Cerebrovascular lesion segmentation method, device, storage medium, and electronic device

Abstract

The present application discloses a method, device, storage medium, and electronic device for segmenting cerebrovascular lesions, wherein the method comprises acquiring MRI image information; obtaining a segmentation result in the MRI image information through a pre-trained image segmentation model, wherein the image segmentation model uses a maximum a posteriori probability model to perform MAP estimation on a posterior distribution model used to describe tissue categories and a spatial prior distribution of leukoplastic tissue; and obtaining a segmentation result that separates the leukoplastic lesions in the cerebrovascular lesions from acute and chronic stroke lesions based on the segmentation result. The present application achieves automatic segmentation of cerebrovascular lesions, and has a strong consistency with the expert segmentation value. The present application can be used for clinical image processing and can improve the registration accuracy of clinical images.

Description

Technical Field

[0001] The present application relates to the fields of medical technology and machine learning, and specifically to a method, device, storage medium, and electronic device for segmenting cerebrovascular lesions. Background Art

[0002] Identifying abnormal cerebrovascular lesions in brain magnetic resonance imaging (MRI) images is crucial for understanding cerebral ischemia (inadequate cerebral blood flow). However, different lesion types, such as leukoaraiosis (small vessel disease) and stroke, cannot be distinguished purely based on image shape or location. Clinicians use anatomical and other medical knowledge to classify and describe these lesions.

[0003] To understand susceptibility to cerebral ischemia and associated risk factors, clinicians manually delineate and analyze vascular lesions, focusing on leukoaraiosis and separating them from stroke lesions. This approach has shown that patients with transient ischemic attack (TIA) have a lower burden of leukoaraiosis compared to those with cerebral infarction. Delineating leukoaraiosis and stroke lesions in each patient takes 30 minutes, yet large studies include hundreds or even thousands of patients. Therefore, automated segmentation is essential.

[0004] The variability of lesion shape and location is one of the main challenges in automatic segmentation of stroke scans. T2-fluid attenuated inversion recovery (FLAIR) sequences show high signal intensity in leukoaraiosis lesions, which are located around the ventricles, vary widely, and are roughly bilaterally symmetrical. Although stroke lesions are also characterized by high intensity, they can occur almost anywhere in the brain and vary greatly in size and shape. In addition, acute stroke (stroke within the past 48 hours) is visible on diffusion-weighted MR (DWI), but chronic stroke (stroke that occurred long before imaging) is not.

[0005] Furthermore, due to extremely limited scanning time, image quality in actual clinical settings is very low, with image slices typically being 5-7 mm thick and even accompanied by bright artifacts. These factors hinder the registration accuracy of clinical images and affect intensity balance.

[0006] Currently, no effective solution has been proposed to the problem that the methods for segmenting cerebral vascular lesions in related technologies cannot achieve automatic segmentation and have poor segmentation accuracy. Summary of the Invention

[0007] The main purpose of this application is to provide a cerebrovascular lesion segmentation method, device, storage medium, and electronic device to solve the problem that the cerebrovascular lesion segmentation method cannot achieve automatic segmentation and has poor segmentation accuracy.

[0008] In order to achieve the above objectives, according to one aspect of the present application, a cerebral vascular lesion segmentation method is provided.

[0009] The cerebrovascular lesion segmentation method according to the present application includes:

[0010] Acquire MRI image information;

[0011] Obtaining a segmentation result in the MRI image information by using a pre-trained image segmentation model, wherein the image segmentation model uses a maximum a posteriori probability model to perform MAP estimation on a posterior distribution model used to describe tissue categories and a spatial prior distribution of leukoaraiosis tissue;

[0012] According to the segmentation result, a segmentation result is obtained for separating the leukoaraiosis lesions from the acute and chronic stroke lesions in the cerebrovascular lesions.

[0013] In some embodiments, the image segmentation model includes intensity features, shape features, and spatial environment features, wherein the intensity features include the intensity distribution of leukoaraiosis and the intensity distribution of stroke, and the spatial environment features include the spatial distribution of leukoaraiosis. The segmentation result further includes:

[0014] Based on the T2 FLAIR sequence, the leukoaraiosis lesions in the cerebrovascular lesions were separated from the acute and chronic stroke lesions.

[0015] In some embodiments, the segmentation result in the MRI image information includes:

[0016] Generative models are used to describe the spatial distribution, shape, and appearance of healthy tissue and cerebrovascular lesions, and a posterior distribution model is established to describe tissue categories, where the prior of the tissue category captures the knowledge of spatial distribution and lesion shape;

[0017] A maximum a posteriori probability model is used to perform MAP estimation on the posterior distribution model used to describe the tissue category to obtain a segmentation result in the MRI image information.

[0018] In some embodiments, the segmentation result in the MRI image information includes:

[0019] PCA was used to construct a probability map for the training set of binary segmentation mapping of manually segmented white matter loose tissue lesions, and the spatial range of white matter loose tissue lesions was modeled to obtain the spatial prior distribution of white matter loose tissue.

[0020] In some embodiments, the MAP estimation is based on an EM algorithm, and the intensity average estimation is modeled as spatially varying while being filtered using a low-pass filter.

[0021] In some embodiments, the image segmentation model uses manually annotated leukoaraiosis lesion results as training images.

[0022] In order to achieve the above objectives, according to another aspect of the present application, a clinical image processing method is provided.

[0023] The clinical image processing method according to the present application includes:

[0024] The cerebrovascular lesion segmentation method is used for image registration.

[0025] In order to achieve the above-mentioned objective, according to another aspect of the present application, a cerebral vascular lesion segmentation device is provided.

[0026] The cerebrovascular lesion segmentation device according to the present application includes:

[0027] An acquisition module, used for acquiring MRI image information;

[0028] a processing module, configured to obtain a segmentation result in the MRI image information by using a pre-trained image segmentation model, wherein the image segmentation model uses a maximum a posteriori probability model to perform MAP estimation on a posterior distribution model used to describe tissue categories and a spatial prior distribution of leukoaraiosis tissue;

[0029] The segmentation module is used to obtain a segmentation result for separating the leukoaraiosis lesions from the acute and chronic stroke lesions in the cerebrovascular lesions according to the segmentation result.

[0030] In the embodiments of the present application, a cerebrovascular lesion segmentation method, apparatus, storage medium, and electronic device are provided. MRI image information is acquired; a segmentation result from the MRI image information is obtained using a pre-trained image segmentation model, wherein the image segmentation model uses a maximum a posteriori probability model to perform MAP estimation on a posterior distribution model used to describe tissue categories and a spatial prior distribution of leukoaraiosis tissue; and a segmentation result is then obtained based on the segmentation result, separating the leukoaraiosis lesions from acute and chronic stroke lesions in the cerebrovascular lesions. The image segmentation model is used to separate the leukoaraiosis lesions from acute and chronic stroke lesions in the cerebrovascular lesions. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The drawings that constitute part of this application are used to provide a further understanding of this application and make other features, objects and advantages of this application more apparent. The illustrative embodiment drawings of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0032] Figure 1 is a flowchart of a cerebrovascular lesion segmentation method according to an embodiment of the present application;

[0033] Figure 2 is a schematic structural diagram of a cerebrovascular lesion segmentation device according to an embodiment of the present application;

[0034] Figure 3 It is a schematic diagram of the implementation principle of the cerebrovascular lesion segmentation method according to an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0036] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0037] In this application, terms such as "upper," "lower," "left," "right," "front," "back," "top," "bottom," "inner," "outer," "center," "vertical," "horizontal," "transverse," and "longitudinal" indicate positions or locations based on the positions or locations shown in the accompanying drawings. These terms are primarily intended to better describe this application and its embodiments and are not intended to limit the devices, elements, or components indicated to having a specific orientation, or to being constructed or operated in a specific orientation.

[0038] Furthermore, some of the above terms may be used to express other meanings besides indicating a position or location. For example, the term "on" may also be used to indicate a dependency or connection in certain circumstances. Those skilled in the art will understand the specific meanings of these terms in this application based on the specific circumstances.

[0039] Furthermore, the terms "installed," "disposed," "provided with," "connected," "connected," and "socketed" should be interpreted broadly. For example, they can refer to fixed connections, removable connections, or integral structures; mechanical connections or electrical connections; direct connections, indirect connections through an intermediary, or internal communication between two devices, elements, or components. Those skilled in the art will understand the specific meanings of these terms in this application based on the specific circumstances.

[0040] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0041] In the embodiments of this application, the intensity, shape and spatial distribution of lesions are modeled to obtain anatomical knowledge of different types of lesions in order to annotate MRI images of stroke patients. This application focuses on segmenting leukoaraiosis and separating it from stroke lesions.

[0042] Furthermore, a probabilistic generative model of the effects of cerebrovascular disease on the brain is introduced. The model integrates the important characteristics of each lesion, resulting in an efficient inference algorithm for segmenting different tissues in stroke patients. The spatial and intensity distributions of leukoaraiosis, as well as the intensity distribution of stroke, are obtained. By training the model on an expert-labeled dataset, it is demonstrated that the modeling choices of this application capture concepts used by clinicians, such as symmetry and covariation of intensity patterns. For example, the segmentation model of this application combines several previously proposed anatomical segmentation methods to accurately establish lesion models.

[0043] Furthermore, during their research, the inventors discovered that intensity-based lesion segmentation algorithms exploit differences in tissue intensity to segment lesions. Spatial priors are sometimes added in the form of Markov random fields or spatial distributions. These methods successfully describe structures that have high or low signals compared to their surroundings, such as MS lesions or tumors. However, these methods cannot be used to distinguish multiple high-intensity structures, such as leukoaraiosis, stroke, and certain artifacts, because these lesions have the same intensity and can occur simultaneously in space. Clinicians use spatial features, such as the bilateral symmetry of leukoaraiosis, to distinguish them.

[0044] Shape-based methods typically model the shape of a structure through explicit or implicit representations. In this application, a shape model is utilized to capture the variability in the spatial distribution of leukoaraiosis, which develops in a consistent pattern around the ventricles. In contrast, strokes can occur at almost any random location in the brain and have no obvious shape or locational contours.

[0045] like Figure 1FIG. 1 is a flow chart of a cerebrovascular lesion segmentation method according to an embodiment of the present application, wherein the method comprises:

[0046] Step S110, obtaining MRI image information.

[0047] The MRI image information usually includes the lesion area, that is, the lesion area of ​​the cerebral blood vessels.

[0048] If it is the training phase, the spatial distribution, shape and appearance of healthy tissue and cerebrovascular lesions need to be included.

[0049] If it is the testing phase, only the MRI image of the lesion area needs to be input.

[0050] Step S120: obtaining a segmentation result in the MRI image information through a pre-trained image segmentation model, wherein the image segmentation model uses a maximum a posteriori probability model to perform MAP estimation on the posterior distribution model used to describe the tissue category and the spatial prior distribution of the leukoaraiosis tissue.

[0051] The pre-trained image segmentation model can be used to obtain a segmentation result in the MRI image information. The pre-trained image segmentation model is based on a probability estimation model, and the corresponding segmentation result can be obtained according to the estimation result.

[0052] Exemplarily, the image segmentation model uses a maximum a posteriori probability model to perform maximum a posteriori probability estimation on the posterior distribution model used to describe the tissue category and the spatial prior distribution of the leukoaraiosis tissue.

[0053] Step S130: obtaining a segmentation result for separating the leukoaraiosis lesions from the acute and chronic stroke lesions in the cerebrovascular lesions based on the segmentation result.

[0054] Based on the segmentation results, a segmentation result can be obtained that separates the leukoaraiosis lesions from the acute and chronic stroke lesions in the cerebrovascular lesions. By segmenting the individual cerebrovascular lesions in the MRI image information, the expert's knowledge of the disease is captured based on a probability model.

[0055] Furthermore, by modeling the spatial distribution of leukoaraiosis lesions and the intensity of leukoaraiosis and stroke lesions, this method automatically segments tissue that is indistinguishable based on intensity alone. Using a stroke model that combines intensity and spatial context can accurately segment leukoaraiosis lesions from acute and chronic stroke lesions.

[0056] As a preference in this embodiment, the image segmentation model includes intensity features, shape features and spatial environment features, wherein the intensity features include the intensity distribution of leukoaraiosis lesions and the intensity distribution of stroke, the spatial environment features include the spatial distribution of leukoaraiosis lesions, and the segmentation results also include: based on the T2 FLAIR sequence, separating the leukoaraiosis lesions in the cerebrovascular lesions from acute and chronic stroke lesions.

[0057] In a specific implementation, the T2-Fluid Attenuated Inversion Recovery (FLAIR) sequence can automatically separate / segment the leukoaraiosis lesions from the acute and chronic stroke lesions in the cerebrovascular lesions.

[0058] As a preference in this embodiment, the segmentation results in the MRI image information include: using a generative model to describe the spatial distribution, shape and appearance of healthy tissue and cerebrovascular lesions, establishing a posterior distribution model for describing tissue categories, wherein the prior of the tissue category captures the knowledge of spatial distribution and lesion shape; using a maximum a posteriori probability model to perform MAP estimation on the posterior distribution model for describing the tissue category to obtain the segmentation results in the MRI image information.

[0059] In a specific implementation, a posterior distribution model for describing tissue categories is established, and a maximum a posteriori probability model is used to perform MAP estimation on the posterior distribution model for describing tissue categories to obtain a segmentation result in the MRI image information.

[0060] Posterior distribution models are used to describe tissue classes, using generative models to describe the spatial distribution, shape, and appearance of healthy tissue and cerebrovascular lesions.

[0061] Exemplarily, let Ω be the set of all spatial locations (voxels) in an image, and I = {Ix}x∈Ω. Let image I be generated by a spatially varying label map C = {Cx}x∈Ω, where C represents the tissue class. For each voxel x, Cx is a binary indicator vector that encodes three tissue labels: leukoaraiosis (L), stroke (S), and healthy (H).

[0062] Furthermore, we use the notation Cx(c) = 1 to indicate that the tissue class at voxel x is c, so c∈{L,S,H}. Conversely, Cx(c) = 0. Given a label map C, the intensity observation Ιx is generated independently from a Gaussian distribution:

[0063]

[0064] in represents the normal distribution, μ is the mean, σ 2is the variance, C={L,S,H}, μ={μ L ,μ S ,μ H}, σ={σ L ,σ S ,σ H The tissue class prior captures the knowledge of spatial distribution and lesion shape.

[0065] Assume that the spatial extent of leukoaraiosis depends on the spatial distribution M = {M x} x∈Ω , where Mx is the prior for leukoaraiosis voxel x, and as described in this application, M is parameterized by the parameter α. If voxel x is not assigned to leukoaraiosis tissue, it is assigned to stroke tissue with a spatially varying probability βx, and healthy tissue with a probability of (1-βx).

[0066] For the continuity of space, a Markov Random Field (MRF) is used as the spatial prior:

[0067]

[0068] where π x =[M x (α), (1-M x (α))β x , (1-M x (α))(1-β x )] T (3)

[0069] π x is the vector of prior probabilities of the three tissue categories mentioned above, is the set of voxel locations adjacent to x, and the 3*3 matrix A is chosen to encourage adjacent voxels to share the same tissue label.

[0070] In practice, the MRF has a greater effect on the interaction between stroke and other tissues than on leukoaraiosis bordering healthy tissue, because stroke tissue is generally found to be more continuous in space, while leukoaraiosis tissue is more dispersed. Using formulas (1), (2), and (3), the posterior distribution of tissue classes is formed.

[0071]

[0072] As a preferred embodiment of this embodiment, the segmentation results in the MRI image information include: using PCA to construct a probability map for the training set of binary segmentation mapping of manually segmented white matter loose tissue lesions, and modeling the spatial range of white matter loose tissue lesions to obtain the spatial prior distribution of white matter loose tissue.

[0073] In the specific implementation, PCA is used to construct a probability map for the training set of binary segmentation mapping of manually segmented white matter loose tissue lesions, so as to model the spatial range of white matter loose tissue lesions and obtain the spatial prior distribution of white matter loose tissue.

[0074] For the spatial prior distribution of leukoaraiosis tissue, the spatial extent of leukoaraiosis lesions is first modeled:

[0075] Exemplarily, a probability map is constructed by using principal component analysis (PCA) on the training set of artificial leukoaraiosis lesion binary segmentation mapping.

[0076] set up is the mean mapping, is the important component corresponding to the K largest eigenvalues, and α k Should be weights (or loadings):

[0077]

[0078] Where Σ is the diagonal covariance matrix containing the K largest eigenvalues. Given α, the spatial prior M={M x} x∈Ω Confirmation is defined as:

[0079]

[0080] As a preference in this embodiment, the MAP estimation is based on the EM algorithm, and the intensity average estimation is modeled as a spatial variation, and a low-pass filter is used for filtering.

[0081] To obtain the segmentation map, perform MAP inference and find:

[0082]

[0083] Since exact calculation becomes infeasible when the MRF weight matrix A is non-zero, the EM (Expectation-maximization, EM) algorithm is used to estimate the solution of MAP.

[0084] For example, the fully factored distribution is used to simulate the posterior distribution P(C|I; μ, σ, α, β)

[0085]

[0086] w x is the probability vector of the three tissue classes at voxel x. Since the prior of PCA loading P(α) is not conjugate with the likelihood P(C|α), the corresponding E-step calculation is approximated using regularized projection:

[0087]

[0088] U = [M1, ....., Mk], using clipping to force the resulting values ​​in M(α) to be between 0 and 1. In the M step, the parameters of the model are updated.

[0089] The update is intuitive. The class mean and variance estimates are computed as weighted averages:

[0090]

[0091] Due to the presence of large intensity pathological changes and severe artifacts, the image inhomogeneity cannot be corrected by the preprocessing step. In order to address the image inhomogeneity of healthy tissue, the intensity average estimate is modeled as spatial variation and a low-pass filter G is introduced. H To enhance spatial smoothness, similar to the original EM segmentation formula. Specifically:

[0092] μ H ←G H *(w x (H)·I) (10)

[0093] Where * represents spatial convolution. The previous healthy tissue β x A small fraction of current estimates of stroke and healthy tissue probability:

[0094]

[0095] Finally, after weighted adjustment with neighboring voxels, we get the variational posterior parameter w x

[0096]

[0097] where π x (c) is defined in (3). The update is performed iteratively until the parameter estimates converge.

[0098] As a preference in this embodiment, the image segmentation model uses manually labeled leukoaraiosis lesion results as training images.

[0099] like Figure 2 As shown, the cerebrovascular lesion segmentation device 200 in the embodiment of the present application includes:

[0100] An acquisition module 210 is used to acquire MRI image information;

[0101] a processing module 220 configured to obtain a segmentation result from the MRI image information using a pre-trained image segmentation model, wherein the image segmentation model uses a maximum a posteriori probability model to perform MAP estimation on a posterior distribution model used to describe tissue categories and a spatial prior distribution of leukoaraiosis tissue;

[0102] The segmentation module 230 is configured to obtain a segmentation result for separating the leukoaraiosis lesions from the acute and chronic stroke lesions in the cerebrovascular lesions according to the segmentation result.

[0103] In the acquisition module 210 of the embodiment of the present application, the MRI image information generally includes the lesion area, that is, the lesion area of ​​the cerebral blood vessels.

[0104] If it is the training phase, the spatial distribution, shape and appearance of healthy tissue and cerebrovascular lesions need to be included.

[0105] If it is the testing phase, only the MRI image of the lesion area needs to be input.

[0106] In the processing module 220 of the embodiment of the present application, the segmentation result of the MRI image information can be obtained by using the pre-trained image segmentation model. The pre-trained image segmentation model is based on a probability estimation model, and the corresponding segmentation result can be obtained according to the estimation result.

[0107] Exemplarily, the image segmentation model uses a maximum a posteriori probability model to perform maximum a posteriori probability estimation on the posterior distribution model used to describe the tissue category and the spatial prior distribution of the leukoaraiosis tissue.

[0108] In the embodiment of the present application, the segmentation module 230 can obtain a segmentation result based on the segmentation result to separate the leukoaraiosis lesions in the cerebrovascular lesions from the acute and chronic stroke lesions. By segmenting the individual cerebrovascular lesions in the MRI image information, the expert's knowledge of the disease is captured based on the probabilistic model.

[0109] Furthermore, by modeling the spatial distribution of leukoaraiosis lesions and the intensity of leukoaraiosis and stroke lesions, this method automatically segments tissue that is indistinguishable based on intensity alone. Using a stroke model that combines intensity and spatial context can accurately segment leukoaraiosis lesions from acute and chronic stroke lesions.

[0110] like Figure 3 FIG. 1 is a schematic diagram showing the implementation principle of the cerebrovascular lesion segmentation method according to an embodiment of the present application, which specifically includes:

[0111] S1, input MRI image.

[0112] S2, Model describing the posterior distribution of tissue types.

[0113] S3, Spatial prior distribution of leukoaraiosis tissue.

[0114] S4, MAP inference.

[0115] S5, test.

[0116] S6, output.

[0117] The specific implementation used results from at least 100 manually delineated leukoaraiosis test images, as well as six additional test volumes, each containing leukoaraiosis lesion regions manually delineated by multiple experts. In this example, the segmentation algorithm was run only in the white matter, where the majority of leukoaraiosis and stroke lesions are expected to be seen.

[0118] In the examples of this application, the scans included T2-FLAIR scans (1 x 1 mm in-plane, slice thickness 5-7 mm, sometimes using a PROPELLER sequence when the patient was moving). TR and TE images were acquired, and T1 images were obtained for each subject and registered to an atlas template using ANTs parameters.

[0119] Parameters: The PCA shape model ({M k The fixed parameters λ and A were manually chosen to optimize the results on a single test example. Specifically, we used λ = 250, A(c,c) = 100 for c∈{L,S,H}, A(L,H) = 97, A(S,L) = 1, and A(S,H) = 20. A simple threshold classifier learned from the training set was used to initialize the posterior estimate.

[0120] An embodiment of the present application further provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps of any of the above method embodiments when run.

[0121] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps:

[0122] S1, obtain MRI image information;

[0123] S2, obtaining a segmentation result in the MRI image information by using a pre-trained image segmentation model, wherein the image segmentation model uses a maximum a posteriori probability model to perform MAP estimation on a posterior distribution model used to describe tissue categories and a spatial prior distribution of leukoaraiosis tissue;

[0124] S3. Obtaining a segmentation result for separating the leukoaraiosis lesions from the acute and chronic stroke lesions in the cerebrovascular lesions based on the segmentation result.

[0125] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store computer programs.

[0126] An embodiment of the present application further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0127] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0128] Optionally, in this embodiment, the processor may be configured to execute the following steps through a computer program:

[0129] S1, obtain MRI image information;

[0130] S2, obtaining a segmentation result in the MRI image information by using a pre-trained image segmentation model, wherein the image segmentation model uses a maximum a posteriori probability model to perform MAP estimation on a posterior distribution model used to describe tissue categories and a spatial prior distribution of leukoaraiosis tissue;

[0131] S3. Obtaining a segmentation result for separating the leukoaraiosis lesions from the acute and chronic stroke lesions in the cerebrovascular lesions based on the segmentation result.

[0132] Optionally, specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0133] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A cerebrovascular lesion segmentation method, characterized in that: The method comprises: Acquire MRI image information; Obtaining a segmentation result in the MRI image information by using a pre-trained image segmentation model, wherein the image segmentation model uses a maximum a posteriori probability model to perform MAP estimation on a posterior distribution model used to describe tissue categories and a spatial prior distribution of leukoaraiosis tissue; Obtaining a segmentation result for separating the leukoaraiosis lesions from the acute and chronic stroke lesions in the cerebrovascular lesions according to the segmentation result; The image segmentation model includes intensity features, shape features, and spatial environment features, wherein the intensity features include the intensity distribution of leukoaraiosis and the intensity distribution of stroke, and the spatial environment features include the spatial distribution of leukoaraiosis. The segmentation result also includes: Based on the T2 FLAIR sequence, the leukoaraiosis lesions in the cerebrovascular lesions were separated from the acute and chronic stroke lesions.

2. The method according to claim 1, characterized in that The segmentation results in the MRI image information include: Generative models are used to describe the spatial distribution, shape, and appearance of healthy tissue and cerebrovascular lesions, and a posterior distribution model is established to describe tissue categories, where the prior of the tissue category captures the knowledge of spatial distribution and lesion shape; A maximum a posteriori probability model is used to perform MAP estimation on the posterior distribution model used to describe the tissue category to obtain a segmentation result in the MRI image information.

3. The method according to claim 1, characterized in that The segmentation results in the MRI image information include: PCA was used to construct a probability map for the training set of binary segmentation mapping of manually segmented white matter loose tissue lesions, and the spatial range of white matter loose tissue lesions was modeled to obtain the spatial prior distribution of white matter loose tissue.

4. The method according to claim 1, wherein The MAP estimation is based on the EM algorithm, and the intensity average estimation is modeled as spatial variation and filtered using a low-pass filter.

5. The method according to claim 1, wherein The image segmentation model uses manually annotated leukoaraiosis lesion results as training images.

6. A clinical image processing method, characterized in that: Image registration is performed using the cerebrovascular lesion segmentation method according to any one of claims 1 to 5.

7. A cerebrovascular lesion segmentation device, characterized in that: The device comprises: An acquisition module, used for acquiring MRI image information; a processing module, configured to obtain a segmentation result in the MRI image information by using a pre-trained image segmentation model, wherein the image segmentation model uses a maximum a posteriori probability model to perform MAP estimation on a posterior distribution model used to describe tissue categories and a spatial prior distribution of leukoaraiosis tissue; a segmentation module, configured to obtain, based on the segmentation results, a segmentation result for separating the leukoaraiosis lesions from the acute and chronic stroke lesions in the cerebrovascular lesions; The image segmentation model includes intensity features, shape features, and spatial environment features, wherein the intensity features include the intensity distribution of leukoaraiosis and the intensity distribution of stroke, and the spatial environment features include the spatial distribution of leukoaraiosis. The segmentation result also includes: Based on the T2 FLAIR sequence, the leukoaraiosis lesions in the cerebrovascular lesions were separated from the acute and chronic stroke lesions.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 5 and / or the method according to claim 6 when executed.

9. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 5, and / or the method according to claim 6.

Images