Brain region grey matter layering method, device and equipment based on diffusion magnetic resonance imaging

Abstract

The invention provides a brain region grey matter layering method, device and equipment based on diffusion magnetic resonance imaging, and can be applied to the technical field of neural image calculation. The method comprises the following steps: acquiring diffusion magnetic resonance imaging data corresponding to a target brain area of a target object; processing the signal feature data by using an encoder to obtain encoding feature vectors corresponding to the N grey matter areas respectively; for the nth grey matter area of the N grey matter areas, clustering at least one voxel located on the nth grey matter area based on a preset clustering number k and the coding feature vector to obtain an nth clustering result; for the mth cluster of the k clusters, determining sorting information of the mth cluster according to the position information of voxels in the mth cluster, the position information of the first interface and the position information of the second interface; and sorting the k clustering clusters according to the sorting information of the k clustering clusters to obtain a brain region grey matter layering result for the target object.

Description

Technical Field

[0001] The present disclosure relates to the field of neuroimaging computing technologies, and more particularly to a method, apparatus, and device for gray matter stratification of brain regions based on diffusion magnetic resonance imaging. Background Art

[0002] In neuroimaging, ideally, the gray matter neurons of the human brain can be divided into a six-layer stratification structure. Different layers of gray matter have different types and numbers of nerve cells and have different functions. Existing research has shown that some neurodegenerative diseases are related to specific gray matter layers. Therefore, studying the stratification structure of the human brain gray matter helps to promote the development of prevention, diagnosis, and treatment technologies for neurodegenerative diseases.

[0003] Existing in-vivo gray matter stratification techniques use the apparent diffusion coefficient (ADC) for clustering. However, the ADC parameter has limited ability to characterize the movement of water molecules under the condition of water molecule restriction and blockage, resulting in poor stratification effects. Summary of the Invention

[0004] In view of the above problems, the present disclosure provides a method, apparatus, and device for gray matter stratification of brain regions based on diffusion magnetic resonance imaging.

[0005] According to a first aspect of the present disclosure, there is provided a method for gray matter stratification of brain regions based on diffusion magnetic resonance imaging, including: obtaining diffusion magnetic resonance imaging data corresponding to a target brain region of a target object, where the target brain region includes N gray matter regions, and the diffusion magnetic resonance imaging data includes signal feature data of at least one voxel corresponding to each of the N gray matter regions; processing the signal feature data by an encoder to obtain an encoded feature vector corresponding to each of the N gray matter regions; for the nth gray matter region of the N gray matter regions, clustering at least one voxel located on the nth gray matter region based on a preset number of clusters k and the encoded feature vector to obtain an nth clustering result, where the nth clustering result includes k clusters, each cluster represents a layer in the nth gray matter region, each cluster includes at least one voxel, and n N; for the mth cluster of the k clusters, determining sorting information of the mth cluster according to the position information of the voxels in the mth cluster, the position information of a first interface, and the position information of a second interface, where the first interface represents the interface between the gray matter region where the voxel is located and an associated white matter region, and the second interface represents the interface between the gray matter region where the voxel is located and an associated cerebrospinal fluid region; sorting the k clusters according to the sorting information of each of the k clusters to obtain a gray matter stratification result of the brain region of the target object.

[0006] According to an embodiment of the present disclosure, the sorting information includes an average depth ratio; wherein, for the m-th clustering cluster among k clustering clusters, determining the sorting information of the m-th clustering cluster according to the position information of the voxels in the m-th clustering cluster, the position information of the first interface, and the position information of the second interface includes: for each voxel in the m-th clustering cluster, determining a first distance from the voxel to the first interface according to the position information of the voxel and the position information of the first interface; determining a second distance between the first interface and the second interface according to the position information of the first interface and the position information of the second interface; obtaining a depth ratio of the voxel according to the ratio between the first distance and the second distance; and averaging the depth ratios corresponding to multiple voxels to obtain the m-th average depth ratio.

[0007] According to an embodiment of the present disclosure, the position information of the first interface and the position information of the second interface are determined based on the following operations: performing a first masking process on the tissue structure distribution information corresponding to the brain region of the target object to generate a first brain mask image, where the first brain mask image is used to distinguish the gray matter region, the white matter region, and the cerebrospinal fluid region; and determining the position information of the first interface between the gray matter region and the white matter region and the position information of the second interface between the gray matter region and the cerebrospinal fluid region based on the first brain mask image.

[0008] According to an embodiment of the present disclosure, clustering at least one voxel located on the n-th gray matter region based on a preset number of clusters k and an encoded feature vector to obtain the n-th clustering result includes: initializing k initial clustering centers according to the preset number of clusters k; clustering at least one voxel based on the distance between the encoded feature vector located on the n-th gray matter region and the initial clustering centers, and determining initial clusters corresponding to the k initial clustering centers respectively; iteratively updating the initial clustering centers of the initial clustering clusters according to multiple encoded feature vectors in the initial clustering clusters to obtain the clustering centers after iteration of the initial clustering clusters; and determining the n-th clustering result based on the clustering centers after iteration and a preset iteration stop condition.

[0009] According to an embodiment of the present disclosure, determining the n-th clustering result based on the clustering centers after iteration and a preset iteration stop condition includes: determining a difference value between the clustering centers after iteration and the clustering centers before iteration; in the case where the difference value satisfies the preset iteration stop condition, determining the clustering centers before iteration as the target clustering centers; and determining the n-th clustering result according to the clusters corresponding to the target clustering centers respectively.

[0010] According to an embodiment of the present disclosure, obtaining diffusion magnetic resonance imaging data corresponding to a target brain region of a target object includes: performing spatial position registration on initial diffusion magnetic resonance imaging data based on tissue structure distribution information corresponding to the brain region of the target object to obtain registered diffusion magnetic resonance imaging data; performing a second masking process on the tissue structure distribution information to generate a second brain mask image, where the second brain mask image is used to distinguish the target brain region and non-target brain regions; and obtaining diffusion magnetic resonance imaging data corresponding to the target brain region of the target object according to the second brain mask image and the registered diffusion magnetic resonance image.

[0011] According to an embodiment of the present disclosure, the encoder is obtained based on the following training operations: obtaining training samples, where the training samples include sample signal feature data; processing the sample signal feature data using an initial encoder to obtain sample encoded feature vectors; processing the sample encoded feature vectors using an initial decoder to obtain sample decoded feature vectors; and training the initial encoder according to the sample signal feature data and the sample decoded feature vectors to obtain the encoder.

[0012] According to an embodiment of the present disclosure, training the initial encoder according to the sample signal feature data and the sample decoded feature vectors to obtain the encoder includes: calculating a loss value between the sample signal feature data and the sample decoded feature vectors using a loss function to obtain a target loss value; and training the initial encoder according to the target loss value to obtain the encoder.

[0013] According to a second aspect of the present disclosure, there is provided a device for gray matter stratification of a brain region based on diffusion magnetic resonance imaging, including: an acquisition module configured to acquire diffusion magnetic resonance imaging data corresponding to a target brain region of a target object, where the target brain region includes N gray matter regions, and the diffusion magnetic resonance imaging data includes signal feature data of at least one voxel corresponding to each of the N gray matter regions; an encoding module configured to process the signal feature data using an encoder to obtain encoded feature vectors corresponding to each of the N gray matter regions; and a stratification module configured to, for the nth gray matter region of the N gray matter regions, perform clustering on at least one voxel located in the nth gray matter region based on a preset number of clusters k and the encoded feature vectors to obtain an nth clustering result, where the nth clustering result includes k clustering clusters, each clustering cluster represents a layer in the nth gray matter region, each clustering cluster includes at least one voxel, n N; a determining module, configured to determine sorting information of the m-th clustering cluster among the k clustering clusters according to the position information of the voxels in the m-th clustering cluster, the position information of the first interface, and the position information of the second interface, where the first interface represents the interface between the gray matter region where the voxels are located and the associated white matter region, and the second interface represents the interface between the gray matter region where the voxels are located and the associated cerebrospinal fluid region; a sorting module, configured to sort the k clustering clusters according to the sorting information of each of the k clustering clusters to obtain a gray matter stratification result of the brain region for the target object.

[0014] A third aspect of the present disclosure provides an electronic device, including: one or more processors; a memory, configured to store one or more programs, where when the one or more programs are executed by the one or more processors, the one or more processors are caused to execute the above-mentioned gray matter stratification method of the brain region.

[0015] According to the gray matter stratification method, device and equipment of the brain region based on diffusion magnetic resonance imaging provided by the present disclosure, by using an encoder to process signal feature data, N encoded feature vectors corresponding to each gray matter region are obtained; for the n-th gray matter region, at least one voxel located on the n-th gray matter region is clustered based on a preset number of clusters k and the encoded feature vector to obtain an n-th clustering result; the k clustering clusters are sorted according to the sorting information of each of the k clustering clusters to obtain a gray matter stratification result of the brain region for the target object. Since compression is performed through the encoder to obtain encoded feature vectors suitable for clustering processing, and then the clustering algorithm is quickly called to cluster the voxels on each gray matter region, thereby realizing the stratification of the gray matter region, the technical problem that the ADC parameter is limited in movement ability under the condition of water molecule blockage, resulting in poor stratification effect when analyzing diffusion magnetic resonance imaging data depending on ADC parameter stratification, is overcome. The encoding process and the clustering process reduce the parameter scale, time cost and operation overhead, and improve the calculation efficiency; in addition, the k clustering clusters are sorted according to the sorting information of the clustering clusters, so as to determine the relative positions between the layers, and the accuracy of stratification is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Through the following description of the embodiments of the present disclosure with reference to the drawings, the above and other objects, features and advantages of the present disclosure will become clearer.

[0017] Figure 1 The flowchart of the gray matter stratification method of the brain region based on diffusion magnetic resonance imaging according to the embodiment of the present disclosure is shown.

[0018] Figure 2 An example schematic diagram of the gray matter stratification result of the brain region according to the embodiment of the present disclosure is shown.

[0019] Figure 3 An example schematic diagram of the encoder and decoder according to the embodiment of the present disclosure is shown.

[0020] Figure 4A Shows an exemplary schematic diagram of the interpretability analysis of the parametric feature vector obtained from the existing computational model according to an embodiment of the present disclosure.

[0021] Figure 4B Shows an exemplary schematic diagram of the interpretability analysis of the encoded feature vector generated by the encoder according to an embodiment of the present disclosure.

[0022] Figure 4C Shows an exemplary schematic diagram of the correlation analysis based on the 2-layer gray matter region segmentation according to an embodiment of the present disclosure.

[0023] Figure 4D Shows an exemplary schematic diagram of the correlation analysis based on the 3-layer gray matter region segmentation according to an embodiment of the present disclosure.

[0024] Figure 5 Shows a structural block diagram of a brain region gray matter stratification device based on diffusion magnetic resonance imaging according to an embodiment of the present disclosure.

[0025] Figure 6 Shows a block diagram of an electronic device suitable for implementing a method for stratifying gray matter in brain regions based on diffusion magnetic resonance imaging according to an embodiment of the present disclosure. Detailed implementation manners

[0026] Hereinafter, embodiments according to the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for the sake of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present disclosure.

[0027] The terms used herein are merely for describing specific embodiments and are not intended to limit the present disclosure. The terms "including", "comprising", etc. used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0028] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0029] In the case of using expressions such as "at least one of A, B, and C", generally, it should be interpreted according to the meaning that those skilled in the art usually understand this expression (for example, "a system having at least one of A, B, and C" should include but not be limited to a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0030] In the process of implementing the present disclosure, it is found that existing gray matter stratification methods usually use a diffusion weighted imaging (DWI) model to analyze diffusion magnetic resonance imaging data to obtain an apparent diffusion coefficient (ADC) for gray matter stratification. The ADC parameter models the diffusion of water molecules in tissues as Gaussian diffusion, thereby reflecting the diffusion properties of tissues. In regions with a higher cell density and a smaller extracellular space, the diffusion of water molecules is more restricted, and the ADC value is lower; conversely, in regions with a lower cell density and a larger extracellular space, the diffusion of water molecules is relatively free, and the ADC value is higher. Based on this principle, gray matter stratification can be achieved by measuring and analyzing the spatial distribution of ADC values. However, due to the limitation and blockage of cell membranes, water molecules in tissues do not always exhibit free Gaussian diffusion, but rather non-Gaussian diffusion and Gaussian diffusion. Therefore, in an environment with a high neuron cell density such as the human brain gray matter, ADC does not always well characterize the microscopic structural properties of tissues, and its accuracy is affected by cell density and cell membrane permeability.

[0031] In view of this, according to an embodiment of the present disclosure, a method, apparatus, and device for gray matter stratification of a brain region based on diffusion magnetic resonance imaging are provided. The method includes: obtaining diffusion magnetic resonance imaging data corresponding to a target brain region of a target object; using an encoder to process signal feature data to obtain encoded feature vectors corresponding to N gray matter regions respectively; for the nth gray matter region among the N gray matter regions, clustering at least one voxel located on the nth gray matter region based on a preset number of clusters k and the encoded feature vector to obtain an nth clustering result; for the mth clustering cluster among the k clustering clusters, determining sorting information of the mth clustering cluster according to the position information of the voxels in the mth clustering cluster, the position information of a first interface, and the position information of a second interface; and sorting the k clustering clusters according to the sorting information of each of the k clustering clusters to obtain a gray matter stratification result of the brain region of the target object.

[0032] In the technical solution of the present disclosure, the user information involved (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) are all information and data authorized by the user or fully authorized by all parties. Moreover, the processing of relevant data, such as collection, storage, use, processing, transmission, provision, disclosure, and application, complies with relevant laws, regulations, and standards, takes necessary confidentiality measures, does not violate public order and good customs, and provides corresponding operation entrances for users to choose to authorize or refuse.

[0033] It should be noted that the serial numbers of each operation in the following methods are only used as representations of the operation for description and should not be regarded as indicating the execution order of each operation. Unless explicitly stated, the method does not need to be executed exactly in the order shown.

[0034] Figure 1 The flowchart of a method for gray matter stratification of a brain region based on diffusion magnetic resonance imaging according to an embodiment of the present disclosure is shown.

[0035] As Figure 1 shown, the method 100 includes operation S110 to operation S150.

[0036] In operation S110, diffusion magnetic resonance imaging data corresponding to a target brain region of a target object is acquired.

[0037] According to an embodiment of the present disclosure, the target object is an object for which gray matter stratification of the brain region to be tested is to be performed.

[0038] According to an embodiment of the present disclosure, the target brain region represents a brain parenchymal tissue region, such as a central nervous system region.

[0039] According to an embodiment of the present disclosure, the target brain region includes N gray matter regions. The gray matter region refers to a region in the brain where nerve cells are concentrated. The nerve cells in the gray matter region are connected to other nerve cells or muscle cells through synapses to form a complex neural network, which is responsible for processing sensory inputs from various parts of the body and also controls the motor output of the body. Different layers of each gray matter region have different types and numbers of nerve cells and have different functions.

[0040] According to an embodiment of the present disclosure, the diffusion magnetic resonance imaging data is Diffusion Magnetic Resonance Imaging (DMRI). The target object is scanned by an imaging device and then the diffusion magnetic resonance imaging data corresponding to the target brain region is extracted. The diffusion magnetic resonance imaging data is four-dimensional data. The four-dimensional data includes the spatial position information of the voxels on the unit volume of the target brain region. Each voxel corresponds to a signal feature data, and the four-dimensional data also includes the length information of the signal feature data.

[0041] According to an embodiment of the present disclosure, the signal feature data characterizes the microscopic motion data of water molecules in tissue, and the characteristics of the diffusion process of water molecules in tissue can reflect rich information about the tissue microstructure.

[0042] In operation S120, the encoder is used to process the signal feature data to obtain the encoded feature vectors corresponding to each of the N gray matter regions.

[0043] According to an embodiment of the present disclosure, the encoder is used to perform compression encoding on the signal feature data of each voxel in the target brain region to obtain the encoded feature vector corresponding to each voxel. The encoded feature vector is rich signal information after dimensionality reduction.

[0044] According to an embodiment of the present disclosure, for each of the N gray matter regions, a vector matrix is formed according to the encoded feature vectors corresponding to the multiple voxels in the gray matter region. The vector matrix is normalized, and each encoded feature vector in the normalized vector matrix is a matrix element.

[0045] In operation S130, for the nth gray matter region among the N gray matter regions, at least one voxel located in the nth gray matter region is clustered based on the preset number of clusters k and the encoded feature vectors to obtain the nth clustering result.

[0046] According to an embodiment of the present disclosure, the preset number of clusters k and k initial cluster centers are initialized, and clustering at least one voxel in the nth gray matter region may include: using the K-means clustering algorithm to process the vector matrix of the nth gray matter region, thereby calculating the distances from each encoded feature vector in the vector matrix to the respective k initial cluster centers, and iteratively updating the cluster centers until the iteration stop condition is satisfied to obtain the nth clustering result corresponding to the nth gray matter region.

[0047] According to an embodiment of the present disclosure, the nth clustering result includes k clusters, each cluster represents a layer in the nth gray matter region, and each cluster includes at least one voxel, n N.

[0048] For example, the first clustering result includes 3 clusters, representing that the first gray matter region is divided into 3 layers. Each cluster represents a layer in the first gray matter region. The first cluster includes voxels 1 and 3; the second cluster includes voxels 4 and 5; the third cluster includes voxels 2 and 6. Layer labels are assigned to each voxel. Voxels 1 and 3 are located in layer 11, voxels 4 and 5 are located in layer 12, and voxels 2 and 6 are located in layer 13.

[0049] In operation S140, for the m-th cluster among the k clusters, sorting information of the m-th cluster is determined according to the position information of the voxels in the m-th cluster, the position information of the first interface, and the position information of the second interface.

[0050] According to an embodiment of the present disclosure, the first interface represents the interface between the gray matter region where the voxel is located and the associated white matter region, and the second interface represents the interface between the gray matter region where the voxel is located and the associated cerebrospinal fluid region.

[0051] According to an embodiment of the present disclosure, the associated white matter region adjacent to the n-th gray matter region and the associated cerebrospinal fluid region adjacent to the n-th gray matter region are determined according to the n-th gray matter region where the voxel is located, so as to determine the first interface and the second interface.

[0052] According to an embodiment of the present disclosure, the m-th cluster includes at least one voxel. Depth information of each voxel in the target brain region is determined according to the position information of each voxel, the position information of the first interface, and the position information of the second interface; the sorting information of the m-th cluster is determined according to the respective depth information of the at least one voxel.

[0053] According to an embodiment of the present disclosure, the sorting information of the m-th cluster represents the relative position of the m-th cluster among the k clusters, so as to determine the relative position of the layer corresponding to the m-th cluster in the n-th gray matter region among the k layers.

[0054] In operation S150, the k clusters are sorted according to the respective sorting information of the k clusters to obtain a gray matter stratification result of the brain region for the target object.

[0055] According to an embodiment of the present disclosure, for the n-th gray matter region, the k clusters are sorted in descending order according to the respective sorting information of the k clusters to obtain a gray matter stratification result of the brain region of the n-th gray matter region. The layer corresponding to the cluster in the first position is used as the first layer, and the layer corresponding to the cluster in the last position is used as the last layer.

[0056] According to an embodiment of the present disclosure, the gray matter stratification results of each gray matter region are integrated to obtain a gray matter stratification result of the brain region for the target object. The gray matter stratification result of the target object is a hierarchical result with relative position relationships.

[0057] For example, the target brain region includes two gray matter regions. The first gray matter region is divided into layer 11, layer 12, and layer 13, and the second gray matter region is divided into layer 21, layer 22, and layer 23. After sorting, the gray matter stratification result of the first gray matter region is layer 12, layer 11, and layer 13; the gray matter stratification result of the second gray matter region is layer 23, layer 21, and layer 22. According to the gray matter stratification results of the first gray matter region and the second gray matter region respectively, the gray matter stratification result of the target brain region is layer 1, layer 2, and layer 3. Among them, layer 1 includes layer 12 and layer 23, layer 2 includes layer 11 and layer 21, and layer 3 includes layer 13 and layer 22.

[0058] According to an embodiment of the present disclosure, the process of calculating the gray matter stratification result of the target object using the method of the present application is about 35 minutes. The calculation time of the existing method is about 4 hours or even higher.

[0059] According to an embodiment of the present disclosure, since compression is performed by the encoder to obtain an encoded feature vector suitable for clustering processing, and then the clustering algorithm is quickly called to cluster the voxels on each gray matter region to achieve accurate stratification of the gray matter region, it overcomes the technical problem that when analyzing diffusion magnetic resonance imaging data and relying on ADC parameter stratification, the movement ability of the ADC parameter is limited under the condition of water molecule blockage, resulting in poor stratification effect. The encoding process and the clustering process reduce the parameter scale, time cost, and operation overhead, and improve the calculation efficiency; in addition, according to the sorting information of the clustering clusters, the K clustering clusters are sorted to determine the relative positions between the layers, improving the accuracy of stratification.

[0060] According to an embodiment of the present disclosure, the position information of the first interface and the position information of the second interface are determined based on the following operations: performing a first masking process on the organizational structure distribution information corresponding to the brain region of the target object to generate a first brain mask image, where the first brain mask image is used to distinguish the gray matter region, the white matter region, and the cerebrospinal fluid region; based on the first brain mask image, determining the position information of the first interface between the gray matter region and the white matter region and the position information of the second interface between the gray matter region and the cerebrospinal fluid region.

[0061] According to an embodiment of the present disclosure, the organizational structure distribution information may be a T1-weighted image (T1WI), which mainly reflects the longitudinal relaxation time of hydrogen protons (1H) in tissues. In the T1-weighted image, different tissues exhibit different signal intensities according to their different T1 longitudinal relaxation times.

[0062] According to an embodiment of the present disclosure, the BET function (Brain Extraction Tool) of the FSL tool (Functional Magnetic Resonance Imaging of the Brain Software Library) is used to strip the organizational structure distribution information of the brain region, removing irrelevant tissues such as the skull, skin, and eyeballs, and obtaining the stripped organizational structure distribution information.

[0063] According to an embodiment of the present disclosure, the FAST function (FMRIB’s Automated Segmentation Tool) of the FSL tool (Functional Magnetic Resonance Imaging of the Brain Software Library) is used to segment the stripped organizational structure distribution information, segmenting the stripped brain region into a gray matter region, a white matter region, and a cerebrospinal fluid region, and then performing a first masking process to generate a first brain mask image.

[0064] According to an embodiment of the present disclosure, the first brain mask image includes the voxel signal intensity information of the gray matter region, the white matter region, and the cerebrospinal fluid region respectively.

[0065] According to an embodiment of the present disclosure, each gray matter region and the adjacent white matter region and cerebrospinal fluid region are determined from the first brain mask image, so as to determine the position information of the first interface formed between the gray matter region and the adjacent white matter region and the position information of the second interface formed between the gray matter region and the adjacent cerebrospinal fluid region.

[0066] According to an embodiment of the present disclosure, the sorting information includes an average depth ratio; wherein, for the m-th cluster of k clusters, the sorting information of the m-th cluster is determined according to the position information of the voxels in the m-th cluster, the position information of the first interface, and the position information of the second interface, including: for each voxel in the m-th cluster, according to the position information of the voxel and the position information of the first interface, determining the first distance from the voxel to the first interface; according to the position information of the first interface and the position information of the second interface, determining the second distance between the first interface and the second interface; obtaining the depth ratio of the voxel according to the ratio between the first distance and the second distance; and averaging the depth ratios corresponding to multiple voxels to obtain the m-th average depth ratio.

[0067] According to an embodiment of the present disclosure, a first interface and a second interface corresponding to the gray matter region are determined according to the gray matter region where the voxel is located. According to the position information of the voxel and the position information of the first interface, a first distance from the voxel to the first interface is calculated; according to the position information of the first interface and the position information of the second interface, a second distance between the first interface and the second interface is calculated.

[0068] According to an embodiment of the present disclosure, based on the ratio between the first distance and the second distance, a depth ratio of the voxel is obtained. The depth ratio of the voxel characterizes the depth of the voxel in the gray matter region.

[0069] According to an embodiment of the present disclosure, the average values of the depth ratios of the multiple voxels in the m-th clustering cluster are calculated to obtain the m-th average depth ratio corresponding to the m-th clustering cluster. The larger the m-th average depth ratio, the closer the layer corresponding to the m-th clustering cluster is to the first interface, and the smaller the m-th average depth ratio, the closer the layer corresponding to the m-th clustering cluster is to the second interface.

[0070] According to an embodiment of the present disclosure, k average depth ratios are obtained according to k clustering clusters. The layer corresponding to the largest average depth ratio is ranked as the first layer, and the layer corresponding to the smallest average depth ratio is ranked as the last layer.

[0071] According to an embodiment of the present disclosure, due to the randomness of the initialization of the clustering center of the clustering algorithm, the layers segmented in each gray matter region are not arranged in order from the surface layer (the second interface between the gray matter region and the cerebrospinal fluid region) to the deep layer (the first interface between the gray matter region and the white matter region). Therefore, sorting the layers according to the average depth ratio to determine the relative positions between the layers improves the accuracy of the layering.

[0072] Figure 2 An exemplary schematic diagram showing the gray matter layering result according to an embodiment of the present disclosure is shown.

[0073] As Figure 2 shown, the layering result 210 of the voxels in the gray matter region is determined according to the clustering result corresponding to each gray matter region. The layering result 210 includes the segmented layers, and there is no order between the layers. For each gray matter region, the average depth ratio corresponding to each of the k clustering clusters is determined, and the k clustering clusters are sorted according to the average depth ratio of each of the k clustering clusters, so as to sort the segmented layers to obtain the sorted gray matter layering result 220 of the brain region. Sorting from the surface layer to the deep layer, combined with the chromaticity reference diagram 230, red represents the first layer, yellow represents the second layer, and white represents the third layer.

[0074] According to an embodiment of the present disclosure, clustering at least one voxel located on the nth gray matter region based on a preset number of clusters k and encoded feature vectors to obtain the nth clustering result includes: initializing k initial cluster centers according to the preset number of clusters k; clustering at least one voxel based on the distance between the encoded feature vectors located on the nth gray matter region and the initial cluster centers, and determining initial clusters corresponding to the k initial cluster centers respectively; iteratively updating the initial cluster centers of the initial clusters according to the multiple encoded feature vectors in the initial clusters to obtain the cluster centers after iteration of the initial clusters; and determining the nth clustering result based on the cluster centers after iteration and a preset iteration stop condition.

[0075] According to an embodiment of the present disclosure, randomly select k elements from the vector matrix as k initial cluster centers according to the preset number of clusters k, calculate the distances from the encoded feature vectors to each initial cluster center to obtain k distances, and classify the voxel corresponding to this encoded feature vector into the cluster where the initial cluster center corresponding to the minimum distance among the k distances is located, and traverse each encoded feature vector, so as to obtain the initial clusters corresponding to the k initial cluster centers respectively.

[0076] In one embodiment, the distance from the encoded feature vector x to the initial cluster center y As shown in formula (1):

[0077] (1);

[0078] Wherein, represents the i-th component in the encoded feature vector x, the vector length of the encoded feature vector is m, represents the i-th component in the initial cluster center y.

[0079] According to an embodiment of the present disclosure, for each initial cluster, re-determine the cluster center after iteration of this initial cluster.

[0080] In one embodiment, the cluster center after iteration of the initial cluster As shown in formula (2):

[0081] (2);

[0082] Wherein, represents the i-th component in the cluster center after iteration c represents the number of voxels included in the initial cluster, represents the i-th component in the encoded feature vector corresponding to the j-th voxel.

[0083] According to an embodiment of the present disclosure, the preset iteration stop condition may be that the cluster centers no longer change or the change value is less than a given threshold. When the preset iteration stop condition is satisfied, the iterative clustering is stopped, so as to determine the final k cluster centers and the clusters corresponding to the cluster centers. The clustering result includes k cluster centers and the clusters corresponding to the cluster centers.

[0084] According to an embodiment of the present disclosure, density clustering algorithms, hierarchical clustering algorithms, etc. may also be used in the clustering process.

[0085] According to an embodiment of the present disclosure, determining the nth clustering result based on the iterated cluster centers and the preset iteration stop condition includes: determining the difference value between the iterated cluster centers and the cluster centers before iteration; when the difference value satisfies the preset iteration stop condition, determining the cluster centers before iteration as the target cluster centers; and determining the nth clustering result according to the clusters corresponding to the respective target cluster centers.

[0086] According to an embodiment of the present disclosure, for each cluster, calculate the difference value between the iterated cluster center and the cluster center before iteration.

[0087] According to an embodiment of the present disclosure, when the difference value satisfies the preset iteration stop condition, stop iterating the cluster centers, and determine the cluster centers before iteration as the target cluster centers of this cluster.

[0088] According to an embodiment of the present disclosure, the nth clustering result includes k target cluster centers and the clusters corresponding to the respective k target cluster centers.

[0089] According to an embodiment of the present disclosure, by quickly calling a clustering algorithm to implement the stratification of the gray matter region cortex, the technical problem that the stratification effect is poor due to the limited movement ability of water molecules under the blocking of ADC parameters or FA parameters is overcome, the parameter scale is reduced, the calculation efficiency is improved, and at the same time, the accuracy of brain gray matter stratification is improved.

[0090] According to an embodiment of the present disclosure, obtaining the diffusion magnetic resonance imaging data corresponding to the target brain region of the target object includes: performing spatial position registration on the initial diffusion magnetic resonance imaging data based on the tissue structure distribution information corresponding to the brain region of the target object to obtain the registered diffusion magnetic resonance imaging data; performing a second masking process on the tissue structure distribution information to generate a second brain mask image, where the second brain mask image is used to distinguish the target brain region and the non-target brain region; and obtaining the diffusion magnetic resonance imaging data corresponding to the target brain region of the target object according to the second brain mask image and the registered diffusion magnetic resonance image.

[0091] According to an embodiment of the present disclosure, the initial diffusion magnetic resonance imaging data is the diffusion magnetic resonance imaging data of the brain region of the target object. Performing spatial position registration on the initial diffusion magnetic resonance imaging data may include eddy current correction, head motion correction, and within-subject image registration to obtain the registered diffusion magnetic resonance imaging data. Among them, eddy current correction can correct the image distortion caused by eddy currents during magnetic resonance scanning; head motion correction can correct the image rotation and offset caused by the movement of the subject during scanning; within-subject image registration registers the initial diffusion magnetic resonance imaging data to the tissue structure distribution information corresponding to the brain region of the target object, so that the two have the same spatial coordinate system.

[0092] According to an embodiment of the present disclosure, the FSL - BET tool (Functional MRI of the Brain Software Library) is used to strip the tissue structure distribution information of the brain region, remove irrelevant tissues such as the skull, skin, and eyeballs, and then perform a second masking process to generate a second brain mask image.

[0093] According to an embodiment of the present disclosure, the pixel value of the target brain region in the second brain mask image is assigned 1, and the pixel value of the non-target brain region is assigned 0.

[0094] According to an embodiment of the present disclosure, the second brain mask image is multiplied by the registered diffusion magnetic resonance image to obtain the diffusion magnetic resonance imaging data corresponding to the target brain region of the target object.

[0095] According to an embodiment of the present disclosure, after the preprocessing operation of performing spatial position registration on the initial diffusion magnetic resonance imaging data, non-parenchymal voxels are masked to obtain the diffusion magnetic resonance imaging data corresponding to the target brain region, thereby reducing the time overhead of the encoder model calculation.

[0096] According to an embodiment of the present disclosure, the encoder is obtained based on the following training operation: obtaining training samples, where the training samples include sample signal feature data; using the initial encoder to process the sample signal feature data to obtain a sample encoded feature vector; using the initial decoder to process the sample encoded feature vector to obtain a sample decoded feature vector; training the initial encoder according to the sample signal feature data and the sample decoded feature vector to obtain the encoder.

[0097] According to an embodiment of the present disclosure, the initial encoder can be an autoencoder model, and the structures of the initial encoder and the initial decoder are symmetric. The vector length of a single sample signal feature data is usually only 60 to 300, so a very large autoencoder model is not required.

[0098] According to an embodiment of the present disclosure, the training samples include a training set and a test set. The sample signal feature data of 40 sample target objects are used as the training set, and the sample signal feature data of 21 sample target objects are used as the test set, and there is no overlap between the two; the number of training rounds is set to 2048; the optimizer uses the Adaptive Moment Estimation (Adam); the learning rate is set to 1e-4.

[0099] According to an embodiment of the present disclosure, the sample encoding feature vector is the feature vector after dimensionality reduction, and the sample decoding feature vector is the upsampled feature vector after size restoration.

[0100] According to an embodiment of the present disclosure, the loss function is used to calculate the difference between the sample signal feature data and the sample decoding feature vector, and the parameters of the initial encoder are adjusted according to the difference, and iterative training is performed. When the model performance is satisfied, the training is stopped, and the trained encoder and decoder are obtained.

[0101] According to an embodiment of the present disclosure, after the model training is completed, the decoder can preferably restore the original sample signal data from the feature representation, which also reflects that the sample decoding feature vector already contains rich information in the sample signal data. When training for 1000 rounds, the training error converges to 6.869e-05, and the validation error is 6.932e-05, and there is no overfitting phenomenon.

[0102] According to an embodiment of the present disclosure, the initial encoder is trained according to the sample signal feature data and the sample decoding feature vector, and the obtained encoder includes: using the loss function to calculate the loss value between the sample signal feature data and the sample decoding feature vector to obtain the target loss value; training the initial encoder according to the target loss value to obtain the encoder.

[0103] According to an embodiment of the present disclosure, the mean square error function can be used as the loss function. When the target loss value satisfies the preset threshold, the training is stopped to obtain the trained encoder.

[0104] Figure 3 An example schematic diagram of an encoder and a decoder according to an embodiment of the present disclosure is shown.

[0105] As Figure 3 shown, the encoder includes a three-layer fully connected layer, and the dimensions of the fully connected layer are 128, 64, and 32 respectively. The decoder and the encoder have a symmetric structure. The encoder encodes each voxel one by one. The encoder encodes the sample signal feature data with an input length of 143 into a sample encoding feature vector with a length of 10. The decoder restores the sample decoding feature vector with a length of 143 according to the sample encoding feature vector. After the model training is completed, the decoder can restore the original sample signal feature data with a low error according to the sample encoding feature vector.

[0106] According to an embodiment of the present disclosure, interpretability analysis is performed on the encoded feature vectors generated by the encoder. Through Canonical Correlation Analysis (CCA), the parametric feature vectors obtained from the existing computational model are used as the X variable group, and the encoded feature vectors generated by the encoder are used as the Y variable group to study the linear relationship between the parametric feature vectors obtained from the existing computational model and the encoded feature vectors obtained from the encoder. The first canonical correlation coefficient of the X variable group and the Y variable group reaches 0.87666981. On the one hand, it shows that there is a strong linear correlation between the principal components of the encoded feature vectors obtained by the encoder and the principal components of the parametric feature vectors obtained from the existing computational model. On the other hand, it shows that the encoder extracts some non-linear components that cannot be characterized by the existing computational model.

[0107] Figure 4A An example schematic diagram of the interpretability analysis of the parametric feature vectors obtained from the existing computational model according to an embodiment of the present disclosure is shown.

[0108] As Figure 4A shown, the canonical variable coefficients of the X variable group are visualized in the form of a heatmap. The horizontal axis represents the X variables, and the vertical axis represents the canonical variable coefficients. It can be seen that the canonical coefficients of the X variable group are concentrated in a few variables, and there is information redundancy among the variables of the X variable group. The introduction of multiple existing computational models does not introduce more additional information.

[0109] Figure 4B An example schematic diagram of the interpretability analysis of the encoded feature vectors generated by the encoder according to an embodiment of the present disclosure is shown.

[0110] As Figure 4B shown, the canonical variable coefficients of the Y variable group are visualized in the form of a heatmap. The horizontal axis represents the Y variables, and the vertical axis represents the canonical variable coefficients. It can be seen that the distribution of the canonical coefficients of each Y variable in the Y variable group is significantly more uniform, which is beneficial to the clustering and classification processing of the clustering algorithm.

[0111] According to an embodiment of the present disclosure, neurophysiological verification is performed on the stratification results of the gray matter region. BigBrain is a hierarchical atlas constructed based on anatomical data, which provides the average thickness information of each layer of the brain region. The BigBrain atlas is divided into 6 layers. The stratification thickness information provided by this atlas is used as the neurophysiological verification of the stratification results of the gray matter region. The average gray matter region thickness and the average thickness of each stratification in the gray matter region are calculated in each brain region, and the rationality of the stratification in this study is verified through correlation analysis.

[0112] Figure 4CShows an exemplary schematic diagram of correlation analysis based on the segmentation of the gray matter region into two layers according to an embodiment of the present disclosure

[0113] As Figure 4C shown, the vertical axis represents the cortical thickness in the BigBrain atlas, and the horizontal axis represents the cortical thickness in the gray matter region of the target object. By comparing the cortical thickness, the gray matter region of the target object is divided into two layers (k = 2). The first layer (surface layer) of the segmentation corresponds to the first, second, and third layers of the BigBrain atlas (correlation coefficient r = 0.67865, p = 0.00005), and the second layer (deep layer) of the segmentation corresponds to the fourth, fifth, and sixth layers of the BigBrain (r = 0.68635, p = 0.00004). P less than 0.05 indicates correlation. For 99.39% of the samples when divided into two layers, the thicknesses of the first layer and the second layer in the gray matter region are significantly correlated with the thicknesses of the first layer and the second layer in the BigBrain stratification atlas.

[0114] Figure 4D Shows an exemplary schematic diagram of correlation analysis based on the segmentation of the gray matter region into three layers according to an embodiment of the present disclosure.

[0115] As Figure 4D shown, the vertical axis represents the cortical thickness in the BigBrain atlas, and the horizontal axis represents the cortical thickness in the gray matter region of the target object. By comparing the thicknesses, when the gray matter region is divided into three layers (k = 3), the first layer (surface layer) of the segmentation corresponds to the first, second, and third layers of the BigBrain (r = 0.59055, p = 0.00037), the second layer (middle layer) of the segmentation corresponds to the fourth and fifth layers of the BigBrain (r = 0.34196, p = 0.005541), and the third layer (deep layer) of the segmentation corresponds to the sixth layer of the BigBrain (r = 0.64356, p = 0.00007). In the database, for 72.6% of the samples when divided into three layers, the thicknesses of the first layer, the second layer, and the third layer in the gray matter region are significantly correlated with the thicknesses of the first layer, the second layer, and the third layer in the BigBrain stratification atlas.

[0116] Based on the above method for gray matter stratification of brain regions based on diffusion magnetic resonance imaging, the present disclosure also provides a device for gray matter stratification of brain regions based on diffusion magnetic resonance imaging. The following will be combined with Figure 5 to describe this device in detail.

[0117] Figure 5 Shows a structural block diagram of a device for gray matter stratification of brain regions based on diffusion magnetic resonance imaging according to an embodiment of the present disclosure.

[0118] As Figure 5As shown, the gray matter stratification device 500 based on diffusion magnetic resonance imaging in this embodiment includes an acquisition module 510, an encoding module 520, a stratification module 530, a determination module 540, and a sorting module 550.

[0119] The acquisition module 510 is configured to acquire diffusion magnetic resonance imaging data corresponding to a target brain region of a target object. Among them, the target brain region includes N gray matter regions, and the diffusion magnetic resonance imaging data includes signal feature data of at least one voxel corresponding to each of the N gray matter regions. In one embodiment, the acquisition module 510 can be used to perform the operation S110 described above, which will not be elaborated here.

[0120] The encoding module 520 is configured to process the signal feature data by using an encoder to obtain an encoded feature vector corresponding to each of the N gray matter regions. In one embodiment, the encoding module 520 can be used to perform the operation S120 described above, which will not be elaborated here.

[0121] The stratification module 530 is configured to, for the nth gray matter region among the N gray matter regions, cluster at least one voxel located on the nth gray matter region based on a preset number of clusters k and the encoded feature vector to obtain an nth clustering result, where the nth clustering result includes k clustering clusters, each clustering cluster represents a layer in the nth gray matter region, each clustering cluster includes at least one voxel, and n N. In one embodiment, the stratification module 530 can be used to perform the operation S130 described above, which will not be elaborated here.

[0122] The determination module 540 is configured to, for the mth clustering cluster among the k clustering clusters, determine the sorting information of the mth clustering cluster according to the position information of the voxels in the mth clustering cluster, the position information of the first interface, and the position information of the second interface, where the first interface represents the interface between the gray matter region where the voxel is located and the associated white matter region, and the second interface represents the interface between the gray matter region where the voxel is located and the associated cerebrospinal fluid region. In one embodiment, the determination module 540 can be used to perform the operation S140 described above, which will not be elaborated here.

[0123] The sorting module 550 is configured to sort the k clustering clusters according to the sorting information of each of the k clustering clusters to obtain the gray matter stratification result of the target object's brain region. In one embodiment, the sorting module 550 can be used to perform the operation S150 described above, which will not be elaborated here.

[0124] According to an embodiment of the present disclosure, the determination module 520 includes a first determination sub-module, a second determination sub-module, a third determination sub-module, and a fourth determination sub-module.

[0125] The first determination sub-module is configured to determine, for each voxel in the m-th clustering cluster, a first distance from the voxel to the first interface according to the position information of the voxel and the position information of the first interface.

[0126] The second determination sub-module is configured to determine a second distance between the first interface and the second interface according to the position information of the first interface and the position information of the second interface.

[0127] The third determination sub-module is configured to obtain a depth ratio of the voxel according to the ratio between the first distance and the second distance.

[0128] The fourth determination sub-module is configured to perform an averaging process on the depth ratios corresponding to multiple voxels respectively to obtain the m-th average depth ratio.

[0129] According to an embodiment of the present disclosure, the layering module 530 includes a first layering sub-module, a second layering sub-module, a third layering sub-module, and a fourth layering sub-module.

[0130] The first layering sub-module is configured to initialize k initial clustering centers according to a preset number of clusters k.

[0131] The second layering sub-module is configured to cluster at least one voxel based on the distance between the encoded feature vectors located on the n-th gray matter region and the initial clustering centers, and determine initial clusters corresponding to the k initial clustering centers respectively.

[0132] The third layering sub-module is configured to iteratively update the initial clustering centers of the initial clustering clusters according to multiple encoded feature vectors in the initial clustering clusters to obtain the clustering centers of the initial clustering clusters after iteration.

[0133] The fourth layering sub-module is configured to determine the n-th clustering result based on the clustering centers after iteration and a preset iteration stop condition.

[0134] According to an embodiment of the present disclosure, the fourth layering sub-module includes a first layering unit, a second layering unit, and a third layering unit.

[0135] The first layering unit is configured to determine a difference value between the clustering centers after iteration and the clustering centers before iteration.

[0136] The second layering unit is configured to determine the clustering centers before iteration as target clustering centers when the difference value meets the preset iteration stop condition.

[0137] The third layering unit is configured to determine the n-th clustering result according to the clusters corresponding to the target clustering centers respectively.

[0138] According to an embodiment of the present disclosure, the acquisition module 510 includes a first acquisition sub-module, a second acquisition sub-module, and a third acquisition sub-module.

[0139] The first acquisition sub-module is configured to perform spatial position registration on the initial diffusion magnetic resonance imaging data based on the tissue structure distribution information corresponding to the brain region of the target object, so as to obtain the registered diffusion magnetic resonance imaging data.

[0140] The second acquisition sub-module is configured to perform a second masking process on the tissue structure distribution information to generate a second brain mask image, where the second brain mask image is used to distinguish the target brain region from the non-target brain region.

[0141] The third acquisition sub-module is configured to obtain the diffusion magnetic resonance imaging data corresponding to the target brain region of the target object according to the second brain mask image and the registered diffusion magnetic resonance image.

[0142] According to an embodiment of the present disclosure, any multiple of the modules, sub-modules, units, and sub-units can be combined and implemented in one module, or any one of them can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the modules, sub-modules, units, and sub-units can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or can be implemented by any other reasonable way of integrating or packaging circuits, etc., in hardware or firmware, or implemented in any one of the three implementation manners of software, hardware, and firmware, or in an appropriate combination of any several of them. Alternatively, at least one of the modules, sub-modules, units, and sub-units can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding functions can be executed.

[0143] Figure 6 The block diagram of an electronic device suitable for implementing the method for gray matter stratification of brain regions based on diffusion magnetic resonance imaging according to an embodiment of the present disclosure is shown.

[0144] Figure 6 The shown electronic device is only an example and should not bring any limitation to the functions and usage scope of the embodiments of the present disclosure.

[0145] Such as Figure 6As shown, the computer electronic device 600 according to an embodiment of the present disclosure includes a processor 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage section 609 into a random access memory (RAM) 603. The processor 601 can include, for example, a general microprocessor (such as a CPU), an instruction set processor, and / or a related chipset, and / or a dedicated microprocessor (such as an application specific integrated circuit (ASIC)), and so on. The processor 601 can also include on-board memory for caching purposes. The processor 601 can include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0146] In the RAM 603, various programs and data required for the operation of the electronic device 600 are stored. The processor 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. The processor 601 performs various operations of the method flow according to an embodiment of the present disclosure by executing the program in the ROM 602 and / or the RAM 603. It should be noted that the program can also be stored in one or more memories other than the ROM 602 and the RAM 603. The processor 601 can also perform various operations of the method flow according to an embodiment of the present disclosure by executing the program stored in one or more memories.

[0147] According to an embodiment of the present disclosure, the electronic device 600 can further include an input / output (I / O) interface 605, and the input / output (I / O) interface 605 is also connected to the bus 604. The electronic device 600 can further include one or more of the following components connected to the input / output (I / O) interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including, for example, a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the input / output (I / O) interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed so that a computer program read from it can be installed into the storage section 608 as needed.

[0148] According to an embodiment of the present disclosure, the method flow according to the embodiments of the present disclosure can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program includes program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 609, and / or installed from the removable medium 611. When the computer program is executed by the processor 601, the above functions defined in the system of the embodiments of the present disclosure are executed. According to an embodiment of the present disclosure, the above-described systems, devices, apparatuses, modules, units, etc. can be implemented by computer program modules.

[0149] The present disclosure also provides a computer-readable storage medium, which may be included in the device / device / system described in the above embodiments; or may exist separately without being assembled into the device / device / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method for stratifying gray matter in brain regions according to the embodiments of the present disclosure is implemented.

[0150] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium. For example, it may include but is not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program, and the program can be used by or combined with an instruction execution system, device, or device.

[0151] For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the above-described ROM 602 and / or RAM 603 and / or one or more memories other than ROM 602 and RAM 603.

[0152] An embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program includes program code for executing the method provided by the embodiments of the present disclosure. When the computer program product runs on an electronic device, the program code is used to cause the electronic device to implement the method for stratifying gray matter in brain regions provided by the embodiments of the present disclosure.

[0153] When the computer program is executed by the processor 601, the above functions defined in the system / apparatus of the embodiments of the present disclosure are executed. According to an embodiment of the present disclosure, the above-described systems, apparatuses, modules, units, etc. can be implemented by computer program modules.

[0154] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 609, and / or installed from the removable medium 611. The program code included in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0155] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiments of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages include but are not limited to, such as Java, C++, python, the "C" language, or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., by connecting through the Internet using an Internet service provider).

[0156] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a portion of code, which contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks may occur in a different order than noted in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and combinations of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions. Those skilled in the art will appreciate that the features described in the various embodiments of the present disclosure can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present disclosure. In particular, without departing from the spirit and teachings of the present disclosure, the features described in the various embodiments of the present disclosure can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of the present disclosure.

[0157] The above describes embodiments according to the present disclosure. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments have been described separately above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Without departing from the scope of the present disclosure, those skilled in the art can make various substitutions and modifications, and all such substitutions and modifications should fall within the scope of the present disclosure.

Claims

1. A method for gray matter stratification of brain regions based on diffusion magnetic resonance imaging, characterized in that, The method includes: Obtaining diffusion magnetic resonance imaging data corresponding to a target brain region of a target object, where the target brain region includes N gray matter regions, and the diffusion magnetic resonance imaging data includes signal feature data of at least one voxel corresponding to each of the N gray matter regions; Processing the signal feature data by an encoder to obtain encoded feature vectors corresponding to each of the N gray matter regions; For the n-th gray matter region among the N gray matter regions, at least one voxel located on the n-th gray matter region is clustered based on a preset number of clusters k and the encoded feature vector to obtain an n-th clustering result, where the n-th clustering result includes k clusters, each cluster represents a layer in the n-th gray matter region, and each cluster includes at least one voxel, n N; For the m-th clustering cluster among the k clustering clusters, determining sorting information of the m-th clustering cluster according to position information of voxels in the m-th clustering cluster, position information of a first interface, and position information of a second interface, where the first interface represents an interface between the gray matter region where the voxel is located and an associated white matter region, and the second interface represents an interface between the gray matter region where the voxel is located and an associated cerebrospinal fluid region; Sorting the k clustering clusters according to the sorting information of each of the k clustering clusters to obtain a gray matter stratification result of the target object's brain region.

2. The method according to claim 1, wherein The sorting information includes an average depth ratio; Wherein, for the m-th clustering cluster among the k clustering clusters, determining the sorting information of the m-th clustering cluster according to position information of voxels in the m-th clustering cluster, position information of a first interface, and position information of a second interface includes: For each voxel in the m-th clustering cluster, determining a first distance from the voxel to the first interface according to the position information of the voxel and the position information of the first interface; Determining a second distance between the first interface and the second interface according to the position information of the first interface and the position information of the second interface; Obtaining a depth ratio of the voxel according to a ratio between the first distance and the second distance; Performing an averaging process on the depth ratios corresponding to the multiple voxels to obtain an m-th average depth ratio.

3. The method according to claim 1, wherein, The position information of the first interface and the position information of the second interface are determined based on the following operations: Performing a first masking process on the organizational structure distribution information corresponding to the brain region of the target object to generate a first brain mask image, where the first brain mask image is used to distinguish gray matter regions, white matter regions, and cerebrospinal fluid regions; Based on the first brain mask image, determining the position information of the first interface between the gray matter region and the white matter region and the position information of the second interface between the gray matter region and the cerebrospinal fluid region.

4. The method according to claim 1, wherein The clustering of at least one voxel located in the n-th gray matter region based on the preset clustering number k and the encoded feature vector to obtain an n-th clustering result includes: Initializing k initial clustering centers according to the preset clustering number k; Clustering at least one voxel based on the distance between the encoded feature vector located in the n-th gray matter region and the initial clustering centers to determine initial clusters corresponding to the k initial clustering centers respectively; Iteratively updating the initial clustering centers of the initial clustering clusters according to the multiple encoded feature vectors in the initial clustering clusters to obtain the clustering centers after iteration of the initial clustering clusters; Determining the n-th clustering result based on the clustering centers after iteration and a preset iteration stop condition.

5. The method according to claim 4, characterized in that Determining the nth clustering result based on the iterated cluster centers and a preset iteration stop condition includes: Determining a difference value between the iterated cluster centers and the cluster centers before iteration; When the difference value meets the preset iteration stop condition, determining the cluster centers before iteration as target cluster centers; Determining the nth clustering result according to the cluster clusters respectively corresponding to the target cluster centers.

6. The method according to claim 1, wherein Obtaining diffusion magnetic resonance imaging data corresponding to a target brain region of a target object includes: Performing spatial position registration on initial diffusion magnetic resonance imaging data based on tissue structure distribution information corresponding to the brain region of the target object to obtain registered diffusion magnetic resonance imaging data; Performing a second masking process on the tissue structure distribution information to generate a second brain mask image, where the second brain mask image is used to distinguish the target brain region from non-target brain regions; Obtaining diffusion magnetic resonance imaging data corresponding to the target brain region of the target object according to the second brain mask image and the registered diffusion magnetic resonance image.

7. The method according to claim 1, wherein The encoder is obtained based on the following training operations: Obtaining training samples, where the training samples include sample signal feature data; Processing the sample signal feature data by using an initial encoder to obtain sample encoded feature vectors; Processing the sample encoded feature vectors by using an initial decoder to obtain sample decoded feature vectors; Training the initial encoder according to the sample signal feature data and the sample decoded feature vectors to obtain the encoder.

8. The method according to claim 7, wherein The training the initial encoder according to the sample signal feature data and the sample decoded feature vectors to obtain the encoder includes: Calculating a loss value between the sample signal feature data and the sample decoded feature vectors by using a loss function to obtain a target loss value; Training the initial encoder according to the target loss value to obtain the encoder.

9. A gray matter stratification device for brain regions based on diffusion magnetic resonance imaging, characterized in that, The apparatus includes: An acquisition module, configured to acquire diffusion magnetic resonance imaging data corresponding to a target brain region of a target object, where the target brain region includes N gray matter regions, and the diffusion magnetic resonance imaging data includes signal feature data of at least one voxel corresponding to each of the N gray matter regions; An encoding module, configured to process the signal feature data by using an encoder to obtain encoded feature vectors corresponding to the N gray matter regions respectively; A hierarchical module is configured to cluster at least one voxel located in the n-th gray matter region out of N gray matter regions based on a preset number of clusters k and the encoded feature vector, to obtain an n-th clustering result, where the n-th clustering result includes k clustering clusters, each clustering cluster represents a layer in the n-th gray matter region, and each clustering cluster includes at least one voxel, n N; A determination module, configured to, for the mth cluster of the k clusters, determine sorting information of the mth cluster according to position information of voxels in the mth cluster, position information of a first interface, and position information of a second interface, where the first interface represents an interface between the gray matter region where the voxel is located and an associated white matter region, and the second interface represents an interface between the gray matter region where the voxel is located and an associated cerebrospinal fluid region; A sorting module, configured to sort the k clusters according to the sorting information of the k clusters respectively to obtain a gray matter stratification result of the brain region of the target object.

10. An electronic device, including: One or more processors; A memory, configured to store one or more computer programs, Characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 8.

Images