Method for structural-functional coupling analysis of brain diseases based on dynamic pulse neural network

Abstract

The application provides a brain disease structure-function coupling analysis method based on a dynamic pulse neural network, belongs to the technical field of intelligent auxiliary medical diagnosis, and effectively solves the technical problems of the existing structure-function coupling analysis method in capturing the dynamic changes of the brain, the nonlinear coupling characteristics and the integration of multi-modal data. The technical scheme is as follows: firstly, a dynamic pulse neural network model is constructed; then, a pulse structure connection matrix is extracted from an original fMRI time sequence; then, a pulse function connection matrix is extracted from a functional brain network, and the pulse function connection matrix is input into a pulse coupling pooling module to obtain a pulse structure-function coupling; finally, the obtained coupling information is input into a classification layer to obtain a disease recognition result. The application has the beneficial effect that the application has the significant advantage of providing a brand-new analysis tool and perspective for brain disease research, promoting the in-depth understanding of the neural mechanism of brain diseases, and showing wide application potential in clinical application.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent auxiliary medical diagnosis, and particularly relates to a brain disease structure-function coupling analysis method based on a dynamic pulse neural network. BACKGROUND

[0002] Brain diseases affect millions of people worldwide, including a variety of diseases and disorders, from stroke and Alzheimer's disease to multiple sclerosis, epilepsy, traumatic brain injury, etc., which are important medical conditions that have a profound impact on human health. Recent studies on the functional and structural networks of brain diseases have revealed specific disruptions in connectivity and network topology, thus leading to a shift in modern brain disease research from "focus" to "network". By analyzing brain networks, a new method can be provided to explore the correlation between brain function defects and related potential structural damage.

[0003] Structural connectivity (SC) is the path between physical white matter regions in the brain, which is usually derived from magnetic resonance imaging (MRI) or diffusion tensor imaging (DTI) images, and functional connectivity (FC) is the temporal co-activation between brain region activities, which is constructed from electroencephalogram (EEG) data and functional magnetic resonance imaging (fMRI) data. There is a close coupling between SC and FC, which is not only reflected in the inter-regional connection strength, but also in the connection patterns supporting various cognitive functions. SC-FC coupling is an emerging comprehensive measure that links functional and structural networks, and has been considered as a modality that is more sensitive to detecting subtle changes in brain activity than any single modality index.

[0004] There are some significant deficiencies in current structure-function coupling analysis, which restricts its in-depth application in the field of brain science. First, the existing methods are mostly based on linear or simple nonlinear models, which are difficult to fully capture the complex nonlinear coupling relationship between brain function and structure. As a complex dynamic system, the structural connection of the brain often affects functional activity through multiple levels, time delays and dynamic processes, and these characteristics are often simplified or ignored. Second, many analysis methods are based on static structural networks and functional connectivity matrices, ignoring the dynamic characteristics of brain function over time, and cannot effectively reveal the time-varying characteristics of the coupling relationship under different tasks or states.

[0005] The document "Relationships between brain structure-function coupling in normal aging and cognition: A cross-ethnicity population-based study" proposes a novel quantitative index, the brain structure-function coupling (C-BSF) index. This index is constructed by evaluating the topological similarity between the structure network and the function network, providing a new perspective and method for exploring the complex relationship between brain structure and function. However, the C-BSF index also has certain limitations, for example, it may be more suitable for capturing static structure-function coupling characteristics, and it is relatively insufficient in revealing the dynamic changes and real-time functional connectivity changes of the brain.

[0006] The document "Representing Functional Connectivity with Structural Detour: A New Perspective to Decipher Structure-Function Coupling Mechanism" proposes a new connectivityomics representation, detour connectivity (DC), which characterizes direct functional connectivity (FC) through weighted connection strength on the path of directed structural connectivity (SC), and also designs a Transformer-style graph neural network framework that effectively quantifies SC-FC coupling through self-attention mechanisms. Although this method can provide quantitative measures of structure-function coupling, the underlying neurobiological mechanisms behind these measures may be difficult to interpret, especially in complex neural networks.

[0007] Therefore, by utilizing the neurobiological mechanisms of dynamic spiking neurons to simulate neuron dynamics in the brain, the complex mechanisms of brain structure-function coupling can be learned and analyzed more deeply, thereby significantly improving the accuracy of brain disease diagnosis. SUMMARY

[0008] In view of the shortcomings of existing structure-function coupling analysis methods in capturing brain dynamic changes, nonlinear coupling characteristics, and integrating multi-modal data, the purpose of the present application is to provide a brain disease structure-function coupling analysis method based on dynamic spiking neural networks, to improve the accuracy of brain disease recognition.

[0009] The application idea of the application is: by combining multi-modal data of brain structure network and function network, using the biological heuristic characteristics of dynamic spiking neural network to simulate the dynamic behavior of neurons, to capture the complex relationship of brain structure-function coupling. Traditional structure-function coupling analysis methods often use static or linear models, which are difficult to fully depict the nonlinear dynamic characteristics of brain network. As a computational model that can simulate biological nervous system, dynamic spiking neural network has the unique advantage of capturing the time-varying activity pattern and nonlinear coupling relationship between neurons.

[0010] In order to better achieve the above application purpose, the application realizes the following technical scheme: a brain disease structure-function coupling analysis method based on dynamic spiking neural network, the specific steps are as follows:

[0011] A, a Leaky Integrate-and-Fire spiking neuron with an infinite impulse response (IIR) filter proposed by Fang et al. is used to construct a dynamic spiking neural network (DSNN) model. Let Axon(·) represent the operation of transmitting pulses to the next neuron, and let represent the behavior of updating membrane potential and firing pulses of neurons, therefore, the DSNN model simulated by the IIR filter can be divided into Axon(·) and operation;

[0012] B, the original functional time series information is extracted from functional magnetic resonance imaging fMRI, let represent the time series matrix of fMRI, N corresponds to the number of node sets of brain regions, T corresponds to the length of time series, and G fc is taken as the functional input of DSNN, and the pulse functional connection (FC) matrix is obtained through Axon(·) and operation;

[0013] C, the structural brain network is extracted from diffusion tensor imaging DTI, let represent the adjacency matrix of the structural brain network, N corresponds to the number of node sets of brain regions, then the edge-to-edge convolution E2E and edge-to-node convolution E2N operations proposed by Kawahara are used to obtain the feature map of the structural brain network information, and the feature map of the structural brain network after E2E convolution is denoted as the feature map after E2N convolution is denoted as where C1 and C2 represent the number of filters of the two convolution operations, and finally the structural brain network feature maps G sc after E2E and E2N convolution operations are taken as the structure input of DSNN, and the pulse structural connection (SC) matrix is obtained through linear network and operation;

[0014] D. Constructing a spike-coupled pooling (SCP) module, inputting the pulse FC and SC matrices into the SCP module to obtain the pulse structure-function (SC-FC) coupling, and finally obtaining the pulse output of each output neuron through a linear network and Operation, the pulse output of each output neuron over time It is a time-varying vector, where m is the number of output categories.

[0015] E. Input to the classification layer, obtain the predicted category by calculating the maximum pulse count of each output neuron, and train the cross-entropy loss of the predicted category and the real category as the loss function, to obtain a brain disease accurate identification model based on pulse structure-function coupling.

[0016] As the brain disease structure-function coupling analysis method based on dynamic pulse neural network provided by the application, the specific steps of step B are as follows:

[0017] a. The original fMRI time series matrix G fc As the functional input of DSNN, the formula is as follows:

[0018]

[0019] Wherein, Axon(·) represents the operation of transmitting pulses to the next neuron, and l is the layer index, when l=1, Y l-1 =G fc ;

[0020] b. Extract the pulse FC matrix using the dynamic pulse neuron model, and the formula is as follows:

[0021]

[0022] Wherein, represents the behavior of updating the membrane potential and firing pulses of neurons, and the final output pulse FC matrix is

[0023] As the brain disease structure-function coupling analysis method based on dynamic pulse neural network provided by the application, the specific steps of step C are as follows:

[0024] a. The SC feature map G sc As the structural input of DSNN, and implicit pulse coding is performed through internal pulse neurons, and the formula is as follows:

[0025]

[0026] Wherein, D is the feature dimension, and T is the time step. It is worth noting that the selection of T is determined according to the length of the fMRI time series, and Vsc G is obtained by sc After T times of repetition, the linear network obtains, is obtained by Therefore, it is a binary pulse tensor;

[0027] b, using a linear network and a dynamic pulse element model to extract a pulse SC matrix, the formula is as follows:

[0028]

[0029] Where, X sc represents the final output pulse SC matrix;

[0030] As the brain disease structure-function coupling analysis method based on a dynamic pulse neural network provided by the application, the specific steps of step D are as follows:

[0031] a, the function and structure input of SCP can be described as Y fc =(y0,...,y t ,...,y T ), and X sc =(x0,...,x t ,...,x T ), For the convenience of description, the processing process at time t (t∈[0,T]) will be described. Specifically, given the FC feature vector y t and the SC feature vector x t , the simplest bilinear transformation is defined as:

[0032]

[0033] Where, is a projection matrix, (tr) represents a transpose operator. In order to obtain an o-dimensional output, it is necessary to learn Thus, a large number of parameters are generated.

[0034] b, decompose W into two low-rank matrices:

[0035]

[0036] Where, k is the decomposition rank, is the element-wise multiplication of two feature vectors, is a full one vector, Therefore, in order to obtain an output feature vector

[0037] Two three-dimensional tensors and Without loss of generality, the above operations can be reshaped with simple reshape operations and can be respectively restated as two-dimensional matrices and

[0038] c, the final fusion vector is obtained by performing sum-pooling on the projection vector with a one-dimensional non-overlapping window of size k using P' and Q':

[0039]

[0040] The above operations generate a feature vector that fuses the FC temporal information and the SC spatial information It only represents the coupling relationship between FC and SC at time t. After repeating the a, b and c operations T times, a pulse matrix composed of T feature vectors can be obtained It is called pulse SC-FC coupling, which not only contains FC information in time, but also retains SC information in space, and comprehensively represents the dynamic coupling relationship between FC and SC.

[0041] d, the final output of each output neuron through the linear network and operation can be obtained

[0042]

[0043] where m is the number of output categories, is a vector that changes over time.

[0044] Compared with the prior art, the present application has the following beneficial effects:

[0045] 1. The brain disease structure-function coupling analysis method based on a dynamic pulse neural network provided by the present application can more effectively capture the nonlinear and dynamic coupling characteristics between brain structure and function compared to most previous structure-function coupling analysis models, overcomes the limitations of traditional methods in static and linear modeling, and thus provides more accurate and comprehensive analysis of brain coupling mechanisms.

[0046] 2. The brain disease structure-function coupling analysis method based on a dynamic pulse neural network provided by the present application innovatively integrates multi-modal data from fMRI and DTI, effectively improves the accuracy and robustness of data integration by combining the advantages of pulse neural networks, and can comprehensively reflect the structure and function interaction of the brain at different time scales, thereby improving the accuracy of brain disease recognition.

[0047] 3、The brain disease structure-function coupling analysis method based on the dynamic spiking neural network provides a new technical means and theoretical framework, can deeply reveal the structure-function coupling mechanism of the brain, has strong clinical transformation potential, and can play an important role in early diagnosis, prediction and treatment of brain diseases. BRIEF DESCRIPTION OF DRAWINGS

[0048] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, are used to explain the application, and do not constitute a limitation on the application.

[0049] Figure 1 The overall framework diagram of the brain disease structure-function coupling analysis method based on the dynamic spiking neural network of the application is shown in the figure.

[0050] Figure 2 The model diagram of the brain disease structure-function coupling analysis method based on the dynamic spiking neural network of the application is shown in the figure. DETAILED DESCRIPTION

[0051] In order to deepen the understanding of the application, the application will be further described below in combination with examples and drawings, and the examples are only used to explain the application, and do not constitute a limitation on the protection scope of the application.

[0052] Example 1

[0053] Referring to Figure 1 , Figure 2 , the embodiment is realized by the following technical solutions:

[0054] As Figure 1 and Figure 2 shown, the brain disease structure-function coupling analysis method based on the dynamic spiking neural network, the specific steps are as follows:

[0055] A, the Leaky Integrate-and-Fire spiking neuron with an infinite impulse response (IIR) filter proposed by Fang et al. is used to construct a dynamic spiking neural network (DSNN) model. Let Axon(·) represent the operation of transmitting pulses to the next neuron, and let represent the behavior of neuron updating membrane potential and firing pulses, therefore, the DSNN model simulated by the IIR filter can be divided into Axon(·) and operation;

[0056] B, the original functional time series information is extracted from functional magnetic resonance imaging fMRI, let represent the time series matrix of fMRI, N corresponds to the number of node sets of brain regions, T corresponds to the length of time series, and G fcAs the functional input of DSNN, the pulse functional connection (FC) matrix is obtained by Axon(·) and The specific steps are as follows:

[0057] a. The original fMRI time series matrix G fc As the functional input of DSNN, the formula is as follows:

[0058]

[0059] Wherein, Axon(·) represents the operation of transmitting pulses to the next neuron, and l is the layer index, when l = 1, Y l-1 = G fc ;

[0060] b. The pulse FC matrix is extracted by using the dynamic pulse neuron model, and the formula is as follows:

[0061]

[0062] Wherein, represents the behavior of neuron updating membrane potential and firing pulses, and the final output pulse FC matrix is denoted as

[0063] The number of samples contained in the brain disease data set is 217, and the decision class attribute is divided into two categories, namely diagnosed brain disease patients and normal people. The fMRI time series matrix is The finally generated pulse FC matrix is

[0064] C. The structural brain network is extracted from the diffusion tensor imaging DTI, and let denotes the adjacency matrix of the structural brain network, and N corresponds to the number of node sets of brain regions. Then the edge-to-edge convolution E2E and edge-to-node convolution E2N operation proposed by Kawahara are used to obtain the structural brain network information feature map, and the feature map of the structural brain network after E2E convolution is denoted as The feature map after E2N convolution is Wherein, C1 and C2 represent the filter number of the two convolution operations, and finally the structural brain network feature map G sc As the structural input of DSNN, the pulse structural connection (SC) matrix is obtained by linear network and The specific steps are as follows:

[0065] a. The SC feature map G sc As the structural input of DSNN, and through the implicit pulse coding of internal pulse neurons, the formula is as follows:

[0066]

[0067] where D is the feature dimension and T is the time step. It is worth noting that the choice of T is determined by the length of the fMRI time series, V sc By G sc After T repetitions, the linear network is used to obtain is obtained, so it is a binary pulse tensor;

[0068] b. Extract the pulse SC matrix using the linear network and the dynamic pulse element model, as follows:

[0069]

[0070] where X sc represents the final output pulse SC matrix;

[0071] The connection matrix of the structural brain network is The E2E convolutional layer consists of 32 1x90 and 32 90x1 filters, and the generated structural feature map is The E2N convolutional layer consists of 64 1x90x32 filters, and the generated structural feature map is

[0072] D. Construct the pulse coupled pooling (SCP) module, input the pulse FC and SC matrices into the SCP module to obtain the pulse structure-function (SC-FC) coupling, and finally pass through the linear network and Operation, the number of pulses emitted by each output neuron over time is a time-varying vector, where m is the number of output classes; the specific steps are as follows:

[0073] a. The function and structure inputs of SCP can be described as Y fc = (y0,..., y t ,..., y T ), and X sc = (x0,..., x t ,..., x T ), For ease of illustration, the processing process at time t (t ∈ [0, T]) will be highlighted. Specifically, given the FC feature vector y t and the SC feature vector x t at time t, the simplest bilinear transformation is defined as:

[0074]

[0075] where,​ is the transpose operator. To obtain an o-dimensional output, one needs to learn which results in a large number of parameters.

[0076] b. Decompose W into two low-rank matrices:

[0077]

[0078] where k is the decomposition rank, is the element-wise multiplication of two eigenvectors, is an all-one vector, Thus, to obtain an output eigenvector

[0079] two three-dimensional tensors and can be respectively reformulated as two-dimensional matrices and without loss of generality. and

[0080] c. The final fused eigenvector is obtained by performing sum-pooling on the projected eigenvector with a one-dimensional non-overlapping window of size k using P' and Q':

[0081]

[0082] The above operation generates a fused eigenvector which only represents the coupling relationship between FC and SC at time t. By repeating the operations a, b and c for T times, one can obtain a spike matrix composed of T eigenvectors which is called the spike SC-FC coupling. It not only contains the FC information in time, but also preserves the SC information in space, and fully characterizes the dynamic coupling relationship between FC and SC.

[0083] d. Finally, through linear network and operations, one can obtain the spikes emitted by each output neuron over time

[0084]

[0085] where m is the number of output classes, is a vector that varies over time.

[0086] E, input to the classification layer, obtain the predicted class by calculating the maximum pulse count of each output neuron, and train the cross-entropy loss of the predicted class and the real class as the loss function, obtain a brain disease accurate identification model based on pulse structure-function coupling.

[0087] The method of five-fold cross-validation is adopted, and the model is trained and tested independently for five rounds, and different training sets, test sets and validation sets are divided in each round. In these rounds, the highest accuracy recorded is 97.89%, and the comprehensive average accuracy is 93.64%. These data show that the model can significantly improve the accuracy of brain disease recognition.

[0088] Embodiment 2

[0089] The present embodiment is trained on an NVIDIA RTX 3090 GPU card, and six comparative models are selected for comparison, such as Statistics, MPCA, Self-calibrated, MMP-GCN, Brain-GRL and Brain-DNN.

[0090] In order to highlight the advantages of the model of the present application in running efficiency, the batch size will be continuously increased until the GPU memory usage exceeds the limit, and the optimal test result of the model is recorded. The initial batch size is set to 32, the epoch is 100, and the learning rate is 0.0001. The performance of the model is evaluated by the following indicators: accuracy (ACC), sensitivity (SEN) and specificity (SPE). The experimental results are shown in Table 1:

[0091] Table 1 Experimental results of the technology proposed by the present application and other technologies

[0092]

[0093] From Table 1, it can be concluded that the DSNN model proposed by the present application has obvious advantages in the recognition of brain diseases, not only is superior to existing methods in multiple evaluation indicators, but also provides a new perspective for the research of SC-FC coupling.

[0094] Embodiment 3

[0095] This embodiment will introduce a suitable scenario of the present application:

[0096] The coupling between structural connectivity (SC) and functional connectivity (FC) (SC-FC coupling) is a core mechanism of brain information processing and functional integration. In recent years, research has found that abnormalities in SC-FC coupling are closely related to the occurrence and development of various brain diseases, such as epilepsy, Alzheimer's disease, and schizophrenia. Therefore, a deep understanding of the mechanism of SC-FC coupling is crucial for early diagnosis and prediction of brain diseases. However, existing SC-FC coupling analysis methods mostly focus on revealing the static association between SC and FC, ignoring the dynamic characteristics of neural activity, especially the pulse timing of information transmission between brain regions.

[0097] To solve this problem, the present application proposes a brain disease structure-function coupling analysis method based on dynamic spiking neural network (DSNN). By quantifying the dynamic characteristics of SC-FC coupling, DSNN can help doctors assess the brain function status of different patients, and thus provide support for personalized treatment. For example, in the treatment process of patients with schizophrenia, DSNN can monitor the effects of drugs or therapies on brain networks, evaluate treatment effectiveness, and adjust treatment plans according to individual brain network states to improve the effectiveness of treatment. DSNN can also be applied to brain-computer interface systems and neurofeedback training. By analyzing the dynamic characteristics of SC-FC coupling between brain regions in real time, DSNN can provide more accurate real-time data for neurofeedback, thereby improving the brain function performance of individuals in specific tasks, such as attention training, emotion regulation, etc.

[0098] In summary, the practical application range of the DSNN model is very wide, covering multiple fields from clinical brain disease diagnosis to personalized treatment, neural regulation, brain imaging analysis, etc. By more accurately capturing and analyzing the spiking neural activity of the brain, it provides innovative ideas and technical support for solving many challenges in current brain science research and neurological disease treatment.

[0099] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for structure-function coupling analysis of brain diseases based on dynamic spiking neural networks, characterized in that, Includes the following steps: S1. A dynamic spiking neural network (DSNN) model is constructed using Leaky Integrate-and-Fire spiking neurons with infinite impulse response (IIR) filters. Let Axon(·) represent the operation of transmitting a pulse to the next neuron. This represents the behavior of neurons updating membrane potentials and firing impulses; therefore, DSNN models simulated by IIR filters are divided into Axon(·) and... operate; S2. Extract the raw functional time-series information from functional magnetic resonance imaging (fMRI), and let This represents the time series matrix of fMRI, where N corresponds to the number of node sets in the brain region, T corresponds to the time series length, and G... fc As functional inputs to the DSNN, via Axon(·) and Operation to obtain the pulse function connection matrix FC; S3. Extract structural brain networks from diffusion tensor imaging (DTI). Let N represent the adjacency matrix of the structural brain network, corresponding to the number of node sets in the brain region. Then, edge-to-edge convolution (E2E) and edge-to-node convolution (E2N) operations, proposed by Kawahara, are used to obtain the feature map of the structural brain network. Let the feature map of the structural brain network after E2E convolution be denoted as... The feature map after E2N convolution is Where C1 and C2 represent the number of filters in the two convolution operations, and finally the structural brain network feature map G after E2E and E2N convolution operations is... sc As the structural input of the DSNN, through a linear network and The operation yields the pulse structure connection matrix SC; S4. Construct a pulse-coupled pooling (SCP) module. Input the pulse matrix FC and the matrix SC into the SCP module to obtain the pulse structure-function SC-FC coupling. Finally, through a linear network and... The operation yields the pulses emitted by each output neuron over time. It is a vector that changes over time, where m is the number of output categories; S5. Input to the classification layer, calculate the maximum pulse count of each output neuron to obtain the predicted category, and use the cross-entropy loss between the predicted category and the true category as the loss function for training to obtain a precise brain disease identification model based on pulse structure-function coupling.

2. The method for structure-function coupling analysis of brain diseases based on dynamic spiking neural networks according to claim 1, characterized in that, The specific steps for obtaining the pulse matrix FC in step S2 are as follows: S21, the original fMRI time series matrix G fc The formula used as the functional input to the DSNN is as follows: Where Axon(·) represents the operation of transmitting the pulse to the next neuron, and l is the layer index. When l = 1, Y l-1 =G fc ; S22. Extract the pulse matrix FC using the dynamic pulse element model, as shown in the following formula: in, This represents the behavior of a neuron updating its membrane potential and firing impulses, denoted as the final output impulse matrix FC.

3. The method for structure-function coupling analysis of brain diseases based on dynamic spiking neural networks according to claim 1, characterized in that, The specific steps for obtaining the pulse matrix SC in step S3 are as follows: S31. The feature map G of the matrix SC after E2E and E2N convolution operations. sc As the structural input of the DSNN, and implicitly spiking encoded through internal spiking neurons, the formula is as follows: Where D is the feature dimension, T is the time step, and the choice of T is determined based on the length of the fMRI time series, V sc By G sc After T repetitions, the result is obtained through a linear network. It was through What we get is a binary pulse tensor; S32. Extract the pulse matrix SC using a linear network and a dynamic pulse element model, as shown in the following formula: Among them, X sc This represents the final output pulse matrix SC.

4. The method for structure-function coupling analysis of brain diseases based on dynamic spiking neural networks according to claim 1, characterized in that, The specific steps for obtaining pulsed SC-FC coupling in step S4 are as follows: The functional and structural inputs of S41 and SCP are respectively described as Y fc =(y0,...,y t ,...,y T ), and X sc =(x0,...,x t ,...,x T ), The focus will be on the processing at time t, where t∈[0,T], given the matrix FC and eigenvector y at time t. t Sum of matrix SC and eigenvectors x t The bilinear transform is defined as: in, Let (tr) be the projection matrix, and (tr) denote the transpose operator. To obtain an o-dimensional output, we need to learn... This results in a large number of parameters; S42. Decompose W into two low-rank matrices: Where k is the rank of the decomposition. It is an element-wise multiplication of two eigenvectors. It is a completely one vector. Therefore, in order to obtain an output feature vector Two 3D tensors are required and Use the reshape operation to and Represented as two-dimensional matrices respectively and S43, Final fusion vector It is obtained by performing summation pooling on the projection vector using P′ and Q′ on a one-dimensional non-overlapping window of size k: The above operations generate an eigenvector that integrates the temporal information of the FC matrix and the spatial information of the SC matrix. It only represents the coupling relationship between matrix FC and matrix SC at time t. After repeating operations S41, S42, and S43 T times, a pulse matrix consisting of T eigenvectors is obtained. This is called pulse SC-FC coupling, which not only contains the temporal matrix FC information, but also retains the spatial matrix SC information, comprehensively characterizing the dynamic coupling relationship between the matrix FC and the matrix SC. S44, ultimately through a linear network and The operation yields the pulses emitted by each output neuron over time. Where m is the number of output categories. It is a vector that changes over time.

Images