Brain function subregion division method and device, equipment and medium

Abstract

The invention discloses a brain function subregion division method and device, equipment and a medium, and relates to the technical field of biomedical imaging signal processing, and the method comprises the steps: obtaining an initial time feature and an initial space feature of target fMRI data after obtaining the initial time feature and the initial space feature of the target fMRI data; performing feature extraction on the initial time feature and the initial space feature by using a trained automatic encoder to obtain a depth time feature and a depth space feature, considering non-linear features of space, time and hiding, and emphasizing the properties of space-time interaction and complementation; and performing spectral clustering analysis on a final similarity matrix constructed by iterative fusion of the time similarity matrix and the space similarity matrix of the similarity network to obtain a brain function sub-region division result of the user, thereby improving comprehensiveness and accuracy of brain function sub-region division.

Description

Technical Field

[0001] The present application relates to the technical field of biomedical imaging signal processing, and in particular to a method, device, equipment and medium for dividing brain functional subregions. Background Art

[0002] Brain parcellation refers to a spatially continuous area that is significantly different from its neighboring areas in terms of anatomy or function, including the whole brain, the whole cerebral cortex, and the parcellation of specific brain areas. Brain parcellation can be divided into anatomical parcellation and functional parcellation. Anatomical parcellation is the division of the brain into regions according to its physical structure. This parcellation is based on the actual anatomical structure of the brain, such as the different lobes of the brain (frontal lobe, parietal lobe, occipital lobe, temporal lobe), neural nuclei, etc. Functional parcellation is the division of functionally consistent regions based on the activity patterns of the brain. Although anatomical parcellation provides a clear map of brain structure, it cannot fully demonstrate the functional activities and interactions of brain regions. In addition, anatomical parcellation may sometimes ignore the plasticity of the brain and the differences between individuals, that is, the same anatomical structure may perform different functions in different individuals or under different circumstances. Functional parcellation can reveal the activity patterns of the brain, thereby providing a deep understanding of brain function and information processing. Functional parcellation divides the brain into functionally homogeneous parcellations. This functional specialization makes it very suitable for exploring brain function and dysfunction. Reliable functional parcellation helps improve the accuracy of predicting brain diseases.

[0003] Functional magnetic resonance imaging (fMRI) plays a key role in exploring brain functional connectivity. Brain functional connectivity refers to the interaction and coordinated activities between different brain regions, which is the basis for understanding how the brain works in an integrated network manner. Resting-state fMRI (rs-fMRI) allows researchers to explore the brain's intrinsic functional networks when it is not performing a specific task. Brain parcellation studies based on rs-fMRI have developed rapidly in recent years. The challenge in this field is the lack of a gold standard for functional parcellation. Therefore, most fMRI-based parcellation studies rely on unsupervised methods.

[0004] In fMRI-based functional zoning, traditional unsupervised methods have certain limitations and cannot fully capture the complexity of brain activity. Therefore, there is an urgent need for a more comprehensive and accurate method, device, equipment and medium for brain functional subregion division. Summary of the invention

[0005] The purpose of this application is to provide a method, device, equipment and medium for dividing brain functional subregions, which can improve the comprehensiveness and accuracy of brain functional subregion division.

[0006] To achieve the above objectives, this application provides the following solutions.

[0007] In a first aspect, the present application provides a method for dividing brain functional subregions, comprising:

[0008] Get the user's raw fMRI data;

[0009] Preprocessing the original fMRI data to obtain target fMRI data;

[0010] Extracting the fMRI signal corresponding to each voxel in the target fMRI data in the area to be divided to obtain the initial time feature of the target fMRI data; extracting the fMRI signal corresponding to each brain region in the target fMRI data to obtain a brain region time signal matrix; determining the initial spatial feature of the target fMRI data according to the initial time feature and the brain region time signal matrix;

[0011] Using the trained autoencoder, extracting the initial time features and the initial spatial features of the target fMRI data respectively to obtain deep time features and deep spatial features;

[0012] Constructing a similarity network; the vertices of the similarity network correspond one-to-one to the voxels in the area to be divided; the weights of the edges of the similarity network are determined by a temporal similarity matrix and a spatial similarity matrix; the temporal similarity matrix is ​​determined by the Pearson correlation coefficient of the deep temporal features of the two voxels connected by the edge; the spatial similarity matrix is ​​determined by the Pearson correlation coefficient of the deep spatial features of the two voxels connected by the edge;

[0013] The final similarity matrix is ​​calculated by iteratively fusing the temporal similarity matrix and the spatial similarity matrix;

[0014] For a user, the final similarity matrix is ​​analyzed using the NCut spectral clustering algorithm to obtain the user's individual functional sub-region division result; the individual functional sub-region division result includes several functional sub-regions.

[0015] In a second aspect, the present application provides a device for dividing brain functional subregions, including the following modules.

[0016] The original fMRI data acquisition module is used to: acquire the original fMRI data of the user;

[0017] A preprocessing module, used to: preprocess the original fMRI data to obtain target fMRI data;

[0018] The initial feature extraction module is used to: extract the fMRI signal corresponding to each voxel in the target fMRI data in the area to be divided to obtain the initial time feature of the target fMRI data; extract the fMRI signal corresponding to each brain region in the target fMRI data to obtain the brain region time signal matrix; determine the initial spatial feature of the target fMRI data according to the initial time feature and the brain region time signal matrix;

[0019] A deep feature extraction module is used to: use the trained autoencoder to extract the initial time features and the initial space features of the target fMRI data respectively, to obtain deep time features and deep space features;

[0020] A similarity network construction module is used to: construct a similarity network; the vertices of the similarity network correspond one-to-one to the voxels in the area to be divided; the weights of the edges of the similarity network are determined by a temporal similarity matrix and a spatial similarity matrix; the temporal similarity matrix is ​​determined by the Pearson correlation coefficient of the deep temporal features of the two voxels connected by the edge; the spatial similarity matrix is ​​determined by the Pearson correlation coefficient of the deep spatial features of the two voxels connected by the edge;

[0021] The final similarity matrix calculation module is used to: calculate the final similarity matrix according to the iterative fusion of the time similarity matrix and the space similarity matrix;

[0022] The functional sub-region division module is used to: for a user, use the NCut spectral clustering algorithm to analyze the final similarity matrix to obtain the user's individual functional sub-region division result; the individual functional sub-region division result includes several functional sub-regions.

[0023] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned method for dividing brain functional subregions.

[0024] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for dividing brain functional subregions.

[0025] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0026] The present application provides a method, apparatus, device and medium for brain functional subregion division, which is implemented based on an end-to-end joint optimization framework. After obtaining the initial time features and initial spatial features of the target fMRI data, the trained autoencoder is used to extract the initial time features and initial spatial features to obtain deep time features and deep spatial features, taking into account the spatial, temporal and hidden nonlinear features (deep time features and deep spatial features), and emphasizing the nature of space-time interaction and complementarity; a spectral clustering analysis is performed on the final similarity matrix iteratively constructed from the temporal similarity matrix and the spatial similarity matrix of the similarity network to obtain the user's individual functional subregion division results. The present application integrates the spatial and temporal features of brain activity into deep embedded clustering. The deep embedded clustering consists of stacked autoencoders and spectral clustering, which can adaptively adjust the contribution of spatial and temporal features, solving the problem that the traditional method relies on the singleness of a single feature and improving the comprehensiveness and accuracy of brain functional subregion division. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 This is a diagram of an application environment of a method for dividing brain functional subregions in an embodiment of the present application;

[0029] Figure 2 A flowchart of a method for dividing brain functional subregions provided in one embodiment of the present application;

[0030] Figure 3 A schematic diagram of the comparison results of the repeatability of different methods under all possible numbers of partitions in an ablation experiment provided in an embodiment of the present application; Figure 3 (a), (b), and (c) are the repeatability comparison results of Dice coefficient, intraclass correlation coefficient, and normalized mutual information of different methods under the repeatability comparison strategy between two scans of the same user under all possible numbers of partitions; Figure 3 (d), (e), and (f) are the repeatability comparison results of Dice coefficient, intra-group correlation coefficient, and normalized mutual information of different methods under all possible numbers of partitions under the repeatability comparison strategy between the two group-level results; Figure 3 (g), (h), and (i) are the repeatability comparison results of Dice coefficient, intraclass correlation coefficient, and normalized mutual information of different methods under the repeatability comparison strategy between paired individuals in all possible numbers of partitions; Figure 3(j), (k), and (l) are the repeatability comparison results of Dice coefficient, intraclass correlation coefficient, and normalized mutual information of different methods under the repeatability comparison strategy of individuals with the leave-one-out method in all possible numbers of partitions;

[0031] Figure 4 A schematic diagram of the comparison results of functional homogeneity of different methods in an ablation experiment provided in an embodiment of the present application;

[0032] Figure 5 A schematic diagram showing the difference between a healthy user and a Parkinson's patient user in terms of graph theory indicators provided in an embodiment of the present application; Figure 5 (a) shows the difference in modularity between healthy users and Parkinson's patients. Figure 5 (b) shows the difference in small-world index between healthy users and Parkinson's patients;

[0033] Figure 6 A schematic diagram of the relationship between the graph theory index changes and the unified Parkinson's disease rating scale provided in one embodiment of the present application; Figure 6 (a) shows the relationship between the changes in modularity indicators and the Unified Parkinson's Disease Rating Scale; Figure 6 (b) shows the relationship between the changes in the small-world index and the Unified Parkinson's Disease Rating Scale;

[0034] Figure 7 A schematic diagram of functional modules of a device for dividing brain functional subregions provided in one embodiment of the present application;

[0035] Figure 8 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0037] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0038] Unsupervised methods are diverse and rich, including various clustering techniques such as hierarchical clustering, K-means clustering and spectral clustering, as well as region growing, boundary mapping and decomposition techniques such as independent component analysis. In recent years, deep neural networks (DNNs) have also begun to make progress in the field of brain parcellation. Although DNNs have become the preferred technology for automatic brain segmentation and brain disease prediction based on structural MRI and supervised methods using anatomical atlases / masks or disease state labels, their potential in the field of fMRI-based parcellation remains to be further explored. Existing studies include supervised brain parcellation using deep residual networks to predict the probability of voxels belonging to functional networks predetermined by ICA, demonstrating the practicality of deep learning algorithms. However, research based on unsupervised deep learning methods is relatively rare.

[0039] Input features can generally be divided into two types: time-based features and space-based features. The temporal feature-based segmentation relies on the blood oxygen level-dependent (BOLD) amplitude (e.g., the time series of voxels or its transformed variant, the activation value of voxels) to characterize the local functional activity of the partitioned region (i.e., the target region). On the other hand, the spatial feature-based segmentation captures the long-range connections between the target region and other regions (i.e., the reference region) to outline the global interaction pattern of the target region.

[0040] In fMRI-based functional parcellation, traditional unsupervised methods usually focus only on temporal features or spatial features. This includes clustering analysis using the correlation or similarity of time series, or defining brain regions based on the distribution of brain spatial patterns. However, these single-dimensional methods have certain limitations, such as the possibility of failing to fully capture the complexity of brain activity. Therefore, it becomes particularly important to consider comprehensive methods that simultaneously integrate spatial and temporal features. This comprehensive analysis strategy is expected to provide more comprehensive and accurate mapping of brain functional regions, providing richer insights for a deeper understanding of neurodegenerative diseases and other brain dysfunctions.

[0041] Therefore, this application integrates local temporal characteristics, global spatial characteristics, and hidden nonlinear characteristics from three perspectives, and considers the interaction between time and space to perform brain functional subregional division. This framework opens up new possibilities for the functional subregional division of brain regions through deep learning and adaptive spatiotemporal feature fusion technology, and provides a new tool for brain science research.

[0042] The brain functional subregion division method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the user's original fMRI data to the server 104. After the server 104 receives the user's original fMRI data, the server 104 pre-processes the original fMRI data for the user's original fMRI data to obtain the target fMRI data, and the target fMRI data is feature extracted to obtain the initial time feature and the initial space feature, and the trained autoencoder is used to extract the initial time feature and the initial space feature respectively to obtain the deep time feature and the deep space feature, and construct a similarity network. According to the time similarity matrix and the space similarity matrix in the similarity network, the final similarity matrix is ​​iteratively calculated, and the final similarity matrix is ​​analyzed using the NCut spectral clustering algorithm to obtain the user's individual functional sub-area division result. The server 104 can feedback the obtained individual functional sub-area division result for the user to the terminal 102. In addition, in some embodiments, the brain functional sub-region division method can also be implemented independently by the server 104 or the terminal 102. For example, the terminal 102 can directly perform functional sub-region division on the user's original fMRI data, or the server 104 can obtain the user's original fMRI data from the data storage system and perform functional sub-region division on the user's original fMRI data.

[0043] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, and tablet computers. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0044] In an exemplary embodiment, Figure 2 As shown, a method for dividing brain functional subregions is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate, including the following steps 201 to 207. Among them:

[0045] Step 201: Obtain the user's original fMRI data.

[0046] Step 202: pre-processing the original fMRI data to obtain target fMRI data.

[0047] Step 203: extract the fMRI signal corresponding to each voxel in the area to be divided in the target fMRI data to obtain the initial time characteristics of the target fMRI data; extract the fMRI signal corresponding to each brain region in the target fMRI data to obtain a brain region time signal matrix; determine the initial spatial characteristics of the target fMRI data based on the initial time characteristics and the brain region time signal matrix.

[0048] Step 204: using the trained autoencoder, extract the initial time features and initial space features of the target fMRI data to obtain deep time features and deep space features.

[0049] Step 205: construct a similarity network; the vertices of the similarity network correspond one-to-one to the voxels in the area to be divided; the weights of the edges of the similarity network are determined by the temporal similarity matrix and the spatial similarity matrix; the temporal similarity matrix is ​​determined by the Pearson correlation coefficient of the deep temporal features of the two voxels connected by the edge; the spatial similarity matrix is ​​determined by the Pearson correlation coefficient of the deep spatial features of the two voxels connected by the edge.

[0050] Step 206: Calculate the final similarity matrix by iteratively fusing the temporal similarity matrix and the spatial similarity matrix.

[0051] Step 207: for a user, using the NCut spectral clustering algorithm, analyze the final similarity matrix to obtain the user's individual functional sub-region division result; the individual functional sub-region division result includes a plurality of functional sub-regions.

[0052] The implementation of the above steps 201 to 207 takes into account the spatial, temporal and hidden nonlinear features (deep temporal features and deep spatial features), and emphasizes the nature of spatial and temporal interaction and complementarity. The present application integrates the spatial and temporal features of brain activity into deep embedding clustering. Deep embedding clustering consists of stacked autoencoders and spectral clustering, which can adaptively adjust the contribution of spatial and temporal features, solve the problem that traditional methods rely on the singleness of a single feature, and improve the comprehensiveness and accuracy of brain functional subregion division. In addition, the present application can also determine the group-level functional subregion division results based on the individual functional subregion division results of different users, and has the flexibility to partition at the individual level and the group level.

[0053] In step 201 of the present application, a group of original fMRI data of 176 healthy users are obtained. The original fMRI data of each user needs to be preprocessed. The preprocessing process in step 202 is as follows:

[0054] (1) Spatial artifact removal: remove artifacts caused by magnetic field inhomogeneity from the original fMRI image through distortion correction and other techniques. (2) Temporal layer correction: solve the problem of time differences between voxels of the same fMRI data layer during scanning. (3) When the user remains still, the head will still move slightly due to psychological factors such as heartbeat or breathing. Head movement correction eliminates the impact of head movement on the acquisition of magnetic resonance data. (4) Registration: align the user's fMRI data to its structural state imaging for subsequent analysis. (5) Since each user's brain has certain differences, each user's brain space needs to be aligned to a standard reference space during group-level analysis. Spatial standardization reduces the differences between users to ensure data consistency and comparability. (6) Since other signals from various sources will interfere with the data, linear drift is performed to reduce noise interference. (7) Finally, spatial smoothing is performed to reduce noise and random variation in the fMRI signal.

[0055] In another exemplary embodiment of the present application, in step 203, the fMRI signal corresponding to each brain region in the target fMRI data is extracted to obtain a brain region time signal matrix, which may specifically include the following steps 301 to 303, wherein:

[0056] Step 301: extracting the fMRI signal corresponding to each brain region in the cortical region of the target fMRI data using a surface-based multimodal brain region atlas, and obtaining a time point matrix of the fMRI signal corresponding to each brain region in the cortical region;

[0057] Step 302: extracting the fMRI signal corresponding to each brain region in the subcortical region of the target fMRI data using the Desikan-Killiany atlas template, and obtaining a time point matrix of the fMRI signal corresponding to each brain region in the subcortical region;

[0058] Step 303: Determine a brain region time signal matrix according to the fMRI signal time point matrix corresponding to each brain region in the cortical region and the fMRI signal time point matrix corresponding to each brain region in the subcortical region.

[0059] In step 203, the initial spatial features of the target fMRI data are determined according to the initial time features and the brain region time signal matrix, specifically including: calculating the Pearson correlation coefficient according to the initial time features and the brain region time signal matrix; and using the Pearson correlation coefficient as the initial spatial features of the target fMRI data.

[0060] The initial time feature refers to the number of voxels in the area to be divided for each user multiplied by the time signal matrix at the time point. The time feature mainly considers the local information of the area to be divided, that is, the area in the brain. The initial spatial feature refers to the Pearson correlation coefficient of the number of voxels in the area to be divided for each user multiplied by the time signal matrix of the area to be divided at the time point and the brain area multiplied by the signal matrix at the time point. The connection feature between the area to be divided and other areas of the brain is the spatial feature. The initial spatial feature takes into account the global information.

[0061] Take the brainstem as an example to divide the functional sub-regions. The existing brainstem data are all voxel-level data. The fMRI data in the HCP dataset is used. For each user, a signal matrix of 3472 (number of brainstem voxels) multiplied by 900 (time points) will be obtained. The signal matrix of 3472 multiplied by 900 is used as the initial time feature. The amount of voxel-level data in other brain regions is too large, and data dimension reduction is required. Since the HCP dataset has been preprocessed with surface generation, the HCP dataset uses the surface-based multimodal brain region atlas HCP-MMP when extracting the region of interest (ROI) signal in the cortical region. The atlas contains 360 multimodal cortical regions, and the fMRI signal time point matrix corresponding to each brain region in the cortical region is extracted. The subcortical part uses the Desikan-Killiany atlas template, which divides the subcortical region into 18 brain regions, and the fMRI signal time point matrix corresponding to each brain region in the subcortical region is extracted. Using these two templates, each user can obtain a signal matrix of 378 (ROI) by 900 (time points), and use this signal matrix of 378 (ROI) by 900 (time points) as the brain region time signal matrix for subsequent generation of functional sub-region data.

[0062] After obtaining the initial time features and brain region time signal matrix, the Pearson correlation coefficient between the two is calculated as the initial spatial features of the target fMRI data.

[0063] An end-to-end joint optimization framework (deep learning framework) integrates a novel adaptive similarity network fusion module to fuse the spatial and temporal features of brain activity into deep embedding clustering, consisting of stacked autoencoders and spectral clustering.

[0064] In another exemplary embodiment of the present application, before step 204, the autoencoder needs to be trained to obtain a trained autoencoder.

[0065] The autoencoder consists of an encoder and a decoder. The encoder represents the initial temporal features or initial spatial features of the input area to be divided as a vector x through nonlinear transformation and maps it to the hidden representation h:

[0066] h=f(W1x+b1) (1);

[0067] Where f is a nonlinear activation function, W1 represents the encoding weight matrix and b1 is the encoding bias. The decoder reconstructs the input data from the hidden representation h, aiming to minimize the reconstruction error:

[0068] x′=f(W2h+b2) (2);

[0069] Where x′ is the reconstructed input, i.e., the output of the decoder; W2 and b2 are the decoding weights and biases, respectively. The training process involves minimizing the loss function mean squared error (MSE). If the input vector x consists of N features, the loss function of the autoencoder is:

[0070]

[0071] Among them, L(x, x′) represents the loss function, N represents all elements of the matrix in the initial time feature or initial spatial feature of the area to be divided, and x i represents the i-th element of the matrix in the initial time feature or initial space feature, x i ′ represents the reconstructed input corresponding to the i-th element of the matrix in the initial time feature or the initial space feature.

[0072] The similarity network is represented as a graph G = (V, E). The vertex V corresponds to the voxel {x1, x2, ...x n}, n represents the number of voxels in the region to be divided, and the edge E is weighted according to the similarity between voxels. The weight is determined by the n×n similarity matrix W, and W(i, j) represents the voxel x i and x j The similarity matrix W is obtained by the Pearson correlation coefficient based on the deep time feature and the deep space feature, respectively, and the similarity is set to zero if the voxels are not adjacent. In this embodiment, the similarity matrix obtained based on the Pearson correlation coefficient of the deep time feature is named as the time similarity matrix, and the similarity matrix obtained based on the Pearson correlation coefficient of the deep space feature is named as the space similarity matrix.

[0073] In another exemplary embodiment of the present application, the above step 206 is replaced by the following steps 401 to 403:

[0074] Step 401: normalizing the temporal similarity matrix and the spatial similarity matrix respectively to obtain a temporal normalized matrix and a spatial normalized matrix;

[0075] Step 402: Calculate a temporal adjacent voxel similarity matrix and a spatial adjacent voxel similarity matrix according to the temporal similarity matrix and the spatial similarity matrix respectively;

[0076] Step 403: iteratively fuse the temporal adjacent voxel similarity matrix, the spatial adjacent voxel similarity matrix, the temporal normalization matrix and the spatial normalization matrix to obtain a final similarity matrix.

[0077] The above temporal similarity matrix and spatial similarity matrix need to be further normalized. One method to achieve normalization is as follows:

[0078]

[0079] Where P(i, j) represents the voxel x i and x j The normalized matrix element value between , W(i, k) represents the voxel x i and x k The similarity between them, k is the voxel x i The adjacent voxel x in the adjacent voxel set of k Serial number.

[0080] When processing the graph G, this method uses the K nearest neighbor algorithm to quantify the similarity matrix and obtains the temporal adjacent voxel similarity matrix and the spatial adjacent voxel similarity matrix as shown below:

[0081]

[0082] Where S(i, j) represents the voxel x i and x j The element value in the similarity matrix between adjacent voxels, N i is the voxel x i The neighboring voxels of each voxel are determined by the K nearest neighbor algorithm.

[0083] The core assumption of the K-nearest neighbor algorithm is that for a given data point, it has a higher similarity with nearby data points, while the similarity with distant data points is less reliable. The key iterative step in the fusion method is to iteratively update the fusion matrix to reflect the similarity between data points through the following formula:

[0084]

[0085] in, Represents the fusion matrix of the first type of data features after t iterations, that is, the time feature fusion matrix; Represents the fusion matrix of the second type of data features after t iterations, that is, the spatial feature fusion matrix; represents the spatial feature fusion matrix obtained after t-1 iterations, represents the temporal feature fusion matrix obtained after t-1 iterations; S (1) represents the temporal neighbor voxel similarity matrix, S (2) represents the spatial adjacent voxel similarity matrix; (S (1) ) T and (S (2) ) T denote the transpose of the temporal neighbor voxel similarity matrix and the spatial neighbor voxel similarity matrix, respectively.

[0086] In the iterative process, the two matrices (temporal feature fusion matrix and spatial feature fusion matrix) will influence and fuse each other in the fusion process. After n steps, the final similarity matrix P (g) The calculation is as follows:

[0087]

[0088] In another exemplary embodiment of the present application, in order to properly consider the interactive relationship between temporal features and spatial features and accurately reflect the role of these two features in data analysis and processing, this paper needs to quantify the respective contributions of temporal and spatial features, not only to determine which feature is more important, but also to have a deeper understanding of how temporal and spatial features interact with each other and show their unique impact in different application scenarios. To this end, an adaptive learning mechanism is introduced, that is, an adjustable parameter α is used to balance the temporal features. and spatial characteristics The importance and contribution between them.

[0089]

[0090] Where P(c) represents the final similarity matrix. Formula (9) describes an adaptive learning framework, where α is a parameter between 0 and 1, representing the relative importance of temporal features and spatial features. When α is close to 1, the model is more biased towards temporal features; and when α is close to 0, it is more biased towards spatial features. This dynamic adjustment mechanism enables the model to flexibly adjust the importance of spatiotemporal features according to different data sets and application requirements.

[0091] In another exemplary embodiment of the present application, the above step 207 includes the following steps 501 to 505:

[0092] Step 501: performing absolute value conversion on the final similarity matrix to construct an adjacency matrix and a degree matrix;

[0093] Step 502: Calculate the Laplacian matrix;

[0094] Step 503: for each voxel, construct a normalized Laplacian matrix corresponding to each voxel according to the degree matrix and the Laplacian matrix;

[0095] Step 504: for each voxel, calculate the eigenvalue of the normalized Laplacian matrix corresponding to the voxel, and arrange all eigenvalues ​​in ascending order, and take the eigenvectors corresponding to the first k eigenvalues ​​to construct a eigenmatrix; k is less than the number of voxels; the eigenvector matrix is ​​an n×k dimensional matrix; n represents the number of voxels;

[0096] Step 505: performing NCut spectral clustering on the feature vector matrix to obtain the user's individual functional sub-region division result, specifically comprising: performing NCut spectral clustering on the feature vector matrix using a K-means clustering algorithm to obtain the user's individual functional sub-region division result.

[0097] Based on the above similarity matrix generation method, the present application provides a given input sample set D = {x1, x2, ...x n} constructs a final similarity matrix P, where x n Represents the nth voxel in the area to be divided.

[0098] The final similarity matrix P is transformed into an absolute value, the similarity values ​​of non-adjacent voxels are set to zero, and the similarity values ​​of adjacent voxels are retained, thereby obtaining the adjacency matrix W and degree matrix D.

[0099] When calculating the Laplace matrix L, a normalized Laplace matrix is ​​constructed. The calculation formula is as follows:

[0100]

[0101] Then, the normalized Laplacian matrix L is calculated n The eigenvalues ​​of are sorted in ascending order. The eigenvectors f corresponding to the first k eigenvalues ​​are selected to form an eigenvector matrix. These eigenvectors f are then normalized to generate an n×k dimensional feature matrix H. In the feature matrix H, each row represents a k-dimensional sample, and there are n samples in total. Finally, the k-dimensional samples are clustered using a clustering method to obtain a set of cluster partitions C = {c1, c2, ...c n}, effectively assigning samples in the dataset D to k different clusters to achieve data clustering, where c n Indicates the functional sub-region corresponding to the nth voxel in the region to be divided. Preferably, the present implementation may use the K-means clustering algorithm to cluster the k-dimensional samples. The cut graph value of the NCut spectral clustering method is regarded as a key indicator in graph theory, which can effectively evaluate the quality and performance of graph segmentation.

[0102] For the above-mentioned adaptive similarity network fusion module, this application proposes a new loss function, which comprehensively considers the reconstruction loss value and NCut cut-graph value of each spatiotemporal feature, and is used to guide the training process of the deep learning framework. The loss function formula is as follows:

[0103]

[0104] Among them, L represents the loss function of the adaptive similarity network fusion module, x1 and are the initial time features of the input and reconstructed region to be divided, x2 and are the input and reconstructed initial spatial features of the region to be divided; F is the subordination matrix of the voxels in the region to be divided, each row of the subordination matrix represents the functional sub-region division result corresponding to the corresponding voxel, the subordination matrix has q columns, q is the number of functional sub-regions in the individual functional sub-region division result, when the voxel belongs to the functional sub-region, the corresponding element value in the matrix is ​​1, when it does not belong to the functional sub-region, the corresponding element value in the matrix is ​​0, so the subordination matrix is ​​an n×q-dimensional matrix; D and L are the degree matrix and normalized Laplace matrix of the similarity matrix after deep spatiotemporal fusion, respectively, and tr() represents the trace of the matrix. This loss function provides a new perspective for the field of brain functional partitioning, which can comprehensively and meticulously analyze and consider various characteristics of the brain.

[0105] In another exemplary embodiment of the present application, after obtaining the individual functional subregion division result of the user, the brain functional subregion division method further includes the following steps 601 to 604:

[0106] Step 601: Obtain individual functional sub-region division results of different users;

[0107] Step 602: for each of the individual functional subregion division results, construct an adjacency matrix corresponding to the individual functional subregion division result according to the region division result corresponding to each voxel in the individual functional subregion division result;

[0108] Step 603: constructing a group-level graph structure according to the adjacency matrix corresponding to all the individual functional sub-region division results; the vertices of the group-level graph structure correspond one-to-one to the voxels in the area to be divided; the weight of each edge of the group-level graph structure is determined by the same allocation frequency of the two vertices connected by the edge; the same allocation frequency indicates the number of voxels corresponding to the two vertices that are divided into the same functional sub-region in all the individual functional sub-region division results;

[0109] Step 604: Perform cluster analysis on the group-level graph structure to obtain group-level functional sub-region division results.

[0110] In order to obtain the division results at the individual level and the group level, this embodiment adopts a two-level analysis method, aiming to deeply explore the function and structure of the brainstem. In the first level analysis, the focus is on the brainstem structure of each individual, and functionally consistent areas are identified by constructing an N×N adjacency matrix (where N represents the number of voxels in the area to be divided). This process covers the analysis of the area to be divided data of each user to extract the similarities between vertices. In the second level analysis, turning to the group level, a stability graph (group-level graph structure) is constructed by integrating the block division results of all individuals. In this group-level graph structure, the weight of the edge is based on the two vertices v i and v j The frequency of the same block allocation in all users is used to determine the group-level graph structure. Finally, the clustering algorithm is applied to subdivide the group-level graph structure, thereby obtaining the group-level block division, and thus obtaining the group-level functional sub-area division result (group-level brainstem functional sub-area template).

[0111] In order to verify the performance of the brain functional subregion division method proposed in this application, two indicators, repeatability and functional homogeneity, were used. Repeatability indicators include Dice coefficient, intraclass correlation coefficient (ICC) and normalized mutual information (NMI). The comparison process mainly involves four different comparison strategies, namely the repeatability between two scans of the same user, the repeatability between the results of two group levels, the repeatability between paired individuals, and the repeatability between individuals using the leave-one-out method. Ablation experiments were carried out using different methods, including deep learning spatiotemporal feature fusion embedding (brain functional subregion division method proposed in this application), deep learning spatiotemporal feature equal importance embedding, deep temporal features (deep learning framework only uses temporal features), deep spatial features (deep learning framework only uses spatial features), spatiotemporal feature fusion embedding (only temporal features and spatial features are used without deep learning framework), temporal features (only temporal features are used without deep learning framework) and spatial features (only spatial features are used without deep learning framework), such as Figure 3 As shown, the repeatability provided by this application in the ablation experiment is the highest. The method using the deep learning framework surpasses the method without deep learning technology. The method that integrates spatiotemporal features can produce better results than the method that only uses spatial features or temporal features alone. This finding emphasizes the importance of deep features, temporal features, and spatial features in the functional division of brain regions, indicating that the combined use of these three is crucial to improving the accuracy of analysis. Figure 4 As shown, the functional homogeneity provided by the present application is the highest in the ablation experiment.

[0112] By fusing temporal and spatial features, information from different dimensions can be better extracted and integrated, thereby improving functional homogeneity.

[0113] In order to study the practicality of the functional area division method, the brain networks of Parkinson disease (PD) users and healthy users were deeply analyzed based on the proposed functional map. Nine graph theory indicators in four categories were selected: modularity, clustering coefficient, local efficiency, transitivity, global efficiency, betweenness centrality, eigenvector centrality, synchronicity and small-world index. By calculating and comparing these indicators, the brain networks of PD patients and healthy users were statistically analyzed using a two-sample t-test. Figure 5 As shown in Figure 2, the modularity and small-world indicators of healthy users were significantly higher than those of PD patients. Figure 6 As shown in the figure, modularity and small-world indexes showed a significant negative correlation with the Unified Parkinson's Disease Rating Scale (UPDRS). The negative correlation between modularity and UPDRS (p = 0.0060, r = -0.6497) means that as PD symptoms worsen, the modularity of the brain network may gradually decrease. This may be because the progression of PD causes the brain network structure to become more complex, and the boundaries between modules that were originally functionally independent become blurred, thus affecting the effectiveness and functional specificity of the brain network. The negative correlation between small-worldness and UPDRS (p = 0.0485, r = -0.4612) indicates that as PD symptoms worsen, the brain network may have problems in the balance between local clustering and global information transmission, resulting in decreased network efficiency.

[0114] The performance of this method was further studied, and the classification prediction of PD was made based on the functional template proposed by the functional area division method. Regardless of the machine learning method used, the functional division-based method proposed in this application consistently outperforms the segmentation based on the anatomical template in terms of accuracy. In K nearest neighbors, the functional template is 0.84 and the anatomical template is 0.79; in discriminant analysis, the functional template is 0.87 and the anatomical template is 0.74; in logistic regression, the accuracy of the functional template is 0.78 and the anatomical template is 0.65. In naive Bayes, the functional template is 0.82 and the anatomical template is 0.72; in random forest, the functional template is 0.83 and the anatomical template is 0.76. In the context of Parkinson's disease, the functional activity of certain brain regions may have changed in the early stages of the disease, even if there are no obvious abnormalities in their structure. By capturing these early functional changes, functional maps can help diagnose and understand the development of the disease earlier.

[0115] This application aims to divide the brain's partition structure according to functional characteristics based on fMRI signals, so as to explore the basic organizational structure and function of the brain. The brain functional sub-region division method proposed in this application is a deep spatiotemporal embedding partitioning method, which is implemented based on an end-to-end joint optimization framework. The end-to-end joint optimization framework simultaneously considers spatial, temporal and hidden nonlinear features, emphasizing the nature of spatiotemporal interaction and complementarity. Based on the end-to-end joint optimization framework, a novel adaptive similarity network fusion module is integrated to fuse the spatial and temporal features of brain activity into deep embedding clustering, which consists of stacked autoencoders and spectral clustering, and can adaptively adjust the contribution of spatial and temporal features, solving the problem that traditional methods rely on the singleness of a single feature. This application also has the flexibility to partition at the individual level and the group level. This application provides a new method for the division of brain functional areas, provides an important foundation for brain function research, and can support clinical application research on brain diseases.

[0116] The present application also provides an application scenario, which applies the above-mentioned brain functional sub-region division method. Specifically: the brain functional sub-region division method provided in this embodiment can be applied in the brain stem functional zoning scenario. The brain stem functional zoning scenario includes a data acquisition link, a functional sub-region division link and a group level division link; the user's original fMRI data enters the functional sub-region division link from the data acquisition link, obtains the corresponding individual functional sub-region division result, and enters the downstream group level division link, and determines the group level functional sub-region division result according to the individual functional sub-region division results of different users. The brain functional sub-region division method provided in this embodiment belongs to the functional sub-region division link. Specifically, in the functional sub-region division link process for the user's original fMRI data, the original fMRI data can be preprocessed to obtain target fMRI data, and feature extraction is performed on the target fMRI data to obtain initial time features and initial space features. The trained autoencoder is used to extract the initial time features and initial space features respectively to obtain deep time features and deep space features, and a similarity network is constructed. The final similarity matrix is ​​iteratively calculated based on the time similarity matrix and the space similarity matrix in the similarity network, and the final similarity matrix is ​​analyzed using the NCut spectral clustering algorithm to obtain the user's individual functional sub-region division result.

[0117] Based on the same inventive concept, the embodiment of the present application also provides a brain function sub-region division device for implementing the above-mentioned brain function sub-region division method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the brain function sub-region division device provided below can refer to the limitations of the brain function sub-region division method above, and will not be repeated here.

[0118] In an exemplary embodiment, Figure 7 As shown, a device for dividing brain functional subregions is provided, including the following modules.

[0119] The original fMRI data acquisition module T1 is used to: acquire the original fMRI data of the user;

[0120] The preprocessing module T2 is used to: preprocess the original fMRI data to obtain target fMRI data;

[0121] The initial feature extraction module T3 is used to: extract the fMRI signal corresponding to each voxel in the target fMRI data in the area to be divided to obtain the initial time feature of the target fMRI data; extract the fMRI signal corresponding to each brain region in the target fMRI data to obtain the brain region time signal matrix; determine the initial spatial feature of the target fMRI data according to the initial time feature and the brain region time signal matrix;

[0122] The deep feature extraction module T4 is used to: use the trained autoencoder to extract the initial time features and the initial space features of the target fMRI data respectively, to obtain the deep time features and the deep space features;

[0123] The similarity network construction module T5 is used to: construct a similarity network; the vertices of the similarity network correspond one-to-one to the voxels in the area to be divided; the weights of the edges of the similarity network are determined by the temporal similarity matrix and the spatial similarity matrix; the temporal similarity matrix is ​​determined by the Pearson correlation coefficient of the deep temporal features of the two voxels connected by the edge; the spatial similarity matrix is ​​determined by the Pearson correlation coefficient of the deep spatial features of the two voxels connected by the edge;

[0124] The final similarity matrix calculation module T6 is used to: calculate the final similarity matrix according to the iterative fusion of the time similarity matrix and the space similarity matrix;

[0125] The functional sub-region division module T7 is used to: for a user, use the NCut spectral clustering algorithm to analyze the final similarity matrix to obtain the user's individual functional sub-region division result; the individual functional sub-region division result includes several functional sub-regions.

[0126] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 8As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store functional sub-area division data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for dividing brain functional sub-areas is implemented.

[0127] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0128] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0129] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0130] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0131] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0132] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0133] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0134] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for dividing brain functional subregions, characterized in that: The method for dividing brain functional subregions comprises: Get the user's raw fMRI data; Preprocessing the original fMRI data to obtain target fMRI data; Extracting the fMRI signal corresponding to each voxel in the target fMRI data in the area to be divided to obtain the initial time feature of the target fMRI data; extracting the fMRI signal corresponding to each brain region in the target fMRI data to obtain a brain region time signal matrix; determining the initial spatial feature of the target fMRI data according to the initial time feature and the brain region time signal matrix; Using the trained autoencoder, extracting the initial time features and the initial spatial features of the target fMRI data respectively to obtain deep time features and deep spatial features; Constructing a similarity network; the vertices of the similarity network correspond one-to-one to the voxels in the area to be divided; the weights of the edges of the similarity network are determined by a temporal similarity matrix and a spatial similarity matrix; the temporal similarity matrix is ​​determined by the Pearson correlation coefficient of the deep temporal features of the two voxels connected by the edge; the spatial similarity matrix is ​​determined by the Pearson correlation coefficient of the deep spatial features of the two voxels connected by the edge; The final similarity matrix is ​​calculated by iteratively fusing the temporal similarity matrix and the spatial similarity matrix; For a user, the final similarity matrix is ​​analyzed using the NCut spectral clustering algorithm to obtain the user's individual functional sub-region division result; the individual functional sub-region division result includes several functional sub-regions.

2. The method for dividing brain functional subregions according to claim 1, characterized in that: After obtaining the individual functional subregion division result of the user, the brain functional subregion division method further includes: Obtain individual functional sub-region division results for different users; For each of the individual functional subregion division results, constructing an adjacency matrix corresponding to the individual functional subregion division result according to the regional division result corresponding to each voxel in the individual functional subregion division result; According to the adjacency matrix corresponding to all the individual functional sub-region division results, a group-level graph structure is constructed; the vertices of the group-level graph structure correspond one-to-one to the voxels in the area to be divided; the weight of each edge of the group-level graph structure is determined by the same allocation frequency of the two vertices connected by the edge; the same allocation frequency indicates the number of voxels corresponding to the two vertices that are divided into the same functional sub-region in all the individual functional sub-region division results; Cluster analysis is performed on the group-level graph structure to obtain group-level functional sub-region division results.

3. The method for dividing brain functional subregions according to claim 1, characterized in that: Extracting the fMRI signal corresponding to each brain region in the target fMRI data to obtain a brain region time signal matrix, specifically including: Extracting the fMRI signal corresponding to each brain region in the cortical region of the target fMRI data using a surface-based multimodal brain region atlas, and obtaining a time point matrix of the fMRI signal corresponding to each brain region in the cortical region; Using the Desikan-Killiany atlas template, extracting the fMRI signal corresponding to each brain region in the subcortical region of the target fMRI data, and obtaining the fMRI signal time point matrix corresponding to each brain region in the subcortical region; The brain region time signal matrix is ​​determined according to the fMRI signal time point matrix corresponding to each brain region in the cortical region and the fMRI signal time point matrix corresponding to each brain region in the subcortical region.

4. The method for dividing brain functional subregions according to claim 1, characterized in that: Determining the initial spatial features of the target fMRI data according to the initial time features and the brain region time signal matrix specifically includes: The Pearson correlation coefficient is calculated according to the initial time feature and the brain region time signal matrix; and the Pearson correlation coefficient is used as the initial spatial feature of the target fMRI data.

5. The method for dividing brain functional subregions according to claim 1, characterized in that: The final similarity matrix is ​​calculated by iterative fusion of the temporal similarity matrix and the spatial similarity matrix, including: Normalizing the temporal similarity matrix and the spatial similarity matrix respectively to obtain a temporal normalized matrix and a spatial normalized matrix; Calculating a temporal adjacent voxel similarity matrix and a spatial adjacent voxel similarity matrix according to the temporal similarity matrix and the spatial similarity matrix respectively; The final similarity matrix is ​​obtained by iteratively fusing the temporal adjacent voxel similarity matrix, the spatial adjacent voxel similarity matrix, the temporal normalization matrix and the spatial normalization matrix.

6. The method for dividing brain functional subregions according to claim 1, characterized in that: The final similarity matrix is ​​analyzed using the NCut spectral clustering algorithm to obtain the user's individual functional sub-region division results, including: Performing absolute value transformation on the final similarity matrix to construct an adjacency matrix and a degree matrix; Calculate the Laplacian matrix; For each voxel, construct a normalized Laplacian matrix corresponding to each voxel according to the degree matrix and the Laplacian matrix; For each voxel, the eigenvalue of the normalized Laplacian matrix corresponding to the voxel is calculated, and all eigenvalues ​​are arranged in ascending order, and the eigenvectors corresponding to the first k eigenvalues ​​are taken to construct a eigenmatrix; k is less than the number of voxels; the eigenvector matrix is ​​an n×k dimensional matrix; n represents the number of voxels; Perform NCut spectral clustering on the feature vector matrix to obtain the user's individual functional sub-region division results.

7. The method for dividing brain functional subregions according to claim 6, characterized in that: The user's individual functional sub-area division results are obtained, including: The K-means clustering algorithm is used to obtain the user's individual functional sub-region division results.

8. A brain function subregion division device based on the brain function subregion division method according to any one of claims 1 to 7, characterized in that: The brain functional sub-region division device comprises: The original fMRI data acquisition module is used to: acquire the original fMRI data of the user; A preprocessing module, used to: preprocess the original fMRI data to obtain target fMRI data; The initial feature extraction module is used to: extract the fMRI signal corresponding to each voxel in the target fMRI data in the area to be divided to obtain the initial time feature of the target fMRI data; extract the fMRI signal corresponding to each brain region in the target fMRI data to obtain the brain region time signal matrix; determine the initial spatial feature of the target fMRI data according to the initial time feature and the brain region time signal matrix; A deep feature extraction module is used to: use the trained autoencoder to extract the initial time features and the initial space features of the target fMRI data respectively, to obtain deep time features and deep space features; A similarity network construction module is used to: construct a similarity network; the vertices of the similarity network correspond one-to-one to the voxels in the area to be divided; the weights of the edges of the similarity network are determined by a temporal similarity matrix and a spatial similarity matrix; the temporal similarity matrix is ​​determined by the Pearson correlation coefficient of the deep temporal features of the two voxels connected by the edge; the spatial similarity matrix is ​​determined by the Pearson correlation coefficient of the deep spatial features of the two voxels connected by the edge; The final similarity matrix calculation module is used to: calculate the final similarity matrix according to the iterative fusion of the time similarity matrix and the space similarity matrix; The functional sub-region division module is used to: for a user, use the NCut spectral clustering algorithm to analyze the final similarity matrix to obtain the user's individual functional sub-region division result; the individual functional sub-region division result includes several functional sub-regions.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for dividing brain functional subregions according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for dividing brain functional subregions according to any one of claims 1 to 7 is implemented.

Images