRI scan image. Background Art

[0002] With the development of technology, the application of image region segmentation technology is becoming more and more extensive. For example, it can be applied to hematoma region segmentation. Currently, when performing regional segmentation, the commonly used method is to segment the required region from the image according to different gray values.

[0003] However, when segmenting the hematoma region from a cranial MRI scan image according to different gray values, the following technical problems often exist:

[0004] Due to the spread of the hematoma region, there is often a fuzzy region with blurred categories between the hematoma region and the normal tissue region, so that the gray value difference at the boundary between the hematoma region and the normal tissue region is not obvious. At this time, if the hematoma region is directly segmented based on different gray values, the accuracy of the hematoma region segmentation is often poor. Summary of the Invention

[0005] In order to solve the technical problem of poor accuracy in hematoma region segmentation, the present invention proposes a three-dimensional segmentation method for a hematoma region in a cranial MRI scan image.

[0006] In a first aspect, the present invention provides a three-dimensional segmentation method for a hematoma region in a cranial MRI scan image, and the method includes:

[0007] Obtain a target 3D image corresponding to the patient to be detected through cranial MRI scanning;

[0008] Determine the T1 gradient performance factor and the T2 gradient performance factor of each voxel point in each preset direction according to the T1 and T2 weighted gray value changes of each voxel point in each preset direction in the target 3D image;

[0009] Screen out the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point according to the T1 gradient performance factor and the T2 gradient performance factor of each voxel point in all preset directions;

[0010] Determine the fuzzy probability factor and the overall main gradient direction corresponding to each voxel point according to the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point, and make a gradient vertical plane to which each voxel point belongs;

[0011] Determine the redness and swelling probability factor corresponding to each voxel point according to the angle between each voxel point and the overall main gradient direction of its neighboring voxel points on the gradient vertical plane to which it belongs;

[0012] Segment the target hematoma region from the target 3D image according to the position, gray value, and swelling and redness possible factor corresponding to the voxel points.

[0013] Combined with the first aspect above, in a possible implementation, the determining the T1 gradient performance factor and the T2 gradient performance factor of each voxel point in each preset direction according to the T1 and T2 weighted gray value changes of each voxel point in the target 3D image in each preset direction includes:

[0014] Determine any voxel point in the target 3D image as a marked voxel point, and determine any preset direction as a marked direction;

[0015] Screen out a preset number of voxel points closest to the marked voxel point in the marked direction of the marked voxel point to form a voxel point sequence of the marked voxel point in the marked direction;

[0016] Determine the T1 gradient performance factor of the marked voxel point in the marked direction according to the difference between the T1 weighted gray values of adjacent voxel points in the voxel point sequence of the marked voxel point in the marked direction, where the T1 weighted gray value corresponding to the voxel point is the gray value of the voxel point corresponding on the T1 weighted image;

[0017] Similarly, determine the T2 gradient performance factor of the marked voxel point in the marked direction according to the difference between the T2 weighted gray values of adjacent voxel points in the voxel point sequence of the marked voxel point in the marked direction, where the T2 weighted gray value corresponding to the voxel point is the gray value of the voxel point corresponding on the T2 weighted image.

[0018] Combined with the first aspect above, in a possible implementation, the determining the T1 gradient performance factor of the marked voxel point in the marked direction according to the difference between the T1 weighted gray values of adjacent voxel points in the voxel point sequence of the marked voxel point in the marked direction includes:

[0019] Determine the absolute value of the accumulated value of the differences between the T1 weighted gray values of all adjacent voxel points in the voxel point sequence of the marked voxel point in the marked direction as the T1 gradient difference index of the marked voxel point in the marked direction;

[0020] Determine the T1 gradient performance factor of the marked voxel point in the marked direction according to the T1 gradient difference index of the marked voxel point in the marked direction, where the T1 gradient difference index and the T1 gradient performance factor are in a positive correlation relationship.

[0021] Combined with the above first aspect, in a possible implementation manner, the step of screening out the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point according to the T1 gradient performance factor and the T2 gradient performance factor of each voxel point in all preset directions includes:

[0022] Screen out the preset directions with the largest T1 gradient performance factor and the largest T2 gradient performance factor from all the preset directions of each voxel point, and use them as the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point respectively.

[0023] Combined with the above first aspect, in a possible implementation manner, the step of determining the fuzzy possibility factor and the overall main gradient direction corresponding to each voxel point according to the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point includes:

[0024] Determine the fuzzy possibility factor corresponding to each voxel point according to the included angle between the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point, the difference between the T1 weighted gray value and the T2 weighted gray value corresponding to each voxel point, and the maximum T1 gradient performance factor and the maximum T2 gradient performance factor corresponding to each voxel point, where the maximum T1 gradient performance factor corresponding to a voxel point is the maximum value among the T1 gradient performance factors of the voxel point in all preset directions, and the maximum T2 gradient performance factor corresponding to a voxel point is the maximum value among the T2 gradient performance factors of the voxel point in all preset directions;

[0025] Construct a T1 unit vector and a T2 unit vector corresponding to each voxel point according to the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point, where the T1 unit vector is a unit vector in the direction of the T1 main gradient direction, and the T2 unit vector is a unit vector in the direction of the T2 main gradient direction;

[0026] Determine the direction of the sum vector between the T1 unit vector and the T2 unit vector corresponding to each voxel point as the overall main gradient direction corresponding to each voxel point.

[0027] Combined with the above first aspect, in a possible implementation manner, the formula corresponding to the fuzzy possibility factor of a voxel point is:

[0028] ; where is the fuzzy possibility factor corresponding to the th voxel point in the target 3D image; is the serial number of the voxel point in the target 3D image; is the maximum T1 gradient performance factor corresponding to the th voxel point in the target 3D image; is the maximum T2 gradient performance factor corresponding to the The maximum T2 gradient performance factor corresponding to each voxel point; is the natural exponential function; is the absolute value of the difference between the T1-weighted gray value and the T2-weighted gray value corresponding to the th voxel point in the target 3D image; is the included angle between the main T1 gradient direction and the main T2 gradient direction corresponding to the

[0029] Combined with the above first aspect, in a possible implementation, the method of making the gradient vertical plane to which each voxel point belongs includes:

[0030] Determine any voxel point in the target 3D image as a marked voxel point, and make a plane perpendicular to the overall main gradient direction corresponding to the marked voxel point through the marked voxel point, which is denoted as the gradient vertical plane to which the marked voxel point belongs.

[0031] Combined with the above first aspect, in a possible implementation, the formula for the possible swelling factor corresponding to a voxel point is:

[0032] ; where is the possible swelling factor corresponding to the th voxel point in the target 3D image; is the serial number of the voxel point in the target 3D image; is the number of neighboring voxel points within the target circular neighborhood corresponding to the th voxel point; the target circular neighborhood corresponding to the th voxel point is the preset circular neighborhood of the th voxel point on its gradient vertical plane; is the serial number of the neighboring voxel point within the target circular neighborhood corresponding to the th voxel point in the target 3D image; is the natural exponential function; is the absolute value function; is the included angle between the overall main gradient direction corresponding to the th neighboring voxel point within the target circular neighborhood corresponding to the th voxel point in the target 3D image, and the overall main gradient direction corresponding to the th voxel point; is the possible fuzziness factor corresponding to the th neighboring voxel point within the target circular neighborhood corresponding to the is the radian measure of 180°;

[0033] Combined with the above first aspect, in a possible implementation manner, the method of segmenting a target hematoma region from a target 3D image according to the position, gray value, and swelling probability factor corresponding to the voxel points includes:

[0034] Clustering the voxel points in the target 3D image according to the position, gray value, and swelling probability factor corresponding to the voxel points to obtain target clustering clusters;

[0035] Determining a target swelling factor corresponding to each target clustering cluster according to the T1-weighted gray value and T2-weighted gray value corresponding to all voxel points in each target clustering cluster, where the T1-weighted gray value has a negative correlation with the target swelling factor, and the T2-weighted gray value has a positive correlation with the target swelling factor;

[0036] If the target swelling factor corresponding to the target clustering cluster is greater than a preset swelling threshold, then determine the target clustering cluster as the target hematoma region.

[0037] Combined with the above first aspect, in a possible implementation manner, during the process of clustering the voxel points in the target 3D image, the formula corresponding to the target distance metric between the voxel point and the clustering center is:

[0038] ;

[0039] ;

[0040] ; where is the target distance metric between the th voxel point and the th clustering center in the target 3D image; is the serial number of the voxel point in the target 3D image; is the serial number of the clustering center; and respectively represent the position difference and gray value difference between the th voxel point and the th clustering center; is the fuzzy probability factor corresponding to the th voxel point in the target 3D image; is the swelling probability factor corresponding to the th voxel point in the target 3D image; is the swelling probability factor corresponding to the th clustering center; is the absolute value function; 、 and respectively are the abscissa, ordinate, and vertical coordinate in the voxel coordinates corresponding to the th voxel point; , and are the abscissa, ordinate, and vertical coordinate in the voxel coordinates corresponding to the th clustering center, respectively; and are the T1-weighted gray value and T2-weighted gray value corresponding to the th voxel point, respectively; and are the T1-weighted gray value and T2-weighted gray value corresponding to the th clustering center, respectively.

[0041] In a second aspect, the present invention provides a three-dimensional segmentation system for a hematoma region in a cranial MRI scan image. The system includes:

[0042] A 3D image acquisition module for acquiring a target 3D image corresponding to a patient to be detected through a cranial MRI scan;

[0043] A gradient performance factor determination module for determining the T1 gradient performance factor and T2 gradient performance factor of each voxel point in each preset direction according to the T1 and T2 weighted gray value changes of each voxel point in the target 3D image in each preset direction;

[0044] A main gradient direction screening module for screening out the T1 main gradient direction and T2 main gradient direction corresponding to each voxel point according to the T1 gradient performance factor and T2 gradient performance factor of each voxel point in all preset directions;

[0045] A factor direction plane acquisition module for determining the fuzzy possible factor and the overall main gradient direction corresponding to each voxel point according to the T1 main gradient direction and T2 main gradient direction corresponding to each voxel point, and creating a gradient vertical plane to which each voxel point belongs;

[0046] A redness and swelling possible factor determination module for determining the redness and swelling possible factor corresponding to each voxel point according to the angle between each voxel point and the overall main gradient direction of its neighboring voxel points in the gradient vertical plane to which it belongs;

[0047] A hematoma region segmentation module for segmenting the target hematoma region from the target 3D image according to the position, gray value, and redness and swelling possible factor corresponding to the voxel point.

[0048] In a third aspect, a server is provided, including a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, so that the device executes the method in the first aspect or any possible implementation manner of the first aspect.

[0049] Fourthly, a computer program product is provided, which includes computer program code that, when running on a computer, causes the computer to execute the method in the first aspect or any possible implementation manner of the first aspect as described above.

[0050] Fifthly, a computer-readable storage medium is provided, which stores computer program code that, when running on a computer, causes the computer to execute the method in the first aspect or any possible implementation manner of the first aspect as described above.

[0051] The present invention has the following beneficial effects:

[0052] A three-dimensional segmentation method for a hematoma area in a cranial MRI scan image of the present invention realizes the segmentation of the hematoma area, solves the technical problem of poor accuracy in segmenting the hematoma area, and improves the accuracy of segmenting the hematoma area. Compared with directly segmenting the hematoma area based on different gray values, in addition to considering gray values, the present invention also comprehensively considers multiple factors related to the segmentation of the hematoma area, such as the T1 gradient performance factor, the T2 gradient performance factor, the fuzzy possibility factor, the red swelling possibility factor, etc., so that the target hematoma area representing the real hematoma area can be relatively accurately segmented from the target 3D image, thereby improving the accuracy of segmenting the hematoma area. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0054] Figure 1 It is a flowchart of a three-dimensional segmentation method for a hematoma area in a cranial MRI scan image of the present invention;

[0055] Figure 2 It is a schematic diagram of the composition structure of a three-dimensional segmentation system for a hematoma area in a cranial MRI scan image of the present invention;

[0056] Figure 3 It is a schematic diagram of the structure of a computer device of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] To further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following specifically describes the specific implementation manners, structures, features, and effects of the technical solutions proposed according to the present invention in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments may be combined in any suitable form.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0059] Reference Figure 1 , which shows the flow of some embodiments of a three-dimensional segmentation method for a hematoma region in a cranial MRI scan image according to the present invention. The three-dimensional segmentation method for the hematoma region in the cranial MRI scan image includes the following steps:

[0060] Step S1, through cranial MRI scanning, obtain the target 3D image corresponding to the patient to be detected.

[0061] Among them, the cranial MRI scan can be the MRI scan in a cranial MRI (Magnetic Resonance Imaging) examination, which can generate a 3D (3 Dimensions) image. The patient to be detected can be a patient to be examined for cranial swelling. The target 3D image can be a 3D image representing the cranial part of the patient to be detected.

[0062] It should be noted that MRI uses the principle of nuclear magnetic resonance (NMR). Based on the different attenuations of the released energy in different structural environments within a substance, by applying an external gradient magnetic field to detect the emitted electromagnetic waves, the positions and types of the atomic nuclei constituting this object can be obtained, and thus the internal structure image of the object can be drawn. The 3D image generated by MRI scanning can be composed of multiple MRI images. For example, according to multiple MRI images of the cranial part of the patient to be detected collected, a three-dimensional image can be constructed by the method of constructing a point cloud. At this time, the constructed three-dimensional image is the target 3D image. Among them, a single MRI image, also known as a slice image, is a two-dimensional image, which shows the cross-sectional information of the detected part.

[0063] Each slice image can correspond to a T1-weighted image and a T2-weighted image. In a T1-weighted image, fat and tissues with less water content usually show high-intensity signals, while tissues with high water content usually show low-intensity signals. In a T2-weighted image, tissues with high water content usually show high-intensity signals, while fat and tissues with low water content show low-intensity signals. For example, tissues with high water content can be hematoma areas. Fat and tissues with low water content can be, but are not limited to, muscle and bone. Generally speaking, a point cloud is a simpler and more unified structure, easier to learn, and easier to operate during geometric transformation and deformation because its connectivity often does not need to be updated.

[0064] As an example, a 3D image of the brain of a patient to be detected can be acquired by an MRI scanner as the target 3D image.

[0065] Step S2: Determine the T1 gradient performance factor and the T2 gradient performance factor of each voxel point in each preset direction according to the T1 and T2 weighted gray-scale change conditions of each voxel point in the target 3D image in each preset direction.

[0066] Among them, the target 3D image is a three-dimensional volume data format, which is a three-dimensional data set composed of a series of pixel points (called voxels in a three-dimensional image). Each voxel point represents a small volume element in space, similar to a small cube in three-dimensional space. The preset direction can be a pre-set gray-scale change analysis direction. For example, there can be 18 preset directions, which can be: 0°, 20°, 40°, 60°, 80°, 100°, 120°, 140°, 160°, 180°, 200°, 220°, 240°, 260°, 280°, 300°, 320°, 340°.

[0067] As an example, this step can include the following steps:

[0068] First step: Determine any voxel point in the target 3D image as the marked voxel point, and determine any preset direction as the marked direction.

[0069] Second step: Screen out the nearest preset number of voxel points in the marked direction of the above-mentioned marked voxel point to form a voxel point sequence of the above-mentioned marked voxel point in the above-mentioned marked direction.

[0070] The preset number may be a preset number of voxel points. For example, the preset number may be 7. The voxel points in the voxel point sequence of the marked voxel point in the above-mentioned marking direction may be sorted according to the distance between them and the marked voxel point, that is, the smaller the distance between them and the marked voxel point, the closer they are to the front in the voxel point sequence.

[0071] The third step is to determine the T1 gradient expression factor of the marked voxel point in the marked direction according to the difference between the T1 weighted grayscale values ​​corresponding to adjacent voxel points in the voxel point sequence of the marked voxel point in the marked direction.

[0072] The T1-weighted grayscale value corresponding to the voxel point may be the grayscale value corresponding to the voxel point on the T1-weighted image, that is, the grayscale value of the voxel point on the T1-weighted image to which it belongs.

[0073] For example, determining the T1 gradient expression factor of the marked voxel point in the marked direction may include the following sub-steps:

[0074] In the first sub-step, the absolute value of the accumulated value of the difference between the T1 weighted grayscale values ​​corresponding to all adjacent voxel points in the voxel point sequence of the marked voxel point in the marked direction is determined as the T1 gradient difference index of the marked voxel point in the marked direction.

[0075] The second sub-step is to determine the T1 gradient expression factor of the marked voxel point in the marked direction according to the T1 gradient difference index of the marked voxel point in the marked direction.

[0076] Among them, the T1 gradient difference index can be positively correlated with the T1 gradient performance factor.

[0077] For example, the formula corresponding to the T1 gradient expression factor of the voxel point in the target 3D image in the preset direction can be determined as follows:

[0078] ;

[0079] ;

[0080] ;in, The target 3D image The individual point is T1 gradient performance factor in a preset direction. It is the serial number of the voxel point in the target 3D image. Is the sequence number of the preset direction. The target 3D image The individual point is The T1 gradient difference index in the preset direction. is the mean of the T1 gradient difference indexes of the ith voxel point in all preset directions in the target 3D image. Generally speaking, the mean of the T1 gradient difference indexes of the voxel point in all preset directions is often not zero. Therefore, it can be used as the denominator. If the mean of the T1 gradient difference indexes of a certain voxel point in all preset directions is zero, a small positive number factor can be added to the denominator. For example, " " can be used to replace the " " in is the absolute value function. is the number of voxel points in the voxel point sequence of the ith voxel point in the target 3D image in the jth preset direction. is the serial number of the ith voxel point in the voxel point sequence of the jth preset direction in the target 3D image. is the T1-weighted gray value corresponding to the ith voxel point in the voxel point sequence of the jth preset direction in the target 3D image, where the kth voxel point is located. is the T1-weighted gray value corresponding to the ith voxel point in the voxel point sequence of the jth preset direction in the target 3D image, where the lth voxel point is located. is the number of preset directions.

[0081] It should be noted that in actual situations, the blurred area generated between the hematoma area and the surrounding normal brain tissue often has a gray value gradient, that is, the gray value of the voxel points in the blurred area often increases or decreases along a direction in different weighted images. For the normal tissue area or the hematoma area, the density of the tissue itself is relatively similar. Therefore, for the normal tissue area or the hematoma area, the gray value of the voxel points in it often does not increase or decrease along a direction in different weighted images. When is larger, it often indicates that the T1-weighted gray value of the ith voxel point in the target 3D image in the jth preset direction is more likely to increase or decrease gradually, which often indicates that the gray value near the lth voxel point in the T1-weighted image corresponding to the ith voxel point is more likely to increase or decrease along the jth preset direction of the ith voxel point, which often indicates that the The more likely a voxel point is to be a voxel point within a fuzzy region. Therefore, when is larger, it often indicates that the voxel point is more likely to be a voxel point within a fuzzy region.

[0082] Fourthly, similarly, according to the differences between the T2-weighted gray values corresponding to adjacent voxel points in the voxel point sequence of the above-mentioned marked voxel points in the above-mentioned marked direction, determine the T2 gradient performance factor of the above-mentioned marked voxel points in the above-mentioned marked direction.

[0083] Among them, the T2-weighted gray value corresponding to a voxel point can be the gray value corresponding to this voxel point on the T2-weighted image, that is, the gray value of this voxel point on its affiliated T2-weighted image. When the T2 gradient performance factor of a marked voxel point in the marked direction is larger, it often indicates that the T2-weighted gray value of the marked voxel point in the marked direction is more likely to gradually increase or decrease, often indicates that the gray value near the marked voxel point in the T2-weighted image is more likely to increase or decrease along the marked direction of the marked voxel point, and often indicates that the marked voxel point is more likely to be a voxel point within a fuzzy region.

[0084] It should be noted that the method for obtaining the T2 gradient performance factor of a marked voxel point in the marked direction can be the same as the method for obtaining the T1 gradient performance factor of the marked voxel point in the marked direction, which will not be elaborated here.

[0085] Step S3, according to the T1 gradient performance factors and T2 gradient performance factors of each voxel point in all preset directions, screen out the T1 main gradient direction and T2 main gradient direction corresponding to each voxel point.

[0086] As an example, the preset directions with the largest T1 gradient performance factor and the largest T2 gradient performance factor can be screened out from all the preset directions of each voxel point, and used as the T1 main gradient direction and T2 main gradient direction corresponding to each voxel point respectively.

[0087] Step S4, according to the T1 main gradient direction and T2 main gradient direction corresponding to each voxel point, determine the fuzzy possibility factor and the overall main gradient direction corresponding to each voxel point, and create a gradient vertical plane to which each voxel point belongs.

[0088] As an example, this step may include the following steps:

[0089] First step, according to the angle between the T1 main gradient direction and T2 main gradient direction corresponding to each voxel point, the difference between the T1-weighted gray value and T2-weighted gray value corresponding to each voxel point, and the maximum T1 gradient performance factor and maximum T2 gradient performance factor corresponding to each voxel point, determine the fuzzy possibility factor corresponding to each voxel point.

[0090] Among them, the method for obtaining the angle between the main gradient direction of T1 and the main gradient direction of T2 can be as follows: Denote the straight line with the main gradient direction of T1 as the straight-line direction as the first straight line, and denote the straight line with the main gradient direction of T2 as the straight-line direction as the second straight line, and determine the angle between the first straight line and the second straight line as the angle between the main gradient direction of T1 and the main gradient direction of T2. The maximum T1 gradient performance factor corresponding to the voxel point can be the maximum value among the T1 gradient performance factors of the voxel point in all preset directions, that is, the T1 gradient performance factor of the voxel point in the main gradient direction of T1. The maximum T2 gradient performance factor corresponding to the voxel point can be the maximum value among the T2 gradient performance factors of the voxel point in all preset directions, that is, the T2 gradient performance factor of the voxel point in the main gradient direction of T2.

[0091] For example, the formula for determining the fuzzy possibility factor corresponding to the voxel point can be:

[0092] ; where is the fuzzy possibility factor corresponding to the th voxel point in the target 3D image. is the serial number of the voxel point in the target 3D image. is the th maximum T1 gradient performance factor corresponding to the voxel point in the target 3D image. is the th maximum T2 gradient performance factor corresponding to the voxel point in the target 3D image. is the natural exponential function. is the absolute value of the difference between the T1-weighted gray value and the T2-weighted gray value corresponding to the th voxel point in the target 3D image. is the th angle between the main gradient direction of T1 and the main gradient direction of T2 corresponding to the voxel point in the target 3D image.

[0093] It should be noted that when is larger, it often indicates that in the T1-weighted image corresponding to the th voxel point, the gray value near the th voxel point is more likely to increase or decrease along a preset direction of the th voxel point, which often indicates that the th voxel point is more likely to be a voxel point in the fuzzy area. When is larger, it often indicates that in the T2-weighted image corresponding to the th voxel point, the gray value near the th voxel point is more likely to increase or decrease along a preset direction of the An increase or decrease in a preset direction of each voxel point often indicates the more likely the voxel point is a voxel point within the fuzzy region.

[0094] In actual situations, fat and tissues with less water content often appear as bright signals in T1-weighted images, while often appearing as low-dark signals in T2-weighted images. Therefore, the gray-scale difference between fat and tissues with less water content between different weighted images is relatively large. Tissues with high water content often appear as low-dark signals in T1-weighted images, while often appearing as bright signals in T2-weighted images. Therefore, the gray-scale difference between tissues with high water content between different weighted images is relatively large. The gray-scale difference of voxel points within the fuzzy region between different weighted images is often not large. When is smaller, it often indicates the smaller the difference between the T1-weighted gray-scale value and the T2-weighted gray-scale value corresponding to each voxel point, often indicates the smaller the gray-scale difference of each voxel point between different weighted images, often indicates the more likely the voxel point is a voxel point within the fuzzy region.

[0095] In actual situations, the angle between the gray-scale gradient directions of the fuzzy region in different weighted images is often small. When is smaller, it often indicates the smaller the angle between the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point, often indicates the more likely the voxel point is a voxel point within the fuzzy region. Therefore, when is larger, it often indicates the more likely the voxel point is a voxel point within the fuzzy region.

[0096] Second step, according to the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point, construct the T1 unit vector and the T2 unit vector corresponding to each voxel point.

[0097] Among them, the T1 unit vector can be a unit vector with the direction of the T1 main gradient direction. The T2 unit vector can be a unit vector with the direction of the T2 main gradient direction.

[0098] Third step, determine the direction of the sum vector between the T1 unit vector and the T2 unit vector corresponding to each voxel point as the overall main gradient direction corresponding to each voxel point.

[0099] Fourth step, determine any voxel point in the target 3D image as the marked voxel point, and draw a plane perpendicular to the overall main gradient direction corresponding to the marked voxel point through the above marked voxel point, denoted as the gradient perpendicular plane to which the marked voxel point belongs.

[0100] Among them, the marked voxel points are within their respective gradually changing vertical planes.

[0101] Step S5: Determine the possible swelling and redness factor corresponding to each voxel point according to the angle between each voxel point and the overall main gradual change direction corresponding to its neighboring voxel points on the gradually changing vertical plane to which it belongs.

[0102] Among them, any voxel point in the target 3D image can be determined as a marked voxel point; the neighboring voxel points of the marked voxel point on the gradually changing vertical plane can be the voxel points within the target circular neighborhood corresponding to the marked voxel point. The target circular neighborhood corresponding to the marked voxel point is a preset circular neighborhood of the marked voxel point on its gradually changing vertical plane. The preset circular neighborhood can be a pre-set circular area, and its radius can be 5. The marked voxel point is located at the center of its corresponding target circular neighborhood. The method for obtaining the angle between two overall main gradual change directions can be: Denote these two overall main gradual change directions as the first direction and the second direction respectively, denote the straight line with the first direction as the straight line direction as the first temporary straight line, denote the straight line with the second direction as the straight line direction as the second temporary straight line, and determine the angle between the first temporary straight line and the second temporary straight line as the angle between these two overall main gradual change directions.

[0103] As an example, the formula for determining the possible swelling and redness factor corresponding to a voxel point can be:

[0104] ; where is the possible swelling and redness factor corresponding to the th voxel point in the target 3D image. is the serial number of the voxel point in the target 3D image. is the th voxel point in the target 3D image. is the number of neighboring voxel points within the target circular neighborhood corresponding to the th voxel point. The target circular neighborhood corresponding to the th voxel point is the preset circular neighborhood of the th voxel point on its gradually changing vertical plane. is the natural exponential function. is the absolute value function. is the th voxel point in the target 3D image. is the angle between the overall main gradual change direction corresponding to the th neighboring voxel point within the target circular neighborhood corresponding to the th voxel point in the target 3D image and the overall main gradual change direction corresponding to the The fuzzy possibility factor corresponding to the th neighborhood voxel point within the target circular neighborhood corresponding to an individual voxel point. is the radian measure of 180°.

[0105] It should be noted that in actual situations, when a hematoma appears, the hematoma area in the cranial brain often spreads gradually. However, due to the limitations of the surrounding brain tissue and structures where the actual lesion area is located, its spreading process is often disorderly, that is, it usually presents an irregular shape. For example, it can often expand according to the anatomical structure of the surrounding brain tissue and can often spread along cavities such as brain sulci and ventricles. Therefore, its outer contour in the 3D image is irregular. For the actual hematoma area located in the fuzzy area that is prone to misjudgment, it often has a main direction during the spreading process. For example, when the hematoma area spreads forward along a certain gap, affected by gravity and pressure, it often spreads in the plane perpendicular to this direction. That is to say, the more a voxel point spreads on the gradient vertical plane to which it belongs, the more likely this voxel point is to represent the actual hematoma located in the fuzzy area. When is smaller, it often indicates that is closer to the maximum direction angle , which often indicates that the individual voxel point is less similar to the overall main gradient direction corresponding to the th neighborhood voxel point, which often indicates that the individual voxel point is more likely to have spread on the gradient vertical plane to which it belongs, and often indicates that the individual voxel point is more likely to represent the actual hematoma located in the fuzzy area. can be used as the weight of . When is larger, it often indicates that the voxel points within the target circular neighborhood corresponding to the individual voxel point are more likely to be voxel points within the fuzzy area, and often indicates that the calculated at this time is more valuable. Therefore, when is larger, it often indicates that the individual voxel point in the target 3D image is more likely to represent the actual hematoma located in the fuzzy area.

[0106] Step S6: Segment the target hematoma area from the target 3D image according to the position, gray value, and swelling and redness possibility factor corresponding to the voxel point.

[0107] As an example, this step may include the following steps:

[0108] First step: Cluster the voxel points in the target 3D image according to the position, gray value, and swelling and redness possibility factor corresponding to the voxel point to obtain the target cluster.

[0109] For example, a preset number of voxel points can be randomly selected from the target 3D image as the initial clustering centers, and the voxel points in the target 3D image can be clustered through the Kmeans (K-means clustering algorithm, a clustering analysis with iterative solution) algorithm. Here, the preset number can be the number of clustering clusters set in advance. For example, the preset number can be 2. During the process of clustering the voxel points in the target 3D image, the formula corresponding to the target distance metric between the voxel points and the clustering centers can be:

[0110] ;

[0111] ;

[0112] ; where, is the target distance metric between the th voxel point and the th clustering center in the target 3D image. is the serial number of the voxel point in the target 3D image. is the serial number of the clustering center. and respectively represent the position difference and gray level difference between the th voxel point and the th clustering center. is the fuzzy possibility factor corresponding to the th voxel point in the target 3D image. is the swelling and redness possibility factor corresponding to the th voxel point in the target 3D image. is the swelling and redness possibility factor corresponding to the th clustering center. is the absolute value function. , and are respectively the abscissa, ordinate, and vertical coordinate in the voxel coordinates corresponding to the th voxel point. , and are respectively the abscissa, ordinate, and vertical coordinate in the voxel coordinates corresponding to the th clustering center. and are respectively the T1-weighted gray level value and T2-weighted gray level value corresponding to the th voxel point. and are respectively the T1-weighted gray level value and T2-weighted gray level value corresponding to the th clustering center.

[0113] It should be noted that when is smaller, it often indicates that the individual voxel point is closer to the th clustering center, and it often indicates that the individual voxel point is more likely to belong to the same cluster as the cluster where the th clustering center is located. When is smaller, it often indicates that the individual voxel point has more similar weighted gray values of the same type with the th clustering center, and it often indicates that the individual voxel point is more likely to belong to the same cluster as the cluster where the th clustering center is located. is the weight. When is larger, it often indicates that the individual voxel point has a greater degree of fuzziness, and it often indicates that the individual voxel point is more affected by the fuzzy situation, and the degree of hematoma change characterized by should be adjusted larger. When is smaller, it often indicates that the individual voxel point has more similar hematoma degrees with the th clustering center, and it often indicates that the individual voxel point is more likely to belong to the same cluster as the cluster where the th clustering center is located. Therefore, when is smaller, it often indicates that the individual voxel point is more likely to belong to the same cluster as the cluster where the th clustering center is located. That is to say, when the target distance metric between the voxel point and the clustering center is smaller, it often indicates that the voxel point should be classified into the cluster where the clustering center is located.

[0114] Second, according to the T1 weighted gray values and T2 weighted gray values corresponding to all voxel points in each target cluster, determine the target swelling factor corresponding to each target cluster.

[0115] Among them, the T1 weighted gray value can have a negative correlation with the target swelling factor. The T2 weighted gray value can have a positive correlation with the target swelling factor.

[0116] For example, the formula for determining the target swelling factor corresponding to the target cluster can be:

[0117] ; where is the target swelling factor corresponding to the th target cluster. is the serial number of the target cluster. is the normalization function. is the mean value of the T1-weighted gray values corresponding to all voxel points within the th target clustering cluster. is the mean value of the T2-weighted gray values corresponding to all voxel points within the th target clustering cluster. is a preset factor greater than 0, mainly used to prevent the denominator from being 0. For example, can be 0.001.

[0118] It should be noted that the hematoma area often shows a low dark signal in the T1-weighted image, and its gray value in the T1-weighted image is often small; while in the T2-weighted image, it often shows a high-bright signal, and its gray value in the T2-weighted image is often large. Therefore, when is larger, it often indicates that the area represented by the th target clustering cluster is more likely to be the hematoma area.

[0119] In the third step, if the target swelling factor corresponding to the target clustering cluster is greater than the preset swelling threshold, then the target clustering cluster is determined as the target hematoma area.

[0120] Among them, the preset swelling threshold can be a threshold preset for screening the hematoma area. For example, the preset swelling threshold can be 0.6.

[0121] Optionally, if it is known that there is a hematoma in the brain of the patient to be detected, the target clustering cluster with the largest corresponding target swelling factor can also be directly determined as the target hematoma area.

[0122] Referring to Figure 2 , based on the same inventive concept as the above method embodiment, the present invention provides a three-dimensional segmentation system for the hematoma area in a cranial MRI scan image. The system includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the above computer program is executed by the processor, it implements the steps of a method for three-dimensional segmentation of the hematoma area in a cranial MRI scan image, which may specifically include:

[0123] A 3D image acquisition module 201, configured to acquire a target 3D image corresponding to the patient to be detected through a cranial MRI scan;

[0124] A gradient performance factor determination module 202, configured to determine the T1 gradient performance factor and the T2 gradient performance factor of each voxel point in each preset direction according to the T1 and T2 weighted gray value changes of each voxel point in the target 3D image in each preset direction;

[0125] The main gradient direction screening module 203 is configured to screen out the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point according to the T1 gradient performance factor and the T2 gradient performance factor of each voxel point in all preset directions;

[0126] The factor direction plane obtaining module 204 is configured to determine the fuzzy possible factor and the overall main gradient direction corresponding to each voxel point according to the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point, and create a gradient vertical plane to which each voxel point belongs;

[0127] The swelling possible factor determining module 205 is configured to determine the swelling possible factor corresponding to each voxel point according to the angle between each voxel point and the overall main gradient direction corresponding to its neighboring voxel points on the gradient vertical plane to which it belongs;

[0128] The hematoma area segmentation module 206 is configured to segment the target hematoma area from the target 3D image according to the position, gray value and swelling possible factor corresponding to the voxel point.

[0129] Figure 3 It is a schematic structural diagram of a computer device provided by an embodiment of the present invention. Exemplarily, as Figure 3 shown, the computer device 300 includes: a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and running on the processor 302. Wherein, when the processor 302 executes the computer program 303, the computer device can execute any one of the three-dimensional segmentation methods of the hematoma area in the foregoing described cranial MRI scan images.

[0130] Based on the same inventive concept as the above method embodiment, the present invention provides a server, including a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, so that the device executes any one of the three-dimensional segmentation methods of the hematoma area in the foregoing described cranial MRI scan images.

[0131] Based on the same inventive concept as the above method embodiment, the present invention provides a computer program product, which includes: computer program code, when the computer program code runs on a computer, the computer is enabled to execute any one of the three-dimensional segmentation methods of the hematoma area in the foregoing described cranial MRI scan images.

[0132] Based on the same inventive concept as the above method embodiment, the present invention provides a computer-readable storage medium, which stores computer program code, and when the computer program code runs on a computer, the computer is enabled to execute any one of the three-dimensional segmentation methods of the hematoma area in the foregoing described cranial MRI scan images.

[0133] In summary, compared with directly segmenting the hematoma region based on the differences in grayscale values, in addition to considering grayscale values, the present invention comprehensively considers multiple factors related to the segmentation of the hematoma region, such as the T1 gradient performance factor, the T2 gradient performance factor, the blurring probability factor, and the swelling probability factor, etc. Thus, the target hematoma region representing the true hematoma region can be relatively accurately segmented from the target 3D image, thereby improving the accuracy of the hematoma region segmentation.

[0134] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A three-dimensional segmentation method for hematoma area in cranial MRI scan images, characterized in that: The following steps are involved: Obtain the target 3D image corresponding to the patient to be tested through cranial MRI scanning; Determine the T1 gradient expression factor and the T2 gradient expression factor of each voxel point in each preset direction according to the T1 and T2 weighted grayscale changes of each voxel point in each preset direction in the target 3D image; According to the T1 gradient expression factor and T2 gradient expression factor of each voxel point in all preset directions, the main T1 gradient direction and the main T2 gradient direction corresponding to each voxel point are screened out; According to the T1 main gradient direction and T2 main gradient direction corresponding to each voxel point, determine the fuzzy possible factor and the overall main gradient direction corresponding to each voxel point, and make the gradient vertical plane to which each voxel point belongs; Determine the possible redness and swelling factor corresponding to each voxel point according to the angle between each voxel point and the overall main gradient direction corresponding to its neighboring voxel point on the gradient vertical plane, and the fuzzy possible factor corresponding to the voxel point; Segment the target hematoma area from the target 3D image according to the corresponding position, gray value and possible factor of redness and swelling of the voxel point; The step of determining the T1 gradient expression factor and the T2 gradient expression factor of each voxel point in each preset direction according to the T1 and T2 weighted grayscale changes of each voxel point in each preset direction in the target 3D image includes: Determine any voxel point in the target 3D image as a marked voxel point, and determine any preset direction as a marked direction; In the marking direction of the marked voxel point, a preset number of voxel points closest to the marked voxel point are screened to form a voxel point sequence of the marked voxel point in the marking direction; Determine the T1 gradient expression factor of the marked voxel point in the marked direction according to the difference between the T1 weighted grayscale values ​​corresponding to adjacent voxel points in the voxel point sequence of the marked voxel point in the marked direction, wherein the T1 weighted grayscale value corresponding to the voxel point is the grayscale value of the voxel point on the T1 weighted image; Similarly, the T2 gradient expression factor of the marked voxel point in the marking direction is determined based on the difference between the T2 weighted grayscale values ​​corresponding to adjacent voxel points in the voxel point sequence of the marked voxel point in the marking direction, wherein the T2 weighted grayscale value corresponding to the voxel point is the grayscale value corresponding to the voxel point on the T2 weighted image.

2. The method for three-dimensional segmentation of hematoma area in cranial MRI scan images according to claim 1, characterized in that: Determining the T1 gradient expression factor of the marked voxel point in the marked direction according to the difference between the T1 weighted grayscale values ​​corresponding to adjacent voxel points in the voxel point sequence of the marked voxel point in the marked direction includes: Determine the absolute value of the accumulated value of the difference between the T1 weighted grayscale values ​​corresponding to all adjacent voxel points in the voxel point sequence of the marked voxel point in the marked direction as the T1 gradient difference index of the marked voxel point in the marked direction; According to the T1 gradient difference index of the marked voxel point in the marked direction, the T1 gradient expression factor of the marked voxel point in the marked direction is determined, wherein the T1 gradient difference index is positively correlated with the T1 gradient expression factor.

3. The method for three-dimensional segmentation of hematoma area in cranial MRI scan images according to claim 1, characterized in that: The method of screening out the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point according to the T1 gradient expression factor and the T2 gradient expression factor of each voxel point in all preset directions includes: The preset directions with the largest T1 gradient expression factor and the largest T2 gradient expression factor are selected from all preset directions of each voxel point, and are used as the main T1 gradient direction and the main T2 gradient direction corresponding to each voxel point, respectively.

4. The method for three-dimensional segmentation of hematoma area in cranial MRI scan images according to claim 1, characterized in that: The determining of the possible fuzzy factor and the overall main gradient direction corresponding to each voxel point according to the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point includes: Determine the fuzzy possible factor corresponding to each voxel point according to the angle between the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point, the difference between the T1 weighted gray value and the T2 weighted gray value corresponding to each voxel point, and the maximum T1 gradient expression factor and the maximum T2 gradient expression factor corresponding to each voxel point, wherein the maximum T1 gradient expression factor corresponding to the voxel point is the maximum value of the T1 gradient expression factors of the voxel point in all preset directions, and the maximum T2 gradient expression factor corresponding to the voxel point is the maximum value of the T2 gradient expression factors of the voxel point in all preset directions; According to the T1 main gradient direction and the T2 main gradient direction corresponding to each voxel point, the T1 unit vector and the T2 unit vector corresponding to each voxel point are constructed, wherein the T1 unit vector is a unit vector with the direction of the T1 main gradient direction, and the T2 unit vector is a unit vector with the direction of the T2 main gradient direction; The direction of the sum vector between the T1 unit vector and the T2 unit vector corresponding to each voxel point is determined as the overall main gradient direction corresponding to each voxel point.

5. The method for three-dimensional segmentation of hematoma area in cranial MRI scan images according to claim 4, characterized in that: The formula corresponding to the fuzzy possible factor corresponding to the voxel point is: ;in, The target 3D image The fuzzy possible factors corresponding to individual pixels; is the serial number of the voxel point in the target 3D image; The target 3D image The maximum T1 gradient performance factor corresponding to the individual pixel; The target 3D image The maximum T2 gradient expression factor corresponding to the individual pixel; is a natural exponential function; The target 3D image The absolute value of the difference between the T1-weighted grayscale value and the T2-weighted grayscale value corresponding to the individual pixel point; The target 3D image The angle between the main gradient direction of T1 and the main gradient direction of T2 corresponding to the individual pixel.

6. The method for three-dimensional segmentation of hematoma area in cranial MRI scan images according to claim 1, characterized in that: The step of making a gradient vertical plane to which each voxel point belongs includes: Any voxel point in the target 3D image is determined as a marked voxel point, and a plane perpendicular to the overall main gradient direction corresponding to the marked voxel point is drawn through the marked voxel point, which is recorded as the gradient vertical plane to which the marked voxel point belongs.

7. The method for three-dimensional segmentation of hematoma area in cranial MRI scan images according to claim 1, characterized in that: The formula corresponding to the possible factor of redness and swelling corresponding to the voxel point is: ;in, The target 3D image Possible factors of redness and swelling corresponding to individual pigment points; is the serial number of the voxel point in the target 3D image; The target 3D image The number of neighboring voxel points in the target circular neighborhood corresponding to the individual voxel point; The target circular neighborhood corresponding to the individual pixel is The preset circular neighborhood of individual pixels on the gradient vertical plane to which they belong; The target 3D image The serial number of the neighboring voxel point in the target circular neighborhood corresponding to the individual voxel point; is a natural exponential function; It is the absolute value function; The target 3D image The first The overall main gradient direction corresponding to the neighborhood voxel point is The angle between the main gradient directions of the whole corresponding to the individual pixels; The target 3D image The first The fuzzy possibility factor corresponding to the neighborhood voxel points; is the number of radians equal to 180°.

8. The method for three-dimensional segmentation of hematoma area in cranial MRI scan images according to claim 1, characterized in that: The method of segmenting a target hematoma area from the target 3D image according to the position, gray value and possible factor of redness and swelling corresponding to the voxel point includes: According to the corresponding position, gray value and possible factor of redness and swelling of the voxel points, the voxel points in the target 3D image are clustered to obtain the target clustering cluster; According to the T1-weighted grayscale value and T2-weighted grayscale value corresponding to all voxel points in each target cluster, the target redness and swelling factor corresponding to each target cluster is determined, wherein the T1-weighted grayscale value is negatively correlated with the target redness and swelling factor, and the T2-weighted grayscale value is positively correlated with the target redness and swelling factor; If the target redness and swelling factor corresponding to the target cluster is greater than the preset redness and swelling threshold, the target cluster is determined as the target hematoma area.

9. The method for three-dimensional segmentation of hematoma area in cranial MRI scan images according to claim 8, characterized in that: In the process of clustering voxel points in the target 3D image, the formula corresponding to the target distance measurement between the voxel points and the cluster center is: ； ； ;in, The target 3D image Individual points and The target distance metric between cluster centers; is the serial number of the voxel point in the target 3D image; is the serial number of the cluster center; and Respectively characterize the Individual points and The position difference and grayscale difference between cluster centers; The target 3D image The fuzzy possible factors corresponding to individual pixels; The target 3D image Possible factors of redness and swelling corresponding to individual pigment points; It is The possible factors of redness and swelling corresponding to the cluster centers; It is the absolute value function; , and They are The horizontal coordinate, vertical coordinate and vertical coordinate in the voxel coordinate corresponding to the individual voxel point; , and They are The horizontal coordinate, vertical coordinate and vertical coordinate in the voxel coordinates corresponding to the cluster center; and They are T1-weighted grayscale value and T2-weighted grayscale value corresponding to individual pixel points; and They are The T1-weighted grayscale value and T2-weighted grayscale value corresponding to the cluster center.

Images