Schizophrenia classification method based on fuzzy hypergraph neural network of evidence theory

Abstract

This invention provides a schizophrenia classification method based on evidence theory using fuzzy hypergraph neural networks, belonging to the technical field of hypergraph neural networks and evidence theory. It solves the technical problem of high heterogeneity in schizophrenia identification tasks. The technical solution is as follows: First, an excellent hyperedge granularity model is constructed using sparse constraint functions to exclude nodes with high heterogeneity from the hyperedge granularity model. Then, during the construction of the fuzzy hypergraph, evidence theory is applied to fuse association quality functions and distance quality functions. Fuzzy membership degrees are used to characterize node weights, increasing the weights of nodes with lower heterogeneity. Finally, a fuzzy hypergraph convolution model is used to identify the labels of schizophrenia patients, improving the classification accuracy of schizophrenia data and optimizing semantic interpretation. The beneficial effects of this invention are: improving the accuracy and interpretability of schizophrenia diagnosis.

Description

Technical Field

[0001] This invention relates to the fields of hypergraph neural networks and evidence theory, specifically to a schizophrenia classification method based on fuzzy hypergraph neural networks of evidence theory. Background Technology

[0002] Schizophrenia is a severe mental disorder that causes hallucinations, delusions, and extremely confused thinking and behavior, affecting patients' daily lives and potentially leading to disability. The paper "Cognitive impairment in schizophrenia: aetiology, pathophysiology, and treatment" (Molecular Psychiatry, 2023) estimates approximately 24 million cases of schizophrenia worldwide. The paper "The Current Status of Alexithymia in Elderly Patients with Schizophrenia and Its Impact on Mental Symptoms" (Henan Medical Research, 2024) proposes that the pathogenesis of schizophrenia is a complex, multi-factorial, and multi-step process, relying heavily on physician experience for diagnosis, resulting in high uncertainty and poor treatment outcomes. Furthermore, due to a lack of experienced physicians and specialized equipment, the predictive accuracy for schizophrenia is low.

[0003] To improve the diagnostic accuracy of schizophrenia, the paper "Brain network analysis of schizophrenia patients based on hypergraph signal processing" (IEEE Transactions on Image Processing, 2023) applied hypergraph brain networks to explore the pathogenesis of schizophrenia. It constructed a realistically weighted hypergraph brain network and converted matrices into weighted adjacency tensors to improve the classification accuracy of schizophrenia. However, due to the high heterogeneity of multi-source, heterogeneous schizophrenia data from different hospitals and institutions, this model cannot model high-order correlations between multimodal data, hindering subsequent diagnostic classification by medical personnel.

[0004] The paper "Multi-Modal imaging genetics data fusion via a hypergraph-based manifold regularization: application to schizophrenia study" (IEEE Transactions on Medical Imaging, 2022) applies hypergraph fusion to multimodal schizophrenia data. By fusing heterogeneous structural information between different modalities, it characterizes higher-order relationships between schizophrenia data, improving the prediction accuracy of schizophrenia. However, due to the irregular distribution and unstable quality of schizophrenia data, the hypergraph modeling in this paper suffers from uncertainty and an increased misdiagnosis rate. With the increasing maturity of deep learning technology, massive amounts of schizophrenia data have been processed and stored, providing technical assurance and data support for the development and research of schizophrenia data. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, this invention provides a schizophrenia classification method based on evidence theory and fuzzy hypergraph neural networks, which reduces the heterogeneity of schizophrenia data, handles the uncertainty in the diagnosis of schizophrenia patients, and improves the prediction accuracy of schizophrenia.

[0006] The inventive concept of this invention is as follows: The method constructs an excellent hyperedge granularity model using sparse constraint functions to handle unevenly distributed schizophrenia data, excluding nodes with high heterogeneity from the hyperedge granularity model and reducing data heterogeneity. Then, during the construction of the fuzzy hypergraph, evidence theory is applied to fuse association quality functions and distance quality functions. Fuzzy membership degrees characterize node weights, increasing the weights of nodes with low heterogeneity and characterizing the classification uncertainty of nodes with high heterogeneity. Finally, a fuzzy hypergraph convolution model is used to identify the labels of schizophrenia patients, improving the classification accuracy of schizophrenia data and optimizing semantic interpretation.

[0007] The technical solution adopted in this invention is: a schizophrenia classification method based on evidence theory and fuzzy hypergraph neural networks, comprising the following steps:

[0008] Step 1: Select data from schizophrenia patients at 5 different sites as the dataset; use the DPARSF tool to process the data, including preprocessing methods such as inter-layer time correction, spatial standardization, global average intensity normalization, interference signal regression, and bandpass filtering; use the widely used automatic anatomical marker atlas template to extract the time series of regions of interest (ROIs). This template divides the cerebral cortex into 90 adjacent and functionally similar brain regions, obtaining time series information of different functional brain regions;

[0009] Step 2: Complete the hyperedge modeling task through sparse constraint functions: Based on the sparse constraint functions proposed in this invention, construct the correlation matrix of schizophrenia data nodes; characterize the optimal k value of different schizophrenia data nodes through the correlation matrix and correlation threshold; construct a hyperedge granular model based on k nearest neighbor samples to characterize the higher-order correlation between schizophrenia data.

[0010] Step 3: Construct a fuzzy hypergraph using evidence theory: Based on the association matrix of the schizophrenia data nodes obtained in Step 2, construct an association quality function; calculate a distance quality function using the distance between the schizophrenia data nodes; apply the quality function to characterize the association coefficient of the schizophrenia data nodes with respect to the hyperedges; apply evidence theory, fuse the quality functions, and calculate the fuzzy membership degree of the schizophrenia data nodes with respect to the hyperedges; obtain the fuzzy adjacency matrix of the fuzzy hypergraph, thus constructing the fuzzy hypergraph.

[0011] Step 4: Based on the fuzzy adjacency matrix H obtained in Step 3 F And a hypergraph neural network model, constructing a fuzzy hypergraph convolution model: based on the fuzzy adjacency matrix H F Features of schizophrenia data nodes and hyperedges are aggregated to construct a fuzzy hypergraph convolution operator. Based on the fuzzy hypergraph convolution operator, a two-layer convolutional structure is designed for schizophrenia data node label prediction to obtain the labels of schizophrenia subjects.

[0012] As a preferred technical solution of the present invention, the specific process of step 2 is as follows:

[0013] Step 2.1: Establish a hyperedge granularity model using sparse constraint functions, where the feature matrix X of the schizophrenia data nodes is represented as... n represents the number of data nodes related to schizophrenia, and d represents the feature dimension of the nodes; by minimizing the reconstruction error of the schizophrenia data nodes, the correlation matrix R between the nodes is obtained; the sparse constraint function is calculated:

[0014]

[0015] Among them, ||·|| FLet ||·||2| denote the Frobenius norm of the matrix, ||·||2 be the l2-norm regularization term, L denote the Laplacian matrix, ρ1 and ρ2 be adjustment parameters, Tr(·) denote the function for calculating the trace of the matrix, and R T With X T Let R be the transpose of X;

[0016] Step 2.2: Based on the sparse constraint function in Step 2.1, obtain the correlation matrix R between the schizophrenia data nodes; based on the correlation threshold σ, determine the optimal k value for all schizophrenia data nodes; for schizophrenia data node x... i compute node x i The optimal k value i :

[0017] k i =num(R(i)>σ) (17)

[0018] Where R(i) is the i-th column of the correlation matrix R, and num is the counting function;

[0019] Step 2.3: Based on the optimal k value obtained in Step 2.2, calculate the k nearest neighbor samples and use them as the schizophrenia data nodes within the hyperedge granularity; for the i-th hyperedge, calculate the hyperedge granularity HG. k (x i ):

[0020]

[0021] Where, k i For schizophrenia data node x i The optimal value of k, d(x m ,x n Let x be any two schizophrenia data nodes. m With x n The Euclidean distance function between them.

[0022] As a preferred technical solution of the present invention, the specific process of step 3 is as follows:

[0023] Step 3.1: Obtain the superedge through step 2. x i Let x be any node in the schizophrenia data set; calculate the schizophrenia data node x. i With the central node x j Euclidean distance d(x) i ,x j ), obtain node x i With super-edge e j Distance similarity re ijCalculate the data node x for schizophrenia i For the distance quality function of hyperedge association:

[0024]

[0025]

[0026] Where m is the distance adjustment parameter, k j For the superedge e j Number of internal nodes;

[0027] Step 3.2: Based on the association vector R(j) obtained in Step 2.2, obtain the hyperedge. The correlation coefficients of all schizophrenia data nodes within the superedge with respect to the hyperedge; based on any schizophrenia data node x within the hyperedge. i For the hyperedge e j The correlation coefficient is used to calculate the schizophrenia data node x. i For the correlation mass function of the hyperedge:

[0028]

[0029]

[0030] Among them, superedge e j The number of nodes is k j Super-edge e j The correlation coefficient between the i-th node and the hyperedge is defined as r. ij ;

[0031] Step 3.3: Based on the schizophrenia data node x from Steps 3.1 and 3.2 i For the hyperedge e j The distance quality function and correlation quality function are used to calculate the schizophrenia data node x. i For the hyperedge e j Fuzzy membership degree u ij :

[0032]

[0033]

[0034]

[0035]

[0036] Among them, X A and X B For the data node x containing schizophrenia data i The set of nodes, where ρ is the evidence adjustment parameter;

[0037] Step 3.4: Obtain the schizophrenia data node x through the calculation in Step 3.3. i For the hyperedge e j Fuzzy membership degree u ij That is, the corresponding element in the j-th column of the fuzzy adjacency matrix of the fuzzy hypergraph; repeat step 3.3 to calculate the fuzzy adjacency matrix H. F :

[0038]

[0039] Where n is the number of rows in the node matrix X, and i and j are any number of rows and columns in the matrix.

[0040] As a preferred embodiment of the present invention, step 4 is specifically performed as follows:

[0041] Step 4.1: Obtain the fuzzy adjacency matrix H based on step 3. F This process obtains the fuzzy membership degree of schizophrenia data nodes to hyperedges and the correlation coefficient of hyperedges to schizophrenia data nodes; based on the fuzzy membership degree and node-edge-node aggregation operation, it aggregates the features of hyperedges and nodes to obtain the fuzzy convolution operator.

[0042]

[0043] Among them, X l With X (l+1) Let H be the feature representation of the fuzzy hypergraph at layers l and l+1, where σ(·) is the activation function and H is the feature representation of the hypergraph at layers l and l+1. F T For matrix H F The transpose of D v and D e For nodes and hyperedges based on H F The degree matrix, W is the identity matrix, Θ l Let be the trainable weight parameter matrix of layer l;

[0044] Step 4.2: Based on the fuzzy convolution operator obtained in Step 4.1, a two-layer convolutional structure is designed for predicting the node labels of schizophrenia data, resulting in the label Z of the schizophrenia subject:

[0045] Z = Softmax(H(ReLU(HXΘ)) (0) ))Θ (1) (29)

[0046]

[0047] Where, Θ (0) Let Θ be the weight matrix from the first layer to the hidden layer. (1)Let be the weight matrix from the hidden layer to the last layer, and Softmax(·) and ReLU(·) be the activation functions.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0049] (1) Compared with traditional subjective treatment that relies on doctors’ experience, this invention can reduce the misdiagnosis rate caused by doctors’ lack of experience and further assist doctors in completing accurate treatment of schizophrenia.

[0050] (2) Compared with the traditional hypergraph neural network classification method for schizophrenia, the fuzzy hypergraph neural network proposed in this invention can reduce the uncertainty caused by the characteristics of schizophrenia data such as multiple data types, irregular distribution, unstable quality, and large differences in data structure between different modalities, and provides interpretability. It has been widely favored by medical staff and provides strong model support for the prediction and identification of schizophrenia.

[0051] (3) Compared with traditional schizophrenia classification methods, the schizophrenia classification method based on evidence theory and fuzzy hypergraph neural network proposed in this invention uses fuzzy hypergraphs to reduce heterogeneity caused by multi-source heterogeneous data, so that medical personnel can obtain more objective and accurate medical data and decision-making solutions. This is the important theoretical value and practical application of this invention. Attached Figure Description

[0052] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0053] Figure 1 This is a schematic diagram of the overall framework of the schizophrenia classification method based on evidence theory and fuzzy hypergraph neural network of the present invention.

[0054] Figure 2 This is a schematic diagram illustrating the specific structure of the schizophrenia classification method based on evidence theory and fuzzy hypergraph neural network of the present invention.

[0055] Figure 3 This is a schematic diagram of the hyperedge modeling of the fuzzy hypergraph neural network based on evidence theory according to the present invention.

[0056] Figure 4 This is a schematic diagram of fuzzy hypergraph convolution in the fuzzy hypergraph neural network based on evidence theory according to the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and experimental results. Of course, the specific experimental results described herein are only used to explain the invention and demonstrate its superiority, and are not intended to limit the invention.

[0058] Example 1

[0059] See Figures 1 to 4 This embodiment provides a schizophrenia classification method based on evidence theory and fuzzy hypergraph neural networks, including the following steps:

[0060] Step 1: In this embodiment, eight nodes are selected from schizophrenia datasets from five different sites as examples; the data is preprocessed using the DPARSF tool; the time series of regions of interest are extracted using an automatic anatomical label atlas template, and the cerebral cortex is divided into 90 functional brain regions that are adjacent in location and similar in function, obtaining time series information of different functional brain regions. The category labels of the eight nodes are 0,0,0,0,0,1,1,1; where 0 represents a healthy subject and 1 represents a schizophrenia patient.

[0061] Step 2: In this embodiment, the hyperedge modeling task is completed through sparse constraint functions: Based on the sparse constraint functions, an association matrix of 8 nodes is constructed; the optimal k value of the 8 nodes is characterized by the association matrix and the correlation threshold; and a hyperedge granular model is constructed based on k-nearest neighbor samples to characterize the higher-order correlations between schizophrenia data.

[0062] Step 2.1: In this embodiment, a hyperedge granularity model is established through a sparse constraint function, where the node feature matrix X is represented as... The number of nodes, n, is 8, and the feature dimension d of the nodes is 90. The association matrix R between the 8 nodes is obtained by minimizing the reconstruction error of the nodes. After calculating the association matrix and removing the diagonal, the association matrix R is:

[0063]

[0064] Step 2.2: Based on the association matrix R among the 8 nodes; set the association threshold σ to 0.1, determine the optimal k value for all nodes; calculate the optimal k value vector K for the 8 nodes using the absolute value of the values ​​in the association matrix and the threshold:

[0065] K = [3 2 3 3 3 4 2 2] (2)

[0066] Step 2.3: Based on the optimal k value obtained in Step 2.2, calculate the k nearest neighbor samples and use the k nearest neighbor samples as nodes within the hyperedge granularity; for example, for the first hyperedge, it contains nodes x2, x3, and x4; for the eighth hyperedge, it contains nodes x3 and x4.

[0067] Step 3: In this embodiment, a fuzzy hypergraph is constructed using evidence theory: Based on the association matrix of the 8 nodes obtained in Step 2, an association quality function is constructed; the distance quality function is calculated using the distance between the 8 nodes; the quality function is applied to characterize the association coefficient of the 8 nodes with respect to the hyperedge; evidence theory is applied to fuse the quality function and calculate the fuzzy membership degree of the 8 nodes with respect to the hyperedge; the fuzzy adjacency matrix of the fuzzy hypergraph is obtained, thus constructing the fuzzy hypergraph.

[0068] Step 3.1: In this embodiment, taking the first hyperedge as an example, through step 2, we obtain the hyperedge e1 = {x2, x3, x4}. The distance mass functions of the three nodes x2, x3, x4 in relation to the hyperedge e1 and the three nodes in relation to the hyperedge e1 are 0.332, 0.332, 0.286, and 0.05, respectively.

[0069] Step 3.2: Based on the association vector R(1) obtained in Step 2.2, obtain the association coefficients of all nodes in hyperedge e1 = {x2, x3, x4} with respect to the hyperedge; the distance quality functions of the association of the three nodes x2, x3, x4 with hyperedge e1 and the complete association of the three nodes with hyperedge e1 are 0.209, 0.532, 0.209, 0.05, respectively;

[0070] Step 3.3: Based on the distance quality function and association quality function of the three nodes x2, x3, x4 with respect to hyperedge e1 from Steps 3.1 and 3.2, calculate the fuzzy membership degree u of the three nodes x2, x3, x4 with respect to hyperedge e1. 21 ,u 31 ,u 41 : 0.256, 0.409, 0.251;

[0071] Step 3.4: Calculate the fuzzy membership degree u of the three nodes x2, x3, x4 with respect to the hyperedge e1 using the method described in Step 3.3. 21 ,u 31 ,u 41 This corresponds to the element in the first column of the fuzzy adjacency matrix of the fuzzy hypergraph; repeat step 3.3 to calculate the fuzzy adjacency matrix H. F :

[0072]

[0073] Step 4: In this embodiment, the fuzzy adjacency matrix H obtained in step 3 is used as a basis. F And a hypergraph neural network model, constructing a fuzzy hypergraph convolution model: based on the fuzzy adjacency matrix H F The features of nodes and hyperedges are aggregated to construct a fuzzy hypergraph convolution operator; based on the fuzzy hypergraph convolution operator, a two-layer convolution structure for node label prediction is designed to obtain node labels.

[0074] Step 4.1: In this embodiment, taking the 1st and 8th hyperedges as examples, the fuzzy adjacency matrix H is obtained based on step 3. F Obtain the fuzzy membership degrees u of the three nodes x2, x3, x4 with respect to the hyperedge e1. 21 ,u 31 ,u 41 And the fuzzy membership degree u of the two nodes x3 and x4 with respect to the hyperedge e8. 38 ,u 48 Based on fuzzy membership and node-edge-node aggregation operations; first, according to nodes x2, x3, x4 and fuzzy membership u 21 ,u 31 ,u 41 Update the features of the superedge e1 based on nodes x3, x4 and fuzzy membership degree u. 38 ,u 48 Update the superedge e8 features; secondly, based on the fuzzy membership degree u... 31 ,u 38 The features of node x3 are updated along with the features of two superedges e1 and e8. The update process of node x4 is similar to that of x3, thus completing the node-edge-node aggregation operation.

[0075] Step 4.2: Based on the fuzzy convolution operator obtained in Step 4.1, design a two-layer convolutional structure for node label prediction; through the two-layer convolutional structure and the updated node feature representation, complete the label prediction, and the prediction result is 0,0,0,0,0,1,1,1, which is consistent with the true label.

[0076] Example 2

[0077] See Figures 1 to 4 This embodiment provides a schizophrenia classification method based on evidence theory and fuzzy hypergraph neural networks, including the following steps:

[0078] Step 1: In this embodiment, 773 nodes were selected from schizophrenia datasets from five different sites as examples; the data were preprocessed using the DPARSF tool; the time series of regions of interest were extracted using an automatic anatomical label atlas template, and the cerebral cortex was divided into 90 functional brain regions that are adjacent in location and have similar functions, so as to obtain time series information of different functional brain regions. This dataset has two labels, namely healthy subjects and schizophrenia patients.

[0079] Step 2: In this embodiment, the hyperedge modeling task is completed through sparse constraint functions: Based on the sparse constraint functions, an association matrix of 773 nodes is constructed; the optimal k value of the 773 nodes is characterized by the association matrix and the correlation threshold; and a hyperedge granular model is constructed based on k-nearest neighbor samples to characterize the higher-order correlations between schizophrenia data.

[0080] Step 2.1: In this embodiment, a hyperedge granularity model is established through a sparse constraint function, where the node feature matrix X is represented as... The number of nodes, n, is 773, and the node feature dimension, d, is 90. The association matrix R among the 773 nodes is obtained by minimizing the node reconstruction error. The sparsity constraint function is then calculated.

[0081]

[0082] Among them, ||·|| F Let ||·||2| represent the Frobenius norm of the matrix, ||·||2 be the l2-norm regularization term, L denote the Laplacian matrix, ρ1 and ρ2 be adjustment parameters defined as 0.1 and 0.01 respectively, Tr(·) denote the function for calculating the trace of the matrix, and R T With X T Let R be the transpose of X;

[0083] Step 2.2: Based on the sparse constraint function in Step 2.1, obtain the association matrix R among the 773 schizophrenia data nodes; based on the association threshold σ = 0.1, determine the optimal k value for the 773 schizophrenia data nodes; for schizophrenia data node x... i compute node x i The optimal k value i :

[0084] k i =num(R(i)>σ) (32)

[0085] Where R(i) is the i-th column of the correlation matrix R, and num is the counting function;

[0086] Step 2.3: Based on the optimal k value obtained in Step 2.2, calculate the k nearest neighbor samples and use them as the schizophrenia data nodes within the hyperedge granularity; for the i-th hyperedge, calculate the hyperedge granularity HG. k (x i ):

[0087]

[0088] Where, k i For schizophrenia data node x i The optimal value of k, d(x m ,x n Let x be any two schizophrenia data nodes. m With x n The Euclidean distance function between them.

[0089] Step 3: In this embodiment, a fuzzy hypergraph is constructed using evidence theory: Based on the association matrix of the 773 nodes obtained in Step 2, an association quality function is constructed; the distance quality function is calculated using the distances between the 773 nodes; the quality function is applied to characterize the association coefficients of the 773 nodes with respect to the hyperedges; evidence theory is applied to fuse the quality functions and calculate the fuzzy membership degrees of the 773 nodes with respect to the hyperedges; the fuzzy adjacency matrix of the fuzzy hypergraph is obtained, thus constructing the fuzzy hypergraph.

[0090] Step 3.1: Obtain the superedge through step 2. x i Let x be any node in the schizophrenia data set; calculate the schizophrenia data node x. i With the central node x j Euclidean distance d(x) i ,x j ), obtain node x i With super-edge e j Distance similarity re ij Calculate the data node x for schizophrenia i For the distance quality function of hyperedge association:

[0091]

[0092]

[0093] Where m is the distance adjustment parameter, defined as 2, k j For the superedge e j Number of internal nodes;

[0094] Step 3.2: Based on the association vector R(j) obtained in Step 2.2, obtain the hyperedge. The correlation coefficients of all schizophrenia data nodes within the superedge with respect to the hyperedge; based on any schizophrenia data node x within the hyperedge. i For the hyperedge e j The correlation coefficient is used to calculate the schizophrenia data node x. i For the correlation mass function of the hyperedge:

[0095]

[0096]

[0097] Among them, superedge e j The number of nodes is k j Super-edge e j The correlation coefficient between the i-th node and the hyperedge is defined as r. ij ;

[0098] Step 3.3: Based on the schizophrenia data node x from Steps 3.1 and 3.2 i For the hyperedge e j The distance quality function and correlation quality function are used to calculate the schizophrenia data node x. i For the hyperedge e j Fuzzy membership degree u ij :

[0099]

[0100]

[0101]

[0102]

[0103] Among them, X A and X B For the data node x containing schizophrenia data i The set of nodes, where ρ is the evidence adjustment parameter, defined as 0.01;

[0104] Step 3.4: Obtain the schizophrenia data node x through the calculation in Step 3.3. i For the hyperedge e j Fuzzy membership degree u ij That is, the corresponding element in the j-th column of the fuzzy adjacency matrix of the fuzzy hypergraph; repeat step 3.3 to calculate the fuzzy adjacency matrix H. F :

[0105]

[0106] Where n is the number of rows in the node matrix X, defined as 773, and i and j are any number of rows and columns in the matrix.

[0107] Step 4: In this embodiment, the fuzzy adjacency matrix H obtained in step 3 is used as a basis. F And a hypergraph neural network model, constructing a fuzzy hypergraph convolution model: based on the fuzzy adjacency matrix H F The features of nodes and hyperedges are aggregated to construct a fuzzy hypergraph convolution operator; based on the fuzzy hypergraph convolution operator, a two-layer convolution structure for node label prediction is designed to obtain node labels.

[0108] Step 4.1: Obtain the fuzzy adjacency matrix H based on step 3. F This process obtains the fuzzy membership degree of schizophrenia data nodes to hyperedges and the correlation coefficient of hyperedges to schizophrenia data nodes; based on the fuzzy membership degree and node-edge-node aggregation operation, it aggregates the features of hyperedges and nodes to obtain the fuzzy convolution operator.

[0109]

[0110] Among them, X l With X (l+1) Let H be the feature representation of the fuzzy hypergraph at layers l and l+1, where σ(·) is the activation function and H is the feature representation of the hypergraph at layers l and l+1. F T For matrix H F The transpose of D v and D e For nodes and hyperedges based on H F The degree matrix, W is the identity matrix, Θ l Let be the trainable weight parameter matrix of layer l;

[0111] Step 4.2: Based on the fuzzy convolution operator obtained in Step 4.1, a two-layer convolutional structure is designed for predicting the node labels of schizophrenia data, resulting in the label Z of the schizophrenia subject:

[0112] Z = Softmax(H(ReLU(HXΘ)) (0) ))Θ (1) (44)

[0113]

[0114] Where, Θ (0) Let Θ be the weight matrix from the first layer to the hidden layer. (1) Let be the weight matrix from the hidden layer to the last layer, and Softmax(·) and ReLU(·) be the activation functions.

[0115] To highlight the superiority of the model constructed in this embodiment, five popular graph neural networks and hypergraph neural networks are selected as comparison algorithms: GCN, HGNN, HyperGCN, and HGNN+. This embodiment uses accuracy metrics to characterize the performance of the model designed in this embodiment and the comparison algorithms.

[0116] GC N is the most widely used model in node classification tasks. It encodes schizophrenia data into a topological graph and inputs it into a neural network to complete semi-supervised classification of schizophrenia patients. The HyperGCN model aims to construct a hypergraph model and transform it into a topological graph for semi-supervised learning tasks based on hypergraphs, further improving the prediction accuracy of schizophrenia patients. HGNN is used to encode high-order data correlations in the hypergraph structure, effectively handling multimodal schizophrenia data and applying the correlations of different modalities to improve the classification accuracy of schizophrenia. HGNN+ is an extension of the original HGNN model, further fusing data from different modalities of schizophrenia to reduce the misdiagnosis rate of schizophrenia.

[0117] The model proposed in this embodiment is based on a hypergraph neural network. Addressing the heterogeneity in schizophrenia data, fuzzy membership is applied to characterize the uncertainty between nodes and hyperedges. Furthermore, evidence theory is applied to fuse different association quality functions, constructing an evidence theory-based fuzzy hypergraph neural network. This embodiment designs an evidence theory-based fuzzy hypergraph neural network classification method with a training set of 653 nodes and a test set of 120 nodes. The performance of different graph and hypergraph neural network classification algorithms is shown in Table 1.

[0118] Table 1. Experimental results of hypergraph neural networks on the schizophrenia dataset.

[0119]

[0120] The results of this embodiment and other hypergraph neural network classification methods are shown in Table 1. FHGNN is the model proposed in this invention. Compared with other hypergraph neural networks GCN, HyperGCN, HGNN, and HGNN+, the classification accuracy of FHGNN is 60.5%, 59.6%, 63.0%, 61.5%, and 71.4%, respectively. The above experimental results demonstrate that FHGNN can reduce the heterogeneity in schizophrenia data and improve the prediction accuracy for schizophrenia patients.

[0121] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A schizophrenia classification method based on fuzzy hypergraph neural networks using evidence theory, characterized in that, Includes the following steps: Step 1: Select data from schizophrenia patients at 5 different sites as the dataset; use the DPARSF tool to process the data, including preprocessing methods such as inter-layer time correction, spatial standardization, global average intensity normalization, interference signal regression, and bandpass filtering; use an automatic anatomical marker atlas template to extract the time series of the Region of Interest (ROI). This template divides the cerebral cortex into 90 adjacent and functionally similar brain regions, obtaining time series information of different functional brain regions; Step 2: Complete the hyperedge modeling task using sparse constraint functions: Construct the association matrix of schizophrenia data nodes based on sparse constraint functions; characterize the optimal k value of different schizophrenia data nodes using the association matrix and correlation threshold; construct a hyperedge granular model based on k-nearest neighbor samples to characterize the higher-order correlations between schizophrenia data. Step 2 includes the following steps: Step 2.1: Establish a hyperedge granularity model using sparse constraint functions, where the feature matrix X of the schizophrenia data nodes is represented as... Where n represents the number of data nodes related to schizophrenia, and d represents the feature dimension of the nodes; the correlation matrix R between nodes is obtained by minimizing the reconstruction error of the schizophrenia data nodes; the sparse constraint function is calculated: ； in, Denotes the Frobenius norm of a matrix. Here, L is the l2-norm regularization term, and L denotes the Laplacian matrix. and To adjust the parameters, The function that calculates the trace of a matrix. and Let R be the transpose of X; Step 2.2: Based on the sparse constraint function in Step 2.1, obtain the correlation matrix R between the schizophrenia data nodes; based on the correlation threshold... Determine the optimal k value for all schizophrenia data nodes; for each schizophrenia data node... compute nodes The optimal k value : ； in, Let num be the i-th column of the correlation matrix R, and num be the counting function. Step 3: Construct a fuzzy hypergraph using evidence theory: Based on the association matrix of the schizophrenia data nodes obtained in Step 2, construct an association quality function; calculate a distance quality function using the distance between the schizophrenia data nodes; apply the quality function to characterize the association coefficient of the schizophrenia data nodes with respect to the hyperedges; apply evidence theory, fuse the quality functions, and calculate the fuzzy membership degree of the schizophrenia data nodes with respect to the hyperedges; obtain the fuzzy adjacency matrix of the fuzzy hypergraph, thus constructing the fuzzy hypergraph. Step 4, obtaining a fuzzy adjacency matrix H based on the fuzzy adjacency matrix H obtained in step 3 F and a hypergraph neural network model, constructing a fuzzy hypergraph convolution model based on the fuzzy adjacency matrix H F Aggregating the features of the schizophrenia data nodes and hyperedges to construct a fuzzy hypergraph convolution operator; based on the fuzzy hypergraph convolution operator, designing a two-layer convolution structure for schizophrenia data node label prediction to obtain the labels of the schizophrenia subjects.

2. The schizophrenia classification method based on evidence theory and fuzzy hypergraph neural network according to claim 1, characterized in that, Step 2 also includes the following steps: Step 2.3: Based on the optimal k value obtained in Step 2.2, calculate the k nearest neighbor samples and use them as the schizophrenia data nodes within the hyperedge granularity; for the i-th hyperedge, calculate the hyperedge granularity. : ； in, Data nodes for schizophrenia The optimal value of k, , For any two schizophrenia data nodes and The Euclidean distance function between them.

3. The schizophrenia classification method based on evidence theory and fuzzy hypergraph neural network according to claim 2, characterized in that, Step 3 includes the following steps: Step 3.1: Obtain the superedge through step 2. , Let be any node in the schizophrenia data set; calculate the schizophrenia data node. With the central node European distance Obtain node With super-edge Distance similarity Calculate schizophrenia data nodes For the distance quality function of hyperedge association: ； ； Where m is the distance adjustment parameter. For super-edge Number of internal nodes; Step 3.2: Based on the association vector obtained in Step 2.2 Obtain super-edge The correlation coefficients of all schizophrenia data nodes within the superedge with respect to the hyperedge; based on any schizophrenia data node within the hyperedge... For hyperedge The correlation coefficient was used to calculate the data nodes for schizophrenia. For the correlation mass function of the hyperedge: ； ； Among them, hyper-edge The number of nodes is Super-edge The correlation coefficient between the i-th node and the hyperedge is defined as r. ij ; Step 3.3: Based on the schizophrenia data nodes in Steps 3.1 and 3.2 For hyperedge The distance quality function and correlation quality function are used to calculate the schizophrenia data nodes. For hyperedge Fuzzy membership degree : ； ； ； ； in, and For nodes containing schizophrenia data The set of nodes, Evidence adjustment parameters; Step 3.4: Obtain the schizophrenia data node through the calculation in Step 3.

3. For hyperedge Fuzzy membership That is, the corresponding element in the j-th column of the fuzzy adjacency matrix of the fuzzy hypergraph; repeat step 3.3 to calculate the fuzzy adjacency matrix. : ； Where n is the number of rows in the node matrix X, and i and j are any number of rows and columns in the matrix.

4. The schizophrenia classification method based on evidence theory and fuzzy hypergraph neural network according to claim 3, characterized in that, Step 4 includes the following steps: Step 4.1: Obtain the fuzzy adjacency matrix based on Step 3. This process obtains the fuzzy membership degree of schizophrenia data nodes to hyperedges and the correlation coefficient of hyperedges to schizophrenia data nodes; based on the fuzzy membership degree and node-edge-node aggregation operation, it aggregates the features of hyperedges and nodes to obtain the fuzzy convolution operator. ； in, and For the feature representation of the fuzzy hypergraph at layers l and l+1, For activation function, For matrix transpose, and D e For nodes and hyperedges based The degree matrix, W is the identity matrix. Let be the trainable weight parameter matrix of layer l; Step 4.2: Based on the fuzzy convolution operator obtained in Step 4.1, a two-layer convolutional structure is designed for predicting the node labels of schizophrenia data, resulting in the label Z of the schizophrenia subject: ； ； in, This is the weight matrix from the first layer to the hidden layer. This is the weight matrix from the hidden layer to the last layer. and This is the activation function.

Images