Heterogeneous brain network classification method and device, electronic equipment and storage medium

Abstract

The invention discloses a heterogeneous brain network classification method and device, electronic equipment and a storage medium, and belongs to the technical field of brain network analysis, and the method comprises the steps: obtaining magnetic resonance imaging data, and constructing a brain function network according to the magnetic resonance imaging data; extracting multi-scale topological features of the brain function network; inputting the multi-scale topological features into a completely trained target heterogeneous brain network classification model, and determining category labels of the magnetic resonance imaging data; the multi-scale topological features of the brain function network are extracted in a targeted manner, so that multi-angle classification basis acquisition and multi-scale topological feature cooperative driving are realized; the category label of the magnetic resonance imaging data is determined based on the target heterogeneous brain network classification model according to the multi-scale topological characteristics, so that the category label of the magnetic resonance imaging data is rapidly and accurately determined on the basis of effectively utilizing a multi-angle classification basis.

Description

Technical Field

[0001] The present invention relates to the technical field of brain network analysis, and in particular to a method, device, electronic device and storage medium for classifying heterogeneous brain networks. Background Art

[0002] In the field of brain network analysis, the classification of heterogeneous brain networks is an important and challenging task. Heterogeneous brain networks refer to networks composed of brain regions (nodes) of different individuals and functional connections (edges) between these regions. These nodes and edges represent the functional connections between different regions in the brain, and the brain networks of different individuals may have significant differences in structure and function, which makes it difficult for traditional classification methods to adapt to these heterogeneous changes, resulting in insufficient accuracy and robustness of classification results.

[0003] At present, existing methods often rely on single-scale features in the process of brain network analysis, which makes it difficult to fully capture the complexity of brain networks. This limitation makes it more difficult to effectively compare and classify brain networks between different individuals.

[0004] Therefore, in the process of classifying heterogeneous brain networks, the existing technology has the problem of low reliability of classification results due to a single classification basis. Summary of the invention

[0005] In view of this, it is necessary to provide a heterogeneous brain network classification method, device, electronic device and storage medium to solve the problem of low reliability of classification results due to a single classification basis in the process of heterogeneous brain network classification in the prior art.

[0006] In order to solve the above problems, the present invention provides a heterogeneous brain network classification method, comprising: Acquire magnetic resonance imaging data and construct a brain functional network based on the magnetic resonance imaging data; Extract multi-scale topological features of brain functional networks; The multi-scale topological features are input into a well-trained target heterogeneous brain network classification model to determine the category labels of magnetic resonance imaging data.

[0007] In one possible implementation, constructing a brain function network based on magnetic resonance imaging data includes: constructing a node set based on magnetic resonance imaging data; Connect any two nodes in the node set with an edge to construct an edge set; The functional connection strength between any two nodes is calculated by the Pearson correlation coefficient to obtain the edge weight set; A brain functional network is constructed based on node sets, edge sets, and weight sets.

[0008] In a possible implementation, the multi-scale topological features include a local feature matrix and a global feature vector; extracting the multi-scale topological features of the brain functional network includes: A non-fully connected brain network based on a minimum connectivity threshold to generate a brain functional network; Calculate multiple local eigenvectors of a non-fully connected brain network by using graph theory methods, and generate a local feature matrix based on the multiple local eigenvectors; Generating global feature vectors from non-fully connected brain networks.

[0009] In a possible implementation, the local feature vector includes degree centrality, clustering coefficient, betweenness centrality and node importance; the global feature vector includes the weight of the edge in the brain functional network.

[0010] In one possible implementation, the multi-scale topological features are input into a well-trained target heterogeneous brain network classification model to determine the category labels of the magnetic resonance imaging data, including: Obtaining deep representations and attention weight matrices of multi-scale topological features; Perform weighted summation on the deep representation and attention weight matrix to obtain the fused feature vector; The fused feature vector is normalized and the category label of the magnetic resonance imaging data is determined.

[0011] In a possible implementation, the deep representation includes a local deep representation and a global deep representation; obtaining the deep representation and the attention weight matrix of the multi-scale topological features includes: Generate a local deep representation of multiple local feature vectors based on the first multi-layer perceptron module; Generate a global deep representation of the global feature vector based on the second multi-layer perceptron module; The deep representation is transposed to generate the attention weight matrix.

[0012] In a possible implementation, acquiring magnetic resonance imaging data includes: sequentially performing time slice correction, motion correction, interference signal regression and bandpass filtering on the initial magnetic resonance imaging data to obtain preprocessed magnetic resonance imaging data; The pre-processed magnetic resonance imaging data are averaged based on the automatic anatomical labeling to obtain magnetic resonance imaging data.

[0013] In order to solve the above problems, the present invention also provides a heterogeneous brain network classification device, comprising: A brain function network construction module, used for acquiring magnetic resonance imaging data and constructing a brain function network according to the magnetic resonance imaging data; Multi-scale topological feature extraction module, used to extract multi-scale topological features of brain functional networks; The category label determination module is used to input multi-scale topological features into a well-trained target heterogeneous brain network classification model to determine the category labels of magnetic resonance imaging data.

[0014] In order to solve the above problem, the present invention further provides an electronic device, including a memory and a processor, wherein: Memory, used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the heterogeneous brain network classification method as described above.

[0015] In order to solve the above problems, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps in the heterogeneous brain network classification method as described above.

[0016] The beneficial effect of adopting the above embodiment is as follows: the present invention provides a heterogeneous brain network classification method. First, by preliminarily combing the magnetic resonance imaging data that can be directly obtained, a brain functional network that is convenient for quantitatively studying the relationship between brain regions is constructed; then, the multi-scale topological features of the brain functional network are extracted in a targeted manner, and multi-angle acquisition of classification basis and multi-scale topological feature collaborative driving are achieved; finally, a well-trained target heterogeneous brain network classification model is used to determine the category label of the magnetic resonance imaging data according to the multi-scale topological features, thereby achieving rapid and accurate determination of the category label of the magnetic resonance imaging data based on the effective use of multi-angle classification basis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of a flow chart of an embodiment of a heterogeneous brain network classification method provided by the present invention; Figure 2 A schematic diagram of a process for constructing a brain function network according to an embodiment of the present invention; Figure 3 A schematic diagram of a process for extracting multi-scale topological features of a brain functional network according to an embodiment of the present invention; Figure 4 A schematic diagram of a process of training an initial heterogeneous brain network classification model according to an embodiment of the present invention; Figure 5 A structural block diagram of an embodiment of a heterogeneous brain network classification device provided by the present invention; Figure 6 This is a structural block diagram of an embodiment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of the present invention and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0019] A heterogeneous brain network is a special network structure that contains multiple types of nodes and edges, which may represent different brain regions, neuron types, connection methods or functional characteristics.

[0020] Topological characteristics refer to the properties of a space or a figure that remain unchanged under topological transformations (such as stretching, bending, compression, etc. that do not change the shape and size of the figure). These properties are independent of the metric properties of the figure (such as length, area, volume, etc.).

[0021] Automatic Anatomical Labeling (AAL) is a population-based average brain map. The AAL map is a standardized brain region division method that divides the brain into multiple regions with specific anatomical and functional characteristics.

[0022] Classification of heterogeneous brain networks is an important part of neuroscience research. By classifying and analyzing different types of heterogeneous brain networks, we can discover the interaction and coordination mechanisms between different brain regions, providing new perspectives and methods for neuroscience research. At the same time, the classification results can also provide strong support for the diagnosis and treatment of neurological diseases and promote the development of neuroscience research. However, in the process of classifying heterogeneous brain networks, the existing technology has the problem of low reliability of classification results due to the single classification basis.

[0023] In order to solve the above problems, the present invention provides a heterogeneous brain network classification method, device, electronic device and storage medium, which are described in detail below.

[0024] Figure 1 A flow chart of an embodiment of a heterogeneous brain network classification method provided by the present invention is shown in FIG. Figure 1 As shown, heterogeneous brain network classification methods include: S101: acquiring magnetic resonance imaging data, and constructing a brain functional network according to the magnetic resonance imaging data; S102: Extract multi-scale topological features of brain functional networks; S103: Input the multi-scale topological features into a well-trained target heterogeneous brain network classification model to determine the category label of the magnetic resonance imaging data.

[0025] In this embodiment, first, the magnetic resonance imaging data that can be directly obtained are preliminarily sorted out to construct a brain functional network that is convenient for quantitative research on the relationship between brain regions; then, the multi-scale topological features of the brain functional network are extracted in a targeted manner, and multi-angle acquisition of classification basis and multi-scale topological feature collaborative driving are achieved; finally, a well-trained target heterogeneous brain network classification model is used to determine the category labels of the magnetic resonance imaging data based on the multi-scale topological features, thereby achieving rapid and accurate determination of the category labels of the magnetic resonance imaging data based on the effective use of multi-angle classification basis.

[0026] It should be noted that the category labels of magnetic resonance imaging data generally include two categories: normal and abnormal, corresponding to a unique digital label, namely 0 or 1.

[0027] In other embodiments, the number of category labels may be further refined according to actual needs, which is not limited here.

[0028] Magnetic resonance imaging (MRI) data is detailed image data about the internal structure of the human body obtained through magnetic resonance imaging technology.

[0029] As a preferred embodiment, in S101, in order to reduce the redundancy of magnetic resonance imaging data, for the initial magnetic resonance imaging data directly acquired, first, it is necessary to perform time slice correction, motion correction, interference signal regression and bandpass filtering on the initial magnetic resonance imaging data in sequence to obtain preprocessed magnetic resonance imaging data; then, the preprocessed magnetic resonance imaging data is averaged based on automatic anatomical labeling to obtain magnetic resonance imaging data.

[0030] In a specific embodiment, resting-state functional magnetic resonance imaging (rs-fMRI) is subjected to time slice correction to ensure that imaging data at different time points are aligned in time; motion correction is performed on rs-fMRI data, all images are corrected to an average image, and two iterations are performed to optimize the correction effect; skull stripping is performed on rs-fMRI data to generate a mask image containing only brain tissue; the global average intensity of rs-fMRI data is normalized to 10,000 to standardize signal intensity and reduce differences between different scans; rs-fMRI data is subjected to interference signal removal regression processing, including motion parameters, the first five principal components from white matter and cerebrospinal fluid signals, and linear and quadratic trends to reduce physiological and non-brain signal interference during scanning; a 0.01-0.1 Hz bandpass filter is applied to rs-fMRI data to reduce the effects of low-frequency drift and high-frequency noise; the functional image is linearly registered to the anatomical space, and then a transformation based on the white matter boundary is applied; the registration parameters previously used for anatomical data are applied to the functional data to transform the functional data to the standard space.

[0031] By performing targeted screening and standardized conversion of rs-fMRI data, data redundancy can be effectively reduced and data availability can be improved to facilitate subsequent classification.

[0032] As a preferred embodiment, in S101, in order to construct a brain function network according to magnetic resonance imaging data, such as Figure 2 As shown, Figure 2 A schematic diagram of a process for constructing a brain function network according to an embodiment of the present invention includes: S201: constructing a node set according to magnetic resonance imaging data; S202: Connect any two nodes in the node set with an edge to construct an edge set; S203: Calculate the functional connection strength between any two nodes through the Pearson correlation coefficient to obtain the edge weight set; S204: Construct a brain function network according to the node set, edge set and weight set.

[0033] In this embodiment, based on the construction of the node set and edge set of the magnetic resonance imaging data, the weight set of the edge is determined by calculating the Pearson correlation coefficient, thereby increasing the data content of the brain function network and specifically improving the reliability of the brain function network.

[0034] In a specific embodiment, for resting-state functional magnetic resonance imaging data, each functional brain area is taken as a node to construct a node set; any two different nodes are connected with edges to construct an edge set; then, the functional connection strength between different nodes is calculated using the Pearson correlation coefficient to construct an edge weight set; finally, a resting-state brain functional connection network is constructed based on the node set, edge set and weight set.

[0035] Among them, the calculation of Pearson correlation coefficient is:

[0036] in, is the i-th node and the jth node The Pearson correlation coefficient between them, Σ is the summation symbol, For Node time series, for The mean of For Node time series, for The mean of is the length of the time series data, is the number of nodes.

[0037] As a preferred embodiment, in S102, the multi-scale topological features include a local feature matrix and a global feature vector; in order to extract the multi-scale topological features of the brain functional network, such as Figure 3 As shown, Figure 3 A schematic diagram of a process for extracting multi-scale topological features of a brain functional network provided by the present invention includes: S301: generating a non-fully connected brain network of a brain functional network based on a minimum connection threshold; S302: Calculate multiple local eigenvectors of the non-fully connected brain network by using a graph theory method, and generate a local feature matrix based on the multiple local eigenvector features; S303: Generate a global feature vector based on the non-fully connected brain network.

[0038] In this embodiment, data screening of the brain functional network based on the minimum connection threshold can further reduce data redundancy and reduce the difficulty of data processing; by calculating the local feature vectors and generating the global feature vectors, multi-dimensional extraction of data information of non-fully connected brain networks is achieved, the diversity of the data is improved, and the reliability of subsequent data classification is enhanced.

[0039] In a specific embodiment, in S301, the minimum connection threshold is set according to actual needs, and the edges with connection strength higher than the minimum connection threshold are screened out to construct a new edge set and weight set. Then, based on the node set, the new edge set and the weight set, a non-fully connected brain network of different individuals is constructed, so that there is a path between any two nodes in the network to reach each other.

[0040] In a specific embodiment, in S302, the local feature vectors include degree centrality, clustering coefficient, betweenness centrality and node importance respectively; the global feature vector includes the weight of the edge in the brain function network.

[0041] Specifically, the number of directly connected edges between different nodes is calculated to generate the degree centrality feature vector; Calculate the tightness of the connection between neighboring nodes of different nodes and generate the clustering coefficient feature vector; Calculate the number of shortest paths passing through different nodes and generate the betweenness centrality eigenvector; Calculate the contribution of different nodes to the overall network function and generate node importance feature vector; Generate a local feature matrix based on degree centrality, clustering coefficient, betweenness centrality and node importance feature vectors; Generate a global feature vector based on the weight set of the non-fully connected brain network; Based on the local feature matrix and the global feature vector, a representation of multi-scale topological features is generated.

[0042] It should be noted that the calculation of node degree centrality is:

[0043] in, For Node The node degree centrality of is an element in the adjacency matrix. If the node and If there is an edge between ,otherwise .

[0044] The calculation of node clustering coefficient is:

[0045] in, For Node The node clustering coefficient of For Node The number of edges between neighbor nodes, Is a node The degree.

[0046] The calculation of node betweenness centrality is:

[0047] in, For Node The node betweenness centrality of For Node To Node The total number of shortest paths, are those that pass through the nodes The number of paths.

[0048] The calculation of node importance is:

[0049] in, For Node The importance of nodes, For Node The number of neighbor nodes, For Node The total number of all participating shortest paths, For Node The degree, For Node The eigenvector centrality of For Node The eigenvector centrality of .

[0050] Subsequently, a local feature matrix is ​​generated based on the degree centrality, clustering coefficient, betweenness centrality, and node importance feature vectors.

[0051] In other embodiments, the local feature vector may include at least two of degree centrality, clustering coefficient, betweenness centrality and node importance, and may be adjusted to other features according to actual needs, which are not limited here.

[0052] In a specific embodiment, in S103, in order to input the multi-scale topological features into the well-trained target heterogeneous brain network classification model and determine the category label of the magnetic resonance imaging data, it is necessary to perform targeted training on the initial heterogeneous brain network classification model, such as Figure 4 As shown, Figure 4 A flow chart of an embodiment of training an initial heterogeneous brain network classification model provided by the present invention includes: S401: Obtaining deep representations and attention weight matrices of multi-scale topological features; S402: Perform weighted summation on the deep representation and the attention weight matrix to obtain a fused feature vector; S403: Normalize the fused feature vector to determine the category label of the magnetic resonance imaging data.

[0053] Among them, in S401, the deep representation includes a local deep representation and a global deep representation; in order to obtain the deep representation and attention weight matrix of multi-scale topological features, first, a local deep representation of multiple local feature vectors is generated based on the first multi-layer perceptron module; then, a global deep representation of the global feature vector is generated based on the second multi-layer perceptron module; finally, the deep representation is transposed to generate an attention weight matrix.

[0054] In a specific embodiment, a first multi-layer perceptron module with two hidden layers is constructed to generate a deep representation of a local feature vector; a second multi-layer perceptron module with one hidden layer is constructed to generate a deep representation of a global feature vector; an attention network module is constructed to generate an attention weight matrix of the feature; the weighted sum of the deep representation of multi-scale features and the attention weight matrix is ​​calculated to generate a fused feature vector; the fused feature vector is used as the input of the Softmax layer to realize the prediction of the category label.

[0055] Among them, the construction of the attention network module includes: first, obtaining input data and calculating the transpose of the input data; second, inputting the transposed data into a fully connected layer and using RELU as the activation function; third, using the output of the fully connected layer as the input of the Softmax layer to generate the attention weight matrix; fourth, calculating and outputting the transpose of the attention weight matrix.

[0056] In this embodiment, a fused feature vector is obtained by weighted summing up the local deep representation, global deep representation and attention weight matrix of the non-fully connected brain network, thereby achieving an organic integration of the classification basis of the heterogeneous brain network, effectively improving the diversity of the classification basis, and thus improving the reliability of the classification results.

[0057] Through the above method, first, a brain functional network is constructed to facilitate the quantitative study of the relationship between brain regions; then, the multi-scale topological features of the brain functional network are extracted in a targeted manner, realizing multi-angle acquisition of classification basis and collaborative driving of multi-scale topological features; finally, the category labels of magnetic resonance imaging data are determined according to multi-scale topological features using a well-trained target heterogeneous brain network classification model, realizing the rapid and accurate determination of category labels of magnetic resonance imaging data based on the effective use of multi-angle classification basis.

[0058] In order to solve the above problems, the present invention also provides a heterogeneous brain network classification device, such as Figure 5 As shown, Figure 5 This is a structural block diagram of an embodiment of a heterogeneous brain network classification device provided by the present invention. The heterogeneous brain network classification device 500 includes: A brain function network construction module 501 is used to obtain magnetic resonance imaging data and construct a brain function network according to the magnetic resonance imaging data; A multi-scale topological feature extraction module 502, used to extract multi-scale topological features of brain functional networks; The category label determination module 503 is used to input the multi-scale topological features into the well-trained target heterogeneous brain network classification model to determine the category label of the magnetic resonance imaging data.

[0059] like Figure 6 As shown, the present invention also provides an electronic device 600. The electronic device 600 includes a processor 601, a memory 602 and a display 603. Figure 6 Only some components of the electronic device 600 are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0060] In some embodiments, the processor 601 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 602, such as the heterogeneous brain network classification method of the present invention.

[0061] In some embodiments, the processor 601 may be a single server or a server group. The server group may be centralized or distributed. In some embodiments, the processor 601 may be local or remote. In some embodiments, the processor 601 may be implemented in a cloud platform. In one embodiment, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-cloud, etc., or any combination thereof.

[0062] In some embodiments, the memory 602 may be an internal storage unit of the electronic device 600, such as a hard disk or memory of the electronic device 600. In other embodiments, the memory 602 may also be an external storage device of the electronic device 600, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc., equipped on the electronic device 600.

[0063] Furthermore, the memory 602 may include both an internal storage unit of the electronic device 600 and an external storage device. The memory 602 is used to store application software installed in the electronic device 600 and various data.

[0064] In some embodiments, the display 603 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 603 is used to display information of the electronic device 600 and to display a visual user interface. The components 601-603 of the electronic device 600 communicate with each other through a system bus.

[0065] In one embodiment, when the processor 601 executes the heterogeneous brain network classification program in the memory 602, the following steps may be implemented: Acquire magnetic resonance imaging data and construct a brain functional network based on the magnetic resonance imaging data; Extract multi-scale topological features of brain functional networks; The multi-scale topological features are input into a well-trained target heterogeneous brain network classification model to determine the category labels of magnetic resonance imaging data.

[0066] It should be understood that: when the processor 601 executes the heterogeneous brain network classification program in the memory 602, in addition to the above functions, other functions can also be implemented. For details, please refer to the description of the corresponding method embodiment above.

[0067] Furthermore, the embodiment of the present invention does not specifically limit the type of the electronic device 600 mentioned, and the electronic device 600 may be a portable electronic device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop computer, etc. Exemplary embodiments of portable electronic devices include but are not limited to portable electronic devices equipped with IOS, Android, Microsoft or other operating systems. The above-mentioned portable electronic device may also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 600 may not be a portable electronic device, but a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0068] Accordingly, an embodiment of the present invention also provides a computer-readable storage medium, which is used to store computer-readable programs or instructions. When the program or instructions are executed by a processor, the steps or functions in the heterogeneous brain network classification method provided by the above-mentioned method embodiments can be implemented.

[0069] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the computer program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0070] The heterogeneous brain network classification method, device, electronic device and storage medium provided by the present invention are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for classifying heterogeneous brain networks, characterized in that: include: Acquiring magnetic resonance imaging data, and constructing a brain functional network based on the magnetic resonance imaging data; Extracting multi-scale topological features of the brain functional network; The multi-scale topological features are input into a well-trained target heterogeneous brain network classification model to determine the category label of the magnetic resonance imaging data.

2. The heterogeneous brain network classification method according to claim 1, characterized in that: The step of constructing a brain function network according to the magnetic resonance imaging data comprises: constructing a node set according to the magnetic resonance imaging data; Connect any two nodes of the node set with an edge to construct an edge set; The functional connection strength between any two nodes is calculated by the Pearson correlation coefficient to obtain an edge weight set; A brain function network is constructed according to the node set, the edge set and the weight set.

3. The heterogeneous brain network classification method according to claim 2, characterized in that: The multi-scale topological features include a local feature matrix and a global feature vector; The step of extracting multi-scale topological features of the brain functional network comprises: generating a non-fully connected brain network of the brain functional network based on a minimum connectivity threshold; Calculating multiple local eigenvectors of the non-fully connected brain network by a graph theory method, and generating a local feature matrix based on the multiple local eigenvectors; A global feature vector is generated according to the non-fully connected brain network.

4. The heterogeneous brain network classification method according to claim 3, characterized in that: The local feature vector includes degree centrality, clustering coefficient, betweenness centrality and node importance; the global feature vector includes the weight of the edge in the brain function network.

5. The heterogeneous brain network classification method according to claim 3, characterized in that: The step of inputting the multi-scale topological features into a well-trained target heterogeneous brain network classification model to determine the category label of the magnetic resonance imaging data includes: Obtaining a deep representation and an attention weight matrix of the multi-scale topological features; Performing weighted summation on the deep representation and the attention weight matrix to obtain a fused feature vector; The fused feature vector is normalized to determine a category label of the magnetic resonance imaging data.

6. The heterogeneous brain network classification method according to claim 5, characterized in that: The deep-level representation includes a local deep-level representation and a global deep-level representation; the deep-level representation and the attention weight matrix of the multi-scale topological features are obtained, including: Generate the local deep representation of the local feature vector based on the first multi-layer perceptron module; Generate the global deep representation of the global feature vector based on a second multi-layer perceptron module; The deep level representation is transposed to generate the attention weight matrix.

7. The heterogeneous brain network classification method according to claim 1, characterized in that: The acquiring of magnetic resonance imaging data comprises: sequentially performing time slice correction, motion correction, interference signal regression and bandpass filtering on the initial magnetic resonance imaging data to obtain preprocessed magnetic resonance imaging data; The pre-processed magnetic resonance imaging data is averaged based on automatic anatomical labeling to obtain the magnetic resonance imaging data.

8. A heterogeneous brain network classification device, characterized in that: include: A brain function network construction module, used to acquire magnetic resonance imaging data and construct a brain function network according to the magnetic resonance imaging data; A multi-scale topological feature extraction module, used to extract the multi-scale topological features of the brain functional network; The category label determination module is used to input the multi-scale topological features into a well-trained target heterogeneous brain network classification model to determine the category label of the magnetic resonance imaging data.

9. An electronic device, characterized in that: comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the heterogeneous brain network classification method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the heterogeneous brain network classification method described in any one of claims 1 to 7.

Images