A brain network coupling analysis system and method based on structural connection topology characterization and task state information transmission mapping

Abstract

The application discloses a brain network coupling analysis system and method based on structural connection topological characterization and task state information transmission mapping, and realizes the collaborative analysis of brain structural network topological features and functional information transmission through structural connection embedding and task state information transmission coupling modeling. The system integrates multi-modal neural image data, extracts the structural features of brain regions by using a random walk graph embedding method, quantifies the task state information transmission by combining an activity flow mapping technology, and finally establishes a structure-function mapping model. The method breaks through the limitation of the separation of structure and function in traditional brain network analysis, can stably represent the information transmission characteristics between brain regions, provides an effective tool for understanding the brain network mechanism in cognitive tasks, and has good universality and scalability.

Description

Technical Field

[0001] This invention relates to the fields of brain image data processing and brain network computational modeling, specifically to a brain network coupling analysis system and method based on structural connectivity topology representation and task-state information transfer mapping. Background Technology

[0002] Existing brain imaging analysis techniques often treat structural connectivity analysis and functional activation analysis as independent processes, making it difficult to characterize the intrinsic relationship between brain region structural topological features and task-state functional information transmission within a unified framework. Furthermore, existing methods generally suffer from unstable modeling, insufficient cross-subject generalization ability, and a lack of characterization of task-state information interaction processes when processing high-dimensional brain imaging data. To address these issues, this invention proposes a method and system for coupled modeling of structural connectivity embedding and task-state information transmission. This method enables collaborative modeling of structural connectivity topological features, functional connectivity weights, and task-state activation patterns within a unified framework, thereby achieving stable representation and prediction of information transmission characteristics between brain regions. Summary of the Invention

[0003] To address the technical problems mentioned above, this invention provides a brain network coupling analysis system based on structural connectivity topological representation and task-state information transfer mapping, comprising: The data acquisition and preprocessing module is used to acquire structural magnetic resonance imaging data, diffusion-weighted imaging data, resting-state functional magnetic resonance imaging data, and task-state functional magnetic resonance imaging data of the subjects, and to preprocess the acquired data. The structural connectivity construction module is used to construct an individual structural connectivity network with brain regions as nodes and inter-brain connectivity weights as edge weights based on preprocessed diffusion-weighted imaging data. The structural embedding module is used to perform low-dimensional vectorization representation of the structural connectivity network and generate a corresponding structural embedding vector for each brain region. The functional connectivity modeling module is used to construct voxel-level resting-state functional connectivity weights based on preprocessed resting-state functional magnetic resonance imaging data to form a voxel-level resting-state functional connectivity matrix. The activity flow mapping module is used to map the task-state activation pattern of the source brain region to the target brain region according to the voxel-level resting-state functional connectivity matrix, so as to obtain the predicted activation pattern of the target brain region. The information transmission calculation module is used to quantify the amount of information transmitted between the source brain region and the target brain region during the task execution process based on the predicted activation mode and the actual activation mode of the target brain region. The mapping modeling module is used to associate the structure embedding vector with the information transmission amount and establish a mapping model between the structure embedding features and the information transmission amount.

[0004] Preferably, the structure embedding module is specifically used to: represent the structure connection network as a weighted graph structure, perform traversal modeling on the weighted graph based on a random walk graph embedding method, and generate a random walk sequence for each node; The random walk sequence is input into the vector mapping model, and the contextual relationships of the nodes are mapped to a unified embedding space through the projection function, thereby generating a corresponding structural embedding vector for each node.

[0005] Preferably, the structure embedding module introduces a first control parameter during the random walk. p Second control parameter q First control parameter p The second control parameter is used to control the probability of returning to the previous node. q This is used to control the probability of exploring new nodes, and different biased embedding strategies can be achieved by adjusting parameters.

[0006] Preferably, the functional connection modeling module is specifically used for: Principal component analysis was performed on the resting-state functional magnetic resonance time series for dimensionality reduction. A multiple linear regression model is established based on the dimensionality-reduced low-dimensional time series to estimate the voxel-to-voxel resting-state functional connectivity weights between any two brain regions, thus forming a voxel-level resting-state functional connectivity matrix.

[0007] Preferably, the information transmission calculation module is specifically used for: Within the cross-validation framework, the predicted activation patterns of the target brain region were obtained in each trial. Calculate the similarity between the predicted activation pattern and the actual activation pattern of the target brain region under matching conditions, and the similarity between the predicted activation pattern and the actual activation pattern of the target brain region under non-matching conditions. The amount of information transmitted between brain regions is quantified by comparing the difference between two types of similarity.

[0008] Preferably, the mapping modeling module is specifically used for: By taking the amount of information transmitted between any pair of brain regions as the dependent variable and the parallel combination of the structural embedding vectors of the corresponding brain regions as the independent variable, a mapping model based on linear regression is constructed. A cross-validation strategy is used to train and evaluate the linear regression model to obtain the predicted information transmission volume corresponding to all connected edges. By calculating the correlation index between the predicted information delivery volume and the actual information delivery volume, the predictive effect of the structural embedding vector on the information delivery volume is quantitatively evaluated.

[0009] Preferably, the data acquisition and preprocessing module is specifically used for: Denoising, artifact correction, motion and eddy current correction, and bias field correction are performed on diffusion-weighted imaging data. Temporal correction, head motion correction, covariate removal, spatial registration, standardization, and frequency domain filtering were performed on the resting-state functional magnetic resonance imaging data. Temporal correction, head motion correction, spatial registration, and standardization were performed on the task-state functional magnetic resonance imaging data. Furthermore, task-time modeling and activation estimation processes were introduced to obtain task-related regression coefficients per voxel and per trial.

[0010] This invention also provides a brain network coupling analysis method based on structural connectivity topology representation and task-state information transfer mapping. The method is applied to the aforementioned system, and the steps include: Structural magnetic resonance imaging (fMRI), diffusion-weighted imaging (DWI), resting-state functional magnetic resonance imaging (fMRI), and task-based fMRI data were collected from the subjects, and the collected data were preprocessed. Based on the preprocessed diffusion-weighted imaging data, an individual structural connection network was constructed with brain regions as nodes and the connection weights between brain regions as edge weights. The structural connectivity network is represented by low-dimensional vectorization to generate a corresponding structural embedding vector for each brain region. Based on the preprocessed resting-state functional magnetic resonance imaging data, voxel-level resting-state functional connectivity weights are constructed to form a voxel-level resting-state functional connectivity matrix. Based on the voxel-level resting-state functional connectivity matrix, the task-state activation patterns of the source brain region are mapped to the target brain region to obtain the predicted activation patterns of the target brain region. Based on the predicted activation pattern and the actual activation pattern of the target brain region, the amount of information transferred between the source brain region and the target brain region during the task execution process is quantified. By associating structural embedding vectors with information transmission volume, a mapping model between structural embedding features and information transmission volume is established.

[0011] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) Achieving unified modeling of structural connection and task state functional information transmission: This invention introduces structural embedding vector and activity flow mapping mechanism to achieve collaborative modeling of structural topological features, functional connection weights and task state information transmission characteristics under the same technical framework, overcoming the problem of the separation between structural and functional analysis in the prior art.

[0012] (2) Improve the stability and generalization ability of high-dimensional brain imaging data modeling: By using structural embedding computation and functional connectivity dimensionality reduction modeling, the impact of high-dimensional data on model stability can be effectively reduced, so that the constructed information transmission model has good cross-subject applicability and engineering feasibility.

[0013] (3) It can characterize the information interaction process between brain regions under task state: Through activity flow mapping and information transmission quantity calculation, this invention can quantitatively describe the information interaction relationship between brain regions during task execution, breaking through the limitations of traditional analysis based only on resting state or activation intensity.

[0014] (4) It has good scalability and application flexibility: The system and method described in this invention do not depend on specific tasks or specific groups of people, and can be applied to different types of task-state functional magnetic resonance data scenarios. It is suitable for various structure-function coupling analysis needs and has good engineering promotion value. Attached Figure Description

[0015] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the brain network coupling analysis system based on structural connection topology representation and task state information transfer mapping of the present invention; Figure 2 This is a schematic diagram of the data processing process of the brain network coupling analysis system based on structural connection topology representation and task state information transfer mapping of the present invention; Figure 3 This is a diagram showing the results obtained from an embodiment of the brain network coupling analysis based on structural connection topology representation and task-state information transfer mapping of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] Example 1 like Figure 1As shown, this embodiment of the invention provides a coupling and decoupling computational system based on morphological-functional brain networks. The system runs on a computer device and includes: a data acquisition and preprocessing module; a structural connectivity construction module; a structural embedding module; a functional connectivity modeling module; an activity flow mapping module; an information transmission computation module; and a mapping modeling module.

[0020] The workflow of the system in this embodiment is as follows: Figure 2 As shown. Specifically, structural magnetic resonance imaging (fMRI) data, diffusion-weighted imaging (DWI) data, task-oriented functional magnetic resonance imaging (fMRI) data, and resting-state fMRI data are collected from the subjects, and preprocessing operations are performed on the acquired data. In this embodiment, the diffusion-weighted imaging data is processed using a preset diffusion imaging preprocessing and structural connectivity construction process. The process includes denoising and artifact correction of diffusion imaging data to reduce acquisition noise and eliminate ringing effects; subsequently, motion and eddy current correction and bias field correction are performed to reduce the impact of subject motion, gradient nonlinearity, and magnetic field inhomogeneity on the diffusion signal; after completing diffusion data quality correction, T1-weighted structural magnetic resonance images are registered to the diffusion imaging space, and tissue segmentation images suitable for anatomically constrained fiber tracing are generated; further, multi-tissue response functions of white matter, gray matter, and cerebrospinal fluid are estimated, and fiber orientation distribution is calculated based on the multi-tissue constrained spherical deconvolution method; on this basis, whole-brain white matter fiber tracing is performed to generate a set of fiber trajectories, and the fiber trajectories are quantitatively weighted and optimized to obtain biologically reasonable connection weights; finally, the weighted fiber trajectories are mapped to a preset brain region partition template to construct an individual structural connection network with brain regions as nodes and inter-brain connection weights as edge weights, providing input data for subsequent structural embedding and information transmission modeling.

[0021] In this embodiment, the resting-state functional magnetic resonance imaging (fMRI) data is first processed using a preset data preprocessing procedure. This preprocessing procedure includes: performing time-level correction on the acquired data to eliminate acquisition time differences between different scanning layers; subsequently correcting for head movements generated by the subject during the scanning process and aligning the fMRI images at each time point to the reference image with the least degree of motion abnormality in the time series; further removing the influence of non-neuronal signals, including head motion parameters, white matter signals, and cerebrospinal fluid; after covariate processing, jointly segmenting and spatially registering the fMRI images with the corresponding T1-weighted structural fMRI images, and standardizing them to a preset standard brain space template; then performing spatial smoothing on the standardized images to improve the signal-to-noise ratio and enhance population consistency; finally, performing frequency domain filtering on the fMRI time series to retain neural activity-related signals within a preset frequency band, thereby obtaining resting-state fMRI data for subsequent functional connectivity modeling.

[0022] In this embodiment, the processing flow of task-state functional magnetic resonance imaging (fMRI) data is basically the same as that of resting-state fMRI data, but the covariate removal step is not performed, and a task-time modeling and activation estimation process is further introduced. Specifically, according to the task experimental design, the time of each trial is marked, and the task time sequence is convolved with a preset blood oxygen dynamics response function to generate corresponding task regression variables. Subsequently, based on a voxel-level generalized linear model, the convolved task regression variables are fitted with preset covariates to obtain task-related regression coefficients per voxel and per trial. The regression coefficients are used to characterize the activation intensity of brain regions induced by the task and serve as input features for subsequent information transfer calculations. In this embodiment, each trial is modeled as an independent variable, rather than being merged according to task conditions, to obtain the activation parameter values ​​corresponding to each voxel under multiple trials; in the case of multiple task stages, only the activation parameters corresponding to the preset task stages are selected for subsequent analysis and processing.

[0023] In this embodiment, the structural embedding module is used to perform low-dimensional vectorization representation of the constructed structural connectivity network to extract structural communication features between brain regions. Specifically, the structural connectivity network is represented as a weighted graph structure, where nodes correspond to preset brain regions and edges correspond to the structural connection strength between brain regions. Based on this, the Node2Vec graph embedding method based on random walks is used to traverse and model the weighted graph. Multiple random walks are performed in the graph structure to obtain the contextual relationship information of each node in the network. The transition probability is determined based on the connection weights between nodes. The larger the connection weight, the higher the probability that the corresponding node will be visited during the random walk. By setting the step size and repetition count of the random walk, each node in the network generates a corresponding random walk sequence to describe the local and higher-order topological relationships of the node in the structural connectivity network.

[0024] Furthermore, the random walk sequence is input into a vector mapping model, and the projection function is used to... f The contextual relationships of nodes are mapped to a unified embedding space, thereby generating a corresponding structural embedding vector for each node. The vector mapping model aims to maximize the similarity between a node's embedding vector and the embedding vectors of its neighboring nodes. By modeling the co-occurrence relationships between nodes, it expresses the structural connection communication patterns. In this process, a first control parameter is introduced. p Second control parameter q To adjust the random walk strategy, where the first control parameter p The second control parameter is used to control the probability of returning to the previous node during a random walk. qUsed to control the probability of exploring new nodes during random walks; by adjusting the value of the parameter, different embedding strategies can be implemented, ranging from biased towards local structures to biased towards global structures. p The smaller the embedding, the more homogeneous it tends to be; q The smaller the embedding, the more structure-oriented it tends to be. The expression for the optimization objective is as follows: (1) in, Represents a node u All neighboring nodes, f This represents a projection operation on the node vector. V Represents the set of all nodes. ni Represents a node u The i One neighboring node, v This represents any node in the graph.

[0025] In this embodiment, the number of random walk steps is set to a preset value. Multiple random walks are performed for each structural connectivity network to enhance the stability of the embedding results. The dimension of the embedding vector is set to a predetermined length, and a random walk strategy primarily focused on preserving local structure is used to generate the structural embedding results. The final obtained structural embedding vector is used to characterize the communication location features of each brain region in the structural connectivity network and serves as the input parameter for subsequent information transfer calculations and mapping modeling modules.

[0026] In this embodiment, the functional connectivity modeling module is used to construct voxel-level resting-state functional connectivity weights to provide a foundation for subsequent activity flow mapping and information transfer calculations. Since the number of voxels in functional magnetic resonance imaging (fMRI) data is significantly greater than the number of time points, directly establishing a regression model based on voxel time series can easily lead to model instability. Therefore, before constructing voxel-level functional connectivity, the resting-state fMRI time series is first subjected to dimensionality reduction processing. Specifically, principal component analysis is performed on the voxel time series to extract low-dimensional representations that can characterize the main temporal variation features, thereby reducing the number of independent variables and suppressing redundant noise. After dimensionality reduction, a multiple linear regression model is established based on the low-dimensional time series to estimate the voxel-to-voxel resting-state functional connectivity weights between any two brain regions. These functional connectivity weights characterize the linear predictive relationship between voxel activity in the source brain region and voxel activity in the target brain region, and constitute an intermediate voxel-level resting-state functional connectivity matrix.

[0027] In this embodiment, the information mapping module is based on voxel-level resting-state functional connectivity, mapping the task-state activation pattern of the source brain region to the target brain region, thereby obtaining the predicted activation pattern of the target brain region. Specifically, for any pair of brain regions... A With brain regions B brain regions AThe task-state activation parameters of endothelial cells in a single trial constitute the source activation vector, based on brain regions. A With brain regions B The resting-state functional connectivity weights at the voxel level constitute a functional connectivity matrix, and the target brain region is generated through matrix multiplication. B The predicted activation pattern is expressed as follows: (2) in, Indicates the target brain region B In the k The predicted activation pattern vector in each trial, the vector being derived from the target brain region... n The predicted activation values ​​of individual elements constitute the composition; Indicates the source brain region A In the k The actual activation pattern vector in each trial, the vector being derived from the source brain region m The activation values ​​of individual elements constitute the composition; Indicates brain regions A With brain regions B The voxel-level resting-state functional connectivity weight matrix between the voxels has dimensions m×n; the operator · represents matrix multiplication. Since each voxel has corresponding activation parameters in multiple trials, the target brain region can obtain multiple sets of predicted activation patterns throughout the task, which are used for subsequent information transfer calculations.

[0028] In this embodiment, the information transfer calculation module is used to quantify the amount of information transferred between the source brain region and the target brain region during task execution. The amount of information transferred is obtained by comparing the similarity between the predicted activation pattern and the actual activation pattern, specifically including the following process: First, under a cross-validation framework, based on the task-state activation of the source brain region, the predicted activation pattern of the target brain region in each trial is obtained through activity flow mapping; then, the similarity between the predicted activation pattern and the actual activation pattern of the target brain region under matching conditions, and the similarity between the predicted activation pattern and the actual activation pattern of the target brain region under non-matching conditions are calculated respectively; finally, the information transfer capability between brain regions is quantified by comparing the difference between the two types of similarity. (Source brain region) A With the target brain region B The overall information transfer volume between them is defined as follows: (3) in, This represents the average similarity value under conditional matching. This represents the average similarity value under conditions of mismatch. The similarity is obtained through relevance calculation and averaged over multiple trials. Specifically, the similarity under matching conditions is defined as: (4) Similarity under non-matching conditions is defined as: (5) in, K This indicates the number of trials for the corresponding stage in the task. corr(⋅) This represents the similarity index obtained by calculating the correlation between two vectors and then performing the Fisher-Z transform. This indicates the target brain region obtained based on source brain region activation prediction. k Activation mode with repeated trials; This represents the conditional prototype obtained by averaging the actual activation patterns of the target brain region under the same task conditions after removing the current trial. This represents the conditional prototype obtained under conditions of mismatch between task and target brain regions. Through the above calculation process, the comprehensive information transfer between the source and target brain regions throughout the entire task phase can be obtained, which is used for subsequent mapping modeling and analysis.

[0029] In this embodiment, the mapping modeling module is used to establish a mapping relationship between structural embedding vectors and information transmission between brain regions, so as to characterize the predictive ability of structural connectivity topological features on functional information transmission. Specifically, any pair of brain regions... A With brain regions B The amount of information transferred between them is used as the dependent variable, and the corresponding brain regions are used as the dependent variable. A With brain regions B We use parallel combinations of structural embedding vectors as independent variables to construct a linear regression-based mapping model to describe the mapping relationship between structural embedding features and information transmission volume.

[0030] To improve the model's generalization ability and stability, a cross-validation strategy is employed for training and evaluation during the mapping modeling process. Specifically, all brain region connectivity edges are randomly divided into multiple subsets. Each time, a subset of these subsets is selected as the test set, while the remaining edges serve as the training set. Parameters of the linear regression model are estimated based on the training set data, and the trained model is used to predict the information transfer volume of the corresponding connectivity edges in the test set. By rotating different subsets as the test set, every connectivity edge participates in the model's training and testing process, thereby obtaining the predicted information transfer volume for all connectivity edges.

[0031] Furthermore, the prediction results obtained from each cross-validation process are summarized, and the cross-validation process is repeated multiple times to reduce the impact of random partitioning on the results and enhance the robustness of the prediction results. Finally, by calculating the correlation index between the predicted information transmission volume and the actual information transmission volume, the prediction effect of the structural embedding vector on the information transmission volume is quantitatively evaluated, thereby completing the mapping model between structural connectivity embedding features and functional information transmission characteristics.

[0032] Example 2 In this embodiment, a batch of publicly available multimodal brain imaging data from normal individuals is used to fully apply the structural connectivity embedding and task-state information transfer mapping system described in this invention. The data originates from the publicly available UCLA brain imaging database, collected from 67 normal subjects. The data types include structural magnetic resonance imaging (MRI) data, diffusion-weighted imaging (DWI) data, and task-state functional magnetic resonance imaging (fMRI) data collected under the episodic memory task paradigm. This comprehensively covers the input requirements of the system in terms of data acquisition, structural modeling, functional modeling, and information transfer analysis.

[0033] During the data processing phase, the participants' structural magnetic resonance imaging (SMRI), diffusion-weighted imaging (DWI), and functional magnetic resonance imaging (fMRI) data were first input into the data acquisition module and the data preprocessing module. Noise reduction, correction, spatial registration, and standardization were performed on the different modalities of data to eliminate systematic errors during acquisition and ensure consistency across different modalities within the same spatial reference frame. For fMRI data, based on its acquisition state, it was categorized into resting-state data and task-state data. Resting-state fMRI data was used for subsequent functional connectivity modeling, while task-state fMRI data was used to describe the activation states of brain regions during the participants' episodic memory task.

[0034] Subsequently, in the structural connectivity construction module, whole-brain white matter fiber pathways are reconstructed from diffusion-weighted imaging data using the MRTrix3 toolkit. The reconstructed fiber connectivity results are then mapped onto a Glasser360 brain template to construct a brain region structural connectivity matrix. This structural connectivity matrix uses brain regions as nodes and the fiber connectivity strength between brain regions as edges, representing the anatomical structural connectivity relationships between different brain regions. This structural connectivity matrix serves as the input to the subsequent structural embedding module, used to extract the topological features of brain regions within the overall structural network.

[0035] In the structural embedding module, graph embedding calculations are performed on the structural connectivity matrix. Using a Node2Vec random walk approach, the high-dimensional structural connectivity network is mapped to a low-dimensional embedding space, generating a corresponding structural embedding vector for each brain region. These structural embedding vectors comprehensively reflect the local connectivity relationships and higher-order topological positions of brain regions within the structural network, characterizing the potential structural communication characteristics between brain regions. By using standardized embedding parameter settings, the comparability of structural embedding results among different subjects is ensured.

[0036] Meanwhile, in the functional connectivity modeling module, voxel-level functional connectivity weights are constructed based on resting-state functional magnetic resonance (fMRI) data. Specifically, by performing dimensionality reduction on the resting-state fMRI time series and establishing a multivariate regression model, a functional connectivity weight matrix between voxels is obtained, which is used to characterize the predictive relationships between the activities of different voxels. These voxel-level functional connectivity weights provide the foundation for subsequent activity flow mapping.

[0037] In the activity flow mapping module, based on the aforementioned functional connectivity weights, the task-state activation patterns of the source brain regions during the episodic memory task are mapped to the target brain regions to predict the activation patterns of the target brain regions in each trial. By performing activation pattern mapping in the spatial dimension, the transmission of task-related information between brain regions is simulated. In the information transmission calculation module, by comparing the similarity difference between the predicted activation patterns and the actual activation patterns of the target brain regions, the amount of information transmitted between brain regions during the execution of the episodic memory task is calculated, thereby constructing a brain region information transmission matrix that reflects the characteristics of task-state information interaction.

[0038] Furthermore, in the mapping modeling module, the calculated information transfer volume is used as the dependent variable, and the structural embedding vector of the corresponding brain region is used as the independent variable to establish a mapping model between structural embedding features and information transfer volume. The mapping model is trained and evaluated using cross-validation to predict information transfer patterns during episodic memory tasks based on structural connectivity embedding features. This embodiment verifies that the system of the present invention can stably achieve collaborative modeling between structural connectivity topological features and task-state functional information transfer in normal human episodic memory task scenarios, demonstrating good versatility, scalability, and engineering application value.

[0039] like Figure 3 As shown, the accuracy of the embedded module in predicting the amount of task-state functional information transmitted is negatively correlated with the BPRS negative scale.

[0040] Example 3 This embodiment also provides a brain network coupling analysis method based on structural connectivity topology representation and task-state information transfer mapping, the steps of which include: S1. Collect structural magnetic resonance imaging data, diffusion-weighted imaging data, resting-state functional magnetic resonance imaging data, and task-state functional magnetic resonance imaging data of the subjects, and preprocess the collected data. S2. Based on the preprocessed diffusion-weighted imaging data, construct an individual structural connection network with brain regions as nodes and the connection weights between brain regions as edge weights. S3. Perform low-dimensional vectorization representation of the structural connectivity network to generate a corresponding structural embedding vector for each brain region. S4. Based on the preprocessed resting-state functional magnetic resonance imaging data, construct voxel-level resting-state functional connectivity weights to form a voxel-level resting-state functional connectivity matrix. S5. Based on the voxel-level resting-state functional connectivity matrix, the task-state activation pattern of the source brain region is mapped to the target brain region to obtain the predicted activation pattern of the target brain region. S6. Based on the predicted activation pattern and the actual activation pattern of the target brain region, quantify the amount of information transferred between the source brain region and the target brain region during the task execution process. S7. Associate the structural embedding vector with the information transmission amount to establish a mapping model between structural embedding features and information transmission amount.

[0041] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A brain network coupling analysis system based on structural connectivity topological representation and task-state information transfer mapping, characterized in that, include: The data acquisition and preprocessing module is used to acquire structural magnetic resonance imaging data, diffusion-weighted imaging data, resting-state functional magnetic resonance imaging data, and task-state functional magnetic resonance imaging data of the subjects, and to preprocess the acquired data. The structural connectivity construction module is used to construct an individual structural connectivity network with brain regions as nodes and inter-brain connectivity weights as edge weights based on preprocessed diffusion-weighted imaging data. The structural embedding module is used to perform low-dimensional vectorization representation of the structural connectivity network and generate a corresponding structural embedding vector for each brain region. The functional connectivity modeling module is used to construct voxel-level resting-state functional connectivity weights based on preprocessed resting-state functional magnetic resonance imaging data, so as to form a voxel-level resting-state functional connectivity matrix. The activity flow mapping module is used to map the task-state activation pattern of the source brain region to the target brain region according to the voxel-level resting-state functional connectivity matrix, so as to obtain the predicted activation pattern of the target brain region. The information transmission calculation module is used to quantify the amount of information transmitted between the source brain region and the target brain region during the task execution process based on the predicted activation mode and the actual activation mode of the target brain region. The mapping modeling module is used to associate the structure embedding vector with the information transmission amount and establish a mapping model between the structure embedding features and the information transmission amount.

2. The brain network coupling analysis system based on structural connectivity topology representation and task-state information transfer mapping according to claim 1, characterized in that, The structure embedding module is specifically used to: represent the structure connection network as a weighted graph structure, perform traversal modeling of the weighted graph based on the graph embedding method of random walk, and generate random walk sequences for each node; The random walk sequence is input into the vector mapping model, and the contextual relationships of the nodes are mapped to a unified embedding space through the projection function, thereby generating a corresponding structural embedding vector for each node.

3. The brain network coupling analysis system based on structural connectivity topology representation and task-state information transfer mapping according to claim 2, characterized in that, The structure embedding module introduces the first control parameter during the random walk. p Second control parameter q First control parameter p The second control parameter is used to control the probability of returning to the previous node. q This is used to control the probability of exploring new nodes, and different biased embedding strategies can be achieved by adjusting parameters.

4. The brain network coupling analysis system based on structural connectivity topology representation and task-state information transfer mapping according to claim 1, characterized in that, The functional connection modeling module is specifically used for: Principal component analysis was performed on the resting-state functional magnetic resonance time series for dimensionality reduction. A multiple linear regression model is established based on the dimensionality-reduced low-dimensional time series to estimate the voxel-to-voxel resting-state functional connectivity weights between any two brain regions, thus forming a voxel-level resting-state functional connectivity matrix.

5. The brain network coupling analysis system based on structural connectivity topology representation and task-state information transfer mapping according to claim 1, characterized in that, The information transmission and calculation module is specifically used for: Within the cross-validation framework, the predicted activation patterns of the target brain region were obtained in each trial. Calculate the similarity between the predicted activation pattern and the actual activation pattern of the target brain region under matching conditions, and the similarity between the predicted activation pattern and the actual activation pattern of the target brain region under non-matching conditions. The amount of information transmitted between brain regions is quantified by comparing the difference between two types of similarity.

6. The brain network coupling analysis system based on structural connectivity topology representation and task-state information transfer mapping according to claim 1, characterized in that, The mapping modeling module is specifically used for: By taking the amount of information transmitted between any pair of brain regions as the dependent variable and the parallel combination of the structural embedding vectors of the corresponding brain regions as the independent variable, a mapping model based on linear regression is constructed. A cross-validation strategy is used to train and evaluate the linear regression model to obtain the predicted information transmission volume corresponding to all connected edges. By calculating the correlation index between the predicted information delivery volume and the actual information delivery volume, the predictive effect of the structural embedding vector on the information delivery volume is quantitatively evaluated.

7. The brain network coupling analysis system based on structural connectivity topology representation and task-state information transfer mapping according to claim 1, characterized in that, The data acquisition and preprocessing module is specifically used for: Denoising, artifact correction, motion and eddy current correction, and bias field correction are performed on diffusion-weighted imaging data. Temporal correction, head motion correction, covariate removal, spatial registration, standardization, and frequency domain filtering were performed on the resting-state functional magnetic resonance imaging data. Temporal correction, head motion correction, spatial registration, and standardization were performed on the task-state functional magnetic resonance imaging data. Furthermore, task-time modeling and activation estimation processes were introduced to obtain task-related regression coefficients per voxel and per trial.

8. A brain network coupling analysis method based on structural connectivity topological representation and task-state information transfer mapping, wherein the method is applied to the system described in any one of claims 1-7, characterized in that the steps... include: Structural magnetic resonance imaging (fMRI), diffusion-weighted imaging (DWI), resting-state functional magnetic resonance imaging (fMRI), and task-based fMRI data were collected from the subjects, and the collected data were preprocessed. Based on the preprocessed diffusion-weighted imaging data, an individual structural connection network was constructed with brain regions as nodes and the connection weights between brain regions as edge weights. The structural connectivity network is represented by low-dimensional vectorization to generate a corresponding structural embedding vector for each brain region. Based on the preprocessed resting-state functional magnetic resonance imaging data, voxel-level resting-state functional connectivity weights are constructed to form a voxel-level resting-state functional connectivity matrix. Based on the voxel-level resting-state functional connectivity matrix, the task-state activation patterns of the source brain region are mapped to the target brain region to obtain the predicted activation patterns of the target brain region. Based on the predicted activation pattern and the actual activation pattern of the target brain region, the amount of information transferred between the source brain region and the target brain region during the task execution process is quantified. By associating structural embedding vectors with information transmission volume, a mapping model between structural embedding features and information transmission volume is established.

Images