Rs-fMRI classification method based on residual block and Bayesian variational reasoning

Abstract

The invention relates to an rs-fMRI classification method based on a residual block and Bayesian variational reasoning. The method comprises the following steps: firstly, acquiring rs-fMRI data and preprocessing the rs-fMRI data to obtain a time signal sequence; then, brain function connection is calculated based on the time signal sequence, and an original brain network is constructed; constructing a line graph based on the original brain network and performing line graph convolution; and finally, based on the line graph after convolution, Bayesian variation posteriori is adopted for classification, and an rs-fMRI classification result is obtained. The brain connection reconstruction model is used as a node, the importance of functional connection is highlighted, and the residual block is combined with Bayesian variational reasoning, so that accurate storage and analysis of brain network connection are facilitated, the problems of original noise and abnormal values are solved, and the accuracy and robustness of brain information extraction by the model for identification and classification are improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an rs-fMRI classification method based on residual blocks and Bayesian variational reasoning. Background Art

[0002] Resting-state functional magnetic resonance imaging (rs-fMRI) reflects the functional connectivity and network characteristics of the brain by measuring spontaneous neural activity in the resting state (i.e., without specific tasks or stimulation). When using fs-fMRI data for the identification and classification of neuropsychiatric diseases, the accurate extraction of disease-related functional connectivity information and the processing of complex noise pose great challenges to correct identification and classification.

[0003] With the continuous development of machine learning and deep learning technologies, more and more researchers have begun to apply these methods to the analysis of rs-fMRI data. By training classification models or prediction models, automatic identification and classification of neuropsychiatric diseases can be achieved. For example, the end-to-end edge predictor based on the variational graph autoencoder identifies and classifies rs-fMRI data. However, since the variational graph autoencoder model has a deeper network structure, multiple nonlinear transformations, and generates implicit representations at the same time, the internal mechanism and decision-making process of the model become more difficult to explain. At the same time, the variational graph autoencoder model mainly relies on learning potential features from graph data rather than explicitly integrating domain knowledge, which may lead to the potential representation learned by the model in the absence of domain knowledge not being fully aligned with the actual clinical or biological mechanism.

[0004] In general, the application of machine learning and deep learning techniques in rs-fMRI data analysis still has the following problems:

[0005] (1) The model extracts and learns irrelevant brain information, resulting in inaccurate prediction results;

[0006] (2) Learning about noise leads to overfitting of the model, thereby reducing the generalization and robustness of the model.

[0007] Therefore, in the relevant technology, there is an urgent need for a method that can improve the accuracy and robustness of the model in extracting brain information for identification and classification. Summary of the invention

[0008] Based on this, it is necessary to provide an rs-fMRI classification method based on residual blocks and Bayesian variational inference, which can improve the accuracy and robustness of the model in extracting brain information for identification and classification, in order to address the above technical problems.

[0009] In a first aspect, the present application provides an rs-fMRI classification method based on residual blocks and Bayesian variational reasoning. The method comprises:

[0010] Acquire rs-fMRI data and preprocess them to obtain a time signal series;

[0011] Calculate brain functional connectivity and construct an original brain network based on the time signal sequence;

[0012] Constructing a line graph based on the original brain network and performing line graph convolution;

[0013] Based on the convolved line graph, Bayesian variational posterior is used for classification to obtain rs-fMRI classification results.

[0014] Optionally, in one embodiment of the present application, the calculating brain functional connectivity based on the time signal sequence and constructing an original brain network comprises:

[0015] Calculating the Pearson correlation coefficient of the time signal sequence as a node feature matrix of the brain map structure;

[0016] A brain functional connection matrix is ​​determined based on the time signal sequence as an adjacency matrix of the brain map structure.

[0017] Optionally, in one embodiment of the present application, the calculating brain functional connectivity based on the time signal sequence and constructing an original brain network further comprises:

[0018] The original brain network is sparsely constructed.

[0019] Optionally, in one embodiment of the present application, constructing a line graph based on the original brain network and performing line graph convolution includes:

[0020] Determine a binary adjacency matrix based on the original brain network, retain the original edge features and the sum of adjacent node features, and obtain a line graph;

[0021] Graph convolution is performed based on the line graph to update the backbone residual block.

[0022] Optionally, in one embodiment of the present application, the classifying based on the convolved line graph using Bayesian variational posterior includes:

[0023] Optimize the network parameters of the Bayesian model using stochastic variational inference;

[0024] The overall loss function is defined based on maximizing the evidence lower bound of the marginal likelihood and the cross entropy loss.

[0025] In a second aspect, the present application also provides an rs-fMRI classification device based on residual blocks and Bayesian variational reasoning. The device comprises:

[0026] A data acquisition module is used to acquire and preprocess rs-fMRI data to obtain a time signal series;

[0027] An original brain network construction module, used to calculate brain functional connections and construct an original brain network based on the time signal sequence;

[0028] A line graph convolution module, used to construct a line graph based on the original brain network and perform line graph convolution;

[0029] The Bayesian variational reasoning and classification module is used to classify the convolved line graph using the Bayesian variational posterior to obtain the rs-fMRI classification result.

[0030] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the steps of the methods described in the above embodiments.

[0031] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in each of the above embodiments are implemented.

[0032] The above rs-fMRI classification method based on residual block and Bayesian variational reasoning first obtains rs-fMRI data and performs preprocessing to obtain a time signal sequence; then, based on the time signal sequence, brain functional connections are calculated and the original brain network is constructed; then, a line graph is constructed based on the original brain network and line graph convolution is performed; finally, based on the convolved line graph, Bayesian variational posterior is used for classification to obtain the rs-fMRI classification result. By remodeling brain connections as nodes, highlighting the importance of functional connections, and combining residual blocks with Bayesian variational reasoning, it is helpful to accurately preserve and analyze brain network connections, overcome the problems of original noise and outliers, and improve the accuracy and robustness of the model in extracting brain information for identification and classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is an application environment diagram of an rs-fMRI classification method based on residual blocks and Bayesian variational inference in one embodiment;

[0034] Figure 2 1 is a flowchart of an rs-fMRI classification method based on residual blocks and Bayesian variational inference in one embodiment;

[0035] Figure 3 1 is a flowchart of specific steps of an rs-fMRI classification method based on residual blocks and Bayesian variational inference in one embodiment;

[0036] Figure 4 is a structural block diagram of an rs-fMRI classification device based on residual blocks and Bayesian variational inference in one embodiment;

[0037] Figure 5 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0039] The rs-fMRI classification method based on residual block and Bayesian variational reasoning provided in the embodiment of the present application can be applied to Figure 1 In 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 integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0040] In one embodiment, Figure 2 As shown in Figure 2, a rs-fMRI classification method based on residual blocks and Bayesian variational inference is provided. Figure 1 The server in the example is used to illustrate the following steps:

[0041] S201: Acquire rs-fMRI data and perform preprocessing to obtain a time signal sequence.

[0042] In the present application embodiment, rs-fMRI data were obtained, mainly from ABIDE (Autism brain imaging Data Exchange) and ADHD-200 brain imaging data sets, including lesion samples of autism spectrum disorder (Autistic Spectrum Disorder, ASD) and attention deficit hyperactivity disorder (Attention-deficit hyperactivity disorder, ADHD). Afterwards, a standard preprocessing pipeline was applied to all image data, including motion correction, slice timing correction, spatial normalization, smoothing, artifact removal, spatial registration and temporal filtering, to remove the original noise, and the original voxel BOLD signal was processed into a time signal sequence of 200 ROIs (regions of interest) in combination with the CC 200 brain template: Where N is the number of ROIs and T is the total number of time points.

[0043] S203: Calculate brain functional connections based on the time signal sequence and construct an original brain network.

[0044] In the embodiment of the present application, after the time signal sequence is obtained, the brain functional connection is calculated based on the time signal sequence and the original brain network is constructed.

[0045] Specifically, in one embodiment of the present application, the calculating brain functional connectivity based on the time signal sequence and constructing the original brain network includes:

[0046] Calculating the Pearson correlation coefficient of the time signal sequence as a node feature matrix of the brain map structure;

[0047] A brain functional connection matrix is ​​determined based on the time signal sequence as an adjacency matrix of the brain map structure.

[0048] In one embodiment of the present application, first, the Pearson correlation coefficient of the BOLD signal in each ROI is calculated and used as the node feature matrix X of the brain map structure. The specific calculation formula is as follows:

[0049]

[0050] The matrix X consists of elements x ij Composition, r it Represents the element at the tth time point of the i-th ROI (row) in the time signal sequence matrix R, i∈{1,2,...,N}, j∈{1,2,...,N}.

[0051] After that, the absolute value of partial correlation between BOLD signals was calculated to obtain the brain functional connection matrix of each subject, which was used as the adjacency matrix A of the brain map structure, thereby constructing the original brain map structure G = (A, X), where V is the number of nodes in the graph structure.

[0052] In one embodiment of the present application, the calculating brain functional connections based on the time signal sequence and constructing the original brain network further comprises:

[0053] The original brain network is sparsely constructed.

[0054] In one embodiment of the present application, the original fully connected brain map is sparsely mapped using KNN (K=1), and the specific formula is as follows:

[0055] d(v i ,v j )=∥x i -x j ∥ 2

[0056]

[0057] Among them, v i refers to the i-th node, x i Refers to v i Corresponding features.

[0058] S205: constructing a line graph based on the original brain network and performing line graph convolution.

[0059] In the embodiment of the present application, in order to highlight the role of edges, after feature selection, all edges of the original brain network are used as nodes in the newly constructed line graph.

[0060] Specifically, in one embodiment of the present application, constructing a line graph based on the original brain network and performing line graph convolution includes:

[0061] Determine a binary adjacency matrix based on the original brain network, retain the original edge features and the sum of adjacent node features, and obtain a line graph;

[0062] Graph convolution is performed based on the line graph to update the backbone residual block.

[0063] In one embodiment of the present application, each edge in the original mind map structure corresponds to a unique node in the line graph. The edges in the line graph do not retain the initial weights, but represent the shared vertices between the nodes in the original graph. This will produce a binary adjacency matrix containing only 0 and 1, where 0 indicates no connection and 1 indicates a connection. At the same time, when constructing the node features of the line graph, the edge features in the original graph and the sum of the features of the adjacent nodes connected by these edges are retained, and the information in the original graph is retained while complying with the properties of the undirected graph, as shown in the following formula:

[0064]

[0065]

[0066] in, represents the feature of the i-th node in the line graph, x i and x j represents the endpoint of the edge in the original graph corresponding to the line graph node i, e ij Represents the weight of the original edge. Thus, a new line graph is constructed. in, This approach does not need to consider the order of node indices, thus producing a stable feature representation.

[0067] Considering that line graphs lack edge features or edge weights, and that both node and graph features must be retained during graph convolution, Residual GraphSAGE emerges as a suitable backbone. Residual GraphSAGE is a graph neural network model that combines residual connections and GraphSAGE algorithms. It introduces residual connections in each layer of GraphSAGE, that is, after aggregating neighbor node features, these features are added to the features of the current node to form a new node representation. For the update of node features at each layer, the message passing function is as follows:

[0068]

[0069] in, represents the learnable weight of the i-th node in the l-th layer, σ represents the activation function, and FC represents the fully connected layer that controls the feature dimension.

[0070] The convolution process is performed by the following matrix calculation method:

[0071]

[0072] in, is the adjacency matrix with self-connection, I is the identity matrix, D′ is the degree matrix of the line graph, and W (l) is the weight matrix of the lth layer.

[0073] The residual representation will then be updated as follows:

[0074]

[0075] in, is the residual representation at the lth layer, is the node feature matrix of the l+1 layer.

[0076] S207: Classify using Bayesian variational posterior based on the convolved line graph to obtain rs-fMRI classification results.

[0077] In the embodiment of the present application, the node feature matrix obtained from the last layer of message passing in the convolution Flatten it and use it as a brain map representation for each participant as follows:

[0078]

[0079] In order to obtain the final classification result, that is, the predicted label y = {y 1 ,y 2 ,...y M}, in C of M participants y In the class, the neural network implements an input mapping:

[0080]

[0081] Among them, D represents the output feature dimension of the line graph, and θ∈Θ represents the parameters of the neural network.

[0082] The unnormalized joint data label log-likelihood for classification in the Bayesian model is:

[0083]

[0084]

[0085]

[0086] in, represents the unnormalized predicted value, represents a zero-mean Gaussian noise term with variance Σ, which specifies a Gaussian prior for W.

[0087] In one embodiment of the present application, the classification based on the convolved line graph using Bayesian variational posterior includes:

[0088] Optimize the network parameters of the Bayesian model using stochastic variational inference;

[0089] The overall loss function is defined based on maximizing the evidence lower bound of the marginal likelihood and the cross entropy loss.

[0090] In one embodiment of the present application, in order to achieve accurate marginalization and avoid the computationally intensive complete marginal likelihood evaluation process, a random variational inference method is used, that is, the posterior distribution of the last layer is simultaneously approximated, and the network parameters are optimized by maximizing the evidence lower bound (ELBO) of the marginal likelihood. The specific process is as follows:

[0091]

[0092] Among them, R -1is the weight regularization term, ξ={W} is the uncertain parameter of the last layer, q(ξ|η) is the approximate posterior, and KL represents the Kullback-Leibler divergence.

[0093] The specific variational posterior form is determined as:

[0094]

[0095] Among them, S k represents the covariance, C y is the number of categories, each q(w k |η) is the weight of a single category w k The entire formula is the product of the variational distributions of all category weights.

[0096] While this decomposition ensures that each class preserves a dense covariance structure, it cannot explain the covariance between different classes. In contrast, adopting a fully dense variational posterior allows the induction of a training objective that incorporates these cross-class relationships.

[0097] In order to update the model parameters to minimize the prediction error, the loss function is defined as follows.

[0098]

[0099] where Σ is the covariance matrix, The formula comes from a double application of Jensen's inequality, and this lower bound is equal to the relaxed standard ELBO.

[0100] In addition, cross entropy is used as a component of the loss function:

[0101]

[0102] Among them, y is the predicted label, is the true label. It can motivate accurate binary classification and optimize the model.

[0103] Therefore, the overall loss is as follows:

[0104]

[0105] The original line graph model is trained using the overall loss, making the Bayesian variational posterior parameter w * and renew.

[0106] The goal of the last layer of Bayesian variation is to enhance the robustness of the model by adding uncertainty factors after training. The training formula of the final layer of Bayesian variation is:

[0107]

[0108] Parameterize the noise inverse covariance matrix Σ via Cholesky decomposition -1 And the covariance matrix of the variational posterior S, find the optimal distribution parameter η * and the covariance matrix Σ * .

[0109] In the above rs-fMRI classification method based on residual block and Bayesian variational reasoning, first, rs-fMRI data is obtained and preprocessed to obtain a time signal sequence; then, brain functional connections are calculated based on the time signal sequence and the original brain network is constructed; then, a line graph is constructed based on the original brain network and line graph convolution is performed; finally, Bayesian variational posterior is used for classification based on the convolved line graph to obtain the rs-fMRI classification result. By remodeling brain connections as nodes, highlighting the importance of functional connections, and combining residual blocks with Bayesian variational reasoning, it is helpful to accurately preserve and analyze brain network connections, overcome the problems of original noise and outliers, and improve the accuracy and robustness of the model in extracting brain information for identification and classification.

[0110] like Figure 3 , which is a flowchart of specific implementation steps of an rs-fMRI classification method based on residual blocks and Bayesian variational inference in a specific embodiment.

[0111] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0112] Based on the same inventive concept, the embodiment of the present application also provides a rs-fMRI classification device based on residual blocks and Bayesian variational reasoning for implementing the above-mentioned rs-fMRI classification method based on residual blocks and Bayesian variational reasoning. 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 rs-fMRI classification device based on residual blocks and Bayesian variational reasoning provided below can refer to the limitations of the rs-fMRI classification method based on residual blocks and Bayesian variational reasoning in the above text, and will not be repeated here.

[0113] In one embodiment, Figure 4 As shown, a rs-fMRI classification device 400 based on residual block and Bayesian variational reasoning is provided, comprising: a data acquisition module 401, an original brain network construction module 403, a line graph convolution module 405 and a Bayesian variational reasoning and classification module 407, wherein:

[0114] The data acquisition module 401 is used to acquire rs-fMRI data and perform preprocessing to obtain a time signal sequence.

[0115] The original brain network construction module 403 is used to calculate brain functional connections and construct an original brain network based on the time signal sequence.

[0116] The line graph convolution module 405 is used to construct a line graph based on the original brain network and perform line graph convolution.

[0117] The Bayesian variational reasoning and classification module 407 is used to perform classification based on the convolved line graph using the Bayesian variational posterior to obtain the rs-fMRI classification result.

[0118] In one embodiment of the present application, the original brain network construction module is also used to:

[0119] Calculating the Pearson correlation coefficient of the time signal sequence as a node feature matrix of the brain map structure;

[0120] A brain functional connection matrix is ​​determined based on the time signal sequence as an adjacency matrix of the brain map structure.

[0121] In one embodiment of the present application, the original brain network construction module is also used to:

[0122] The original brain network is sparsely constructed.

[0123] In one embodiment of the present application, the line graph convolution module is further used to:

[0124] Determine a binary adjacency matrix based on the original brain network, retain the original edge features and the sum of adjacent node features, and obtain a line graph;

[0125] Graph convolution is performed based on the line graph to update the backbone residual block.

[0126] In one embodiment of the present application, the Bayesian variational reasoning and classification module is also used to:

[0127] Optimize the network parameters of the Bayesian model using stochastic variational inference;

[0128] The overall loss function is defined based on maximizing the evidence lower bound of the marginal likelihood and the cross entropy loss.

[0129] Each module in the above-mentioned rs-fMRI classification device based on residual block and Bayesian variational inference can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0130] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, 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 and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, an rs-fMRI classification method based on residual blocks and Bayesian variational inference is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0131] Those skilled in the art will understand that Figure 5 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.

[0132] In one embodiment, a computer device is 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.

[0133] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0134] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0135] 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.

[0136] Those skilled 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), magnetoresistive 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. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. 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 are not limited to this.

[0137] The technical features of the above embodiments may be combined arbitrarily. 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.

[0138] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A rs-fMRI classification method based on residual blocks and Bayesian variational inference, characterized in that: The method comprises: Acquire rs-fMRI data and preprocess them to obtain a time signal series; Calculate brain functional connectivity and construct an original brain network based on the time signal sequence; Constructing a line graph based on the original brain network and performing line graph convolution; Based on the convolved line graph, Bayesian variational posterior is used for classification to obtain rs-fMRI classification results.

2. The rs-fMRI classification method based on residual block and Bayesian variational inference according to claim 1, characterized in that: The calculating brain functional connections based on the time signal sequence and constructing the original brain network comprises: Calculating the Pearson correlation coefficient of the time signal sequence as a node feature matrix of the brain map structure; A brain functional connection matrix is ​​determined based on the time signal sequence as an adjacency matrix of the brain map structure.

3. The rs-fMRI classification method based on residual block and Bayesian variational inference according to claim 2, characterized in that: The calculating brain functional connections based on the time signal sequence and constructing the original brain network also includes: The original brain network is sparsely constructed.

4. The rs-fMRI classification method based on residual block and Bayesian variational inference according to claim 1, characterized in that: The constructing a line graph based on the original brain network and performing line graph convolution comprises: Determine a binary adjacency matrix based on the original brain network, retain the original edge features and the sum of adjacent node features, and obtain a line graph; Graph convolution is performed based on the line graph to update the backbone residual block.

5. The rs-fMRI classification method based on residual block and Bayesian variational inference according to claim 1, characterized in that: The classifying based on the convolved line graph using Bayesian variational posterior includes: Optimize the network parameters of the Bayesian model using stochastic variational inference; The overall loss function is defined based on maximizing the evidence lower bound of the marginal likelihood and the cross entropy loss.

6. An rs-fMRI classification device based on residual blocks and Bayesian variational inference, characterized in that: The device comprises: A data acquisition module is used to acquire and preprocess rs-fMRI data to obtain a time signal series; An original brain network construction module, used to calculate brain functional connections and construct an original brain network based on the time signal sequence; A line graph convolution module, used to construct a line graph based on the original brain network and perform line graph convolution; The Bayesian variational reasoning and classification module is used to classify the convolved line graph using the Bayesian variational posterior to obtain the rs-fMRI classification result.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images