Double-current space-time brain network analysis method with embedded group prior

Abstract

The invention relates to the technical field of brain network construction, in particular to a group prior embedded double-flow space-time brain network analysis method, which comprises the following steps of: S1, preprocessing a brain function image; s2, dividing the brain into a plurality of brain regions, and extracting an average time sequence; s3, calculating edges of a connection weight construction brain map, and outputting a symmetric correlation matrix # imgabs0 #; s4, defining a graph isomorphic network under spatial features, taking the correlation matrix R as the input of the graph isomorphic network, and outputting to obtain a tag Z1 related to the spatial features; S5, collecting BOLD signals of brain function images, and inputting the BOLD signals into a # imgabs1 # model to a # imgabs2 # model to obtain a tag Z2 related to the time features; s6, performing dimensionality reduction and aggregation on the spatial feature tag Z1 and the time feature tag Z2 to obtain a same-dimensional tag Z; and S7, establishing a group-based attraction graph # imgabs3 # by using the same-dimensional label Z to realize classification and identification. According to the method, the spatial-temporal feature tags are utilized to construct the group graph Gp, and node feature updating is matched to obtain new tags embedded into group priori, so that the classification and recognition accuracy of the brain function image can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of brain network construction, and in particular to a dual-stream spatiotemporal brain network analysis method embedded with group priors. Background Art

[0002] Resting-State Functional Magnetic Resonance Imaging (rs-fMRI) is a non-invasive neuroimaging technique used to measure the functional connectivity of the brain when not performing a specific task. It uses fluctuations in blood oxygenation-dependent signals to generate functional maps of brain activity, which are used to infer functional interactions between different brain regions. It has been widely used in early research on autism. Because autism is associated with impaired neural connectivity between brain regions, constructing high-quality brain networks has become a key tool in assisting clinicians in clinical autism research.

[0003] The brain regions of interest (ROIs) obtained after resting-state magnetic resonance imaging preprocessing are represented by blood oxygen level-dependent signals (BOLDs), which consist of time series. Mamba, a novel sequence modeling architecture based on the state-space model (SSM), excels in modeling long sequences and can efficiently process large amounts of data. It also addresses the difficulties of traditional recurrent neural networks (RNNs) in capturing long-term dependencies and the quadratic computational cost caused by the Transformer's reliance on attention mechanisms, providing a more efficient sequence modeling approach with linear complexity. The Graph Isomorphism Network (GIN) is a graph neural network primarily used to improve graph representation capabilities. Its core concept is to mimic the Weisfeiler-Lehman graph isomorphism test, thereby improving the GNN's ability to distinguish between different graph structures. GIN, through methods such as MLP transformation and neighborhood feature aggregation, can more effectively capture spatial features than traditional GNNs.

[0004] Most existing diagnostic models for brain disorders focus solely on single spatial features in the BOLD signal, failing to capture its deeper, temporal dynamics. Furthermore, they suffer from low processing efficiency when dealing with large samples, limiting their widespread clinical application. Furthermore, most existing methods for constructing brain functional networks map individual subjects independently, ignoring potential shared structural and functional connectivity patterns across subjects. This can easily lead to overfitting of diagnostic models to noise or individual differences, resulting in poor generalization. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a dual-stream spatiotemporal brain network analysis method embedded with group priors, which obtains labels containing spatiotemporal features by establishing a spatiotemporal dual-stream model and constructs a group graph G p Using the group attraction graph to update node features to obtain new labels embedded with group priors can effectively improve the classification and recognition accuracy of brain functional images.

[0006] The present invention is achieved through the following technical solutions: A two-stream spatiotemporal brain network analysis method with embedded population priors is provided, comprising the following steps: S1. Preprocessing of acquired brain functional images; S2. Based on the automatic anatomical labeling atlas, the brain is divided into several brain regions, and the average time series of each brain region is extracted and expressed as ,in Indicates the Time series of brain regions, represents the length of the time series, Indicates the number of brain regions; S3. The connection weights are calculated by Pearson correlation coefficient to construct the edges of the brain map as follows: For brain areas and , and its Pearson correlation coefficient: ; in: It is the brain area The time series is z-score standardized to make the mean 0 and the standard deviation 1: ; in: For brain area At the time point The signal value of and brain regions The mean and standard deviation of , a total of Secondary correlation coefficient, output symmetrical correlation matrix ; S4. Define a graph isomorphism network under spatial features, use the correlation matrix R output in step S3 and the original data X as inputs to the graph isomorphism network, and output the label Z1 that has learned the spatial features; S5. Collect BOLD signal input of pre-processed brain functional images Give Model, obtains the label Z2 that learns the time feature; S6, respectively perform dimensionality reduction aggregation on the label Z1 of the learned spatial features obtained in step S4 and the label Z2 of the learned temporal features obtained in step S5 to obtain a label Z that integrates the spatial features and temporal features in the original data; S7. Build a group diagram , input the fused label Z as a node into the group graph , the group attraction graph is used to update the features of node Z to obtain a new label Z_cat embedded with group priors, thereby realizing the classification and recognition of brain functional images.

[0007] Further, in step S5, the selective mechanism mapping is used to obtain 、 and , with the state space model SSM formula, using parameters Through zero-order hold and After discretization, we get and input Multiply and discretize to get and original state Multiply and add these two terms to get the new state , and finally the new state and Multiply to get the output : ; ; in: , , is the identity matrix, and the discretized SSM is equivalent to the following convolution: ; .

[0008] Furthermore, when calculating the loss in steps S5 and S6, a regularizer is defined: ; During training, the regularization term is added to the overall loss function: ; ; in: is the standard cross entropy loss, is the number of categories; is the real label, represented by one-hot encoding, where only the position corresponding to the target category is 1 and the other positions are 0; is the label probability predicted by the model, i.e. The output value of It is a learnable parameter used to control the influence of the regularization term; is the regularized loss defined by the regularizer, which should be in vector form ,in It is thermonuclear rights.

[0009] A semi-supervised regularizer with thermal kernel weights is introduced into the standard cross entropy loss to optimize the loss function.

[0010] Furthermore, in step S7, the label Z obtained according to the spatiotemporal dual-stream model is embedded into the patient node, and each patient in the healthy group is regarded as a positive sample, so that they are clustered and pulled together in the feature space, while each patient in the patient group is regarded as a negative sample, so that they are sparse and pushed away in the feature space; each patient in the patient group is regarded as a positive sample, so that they are clustered and pulled together in the feature space, while each patient in the healthy group is regarded as a negative sample, so that they are sparse and pushed away in the feature space, and finally a group graph G consisting of two groups of clusters is obtained. p .

[0011] Furthermore, in step S7, the group graph The edge is constructed using the Top-k neighborhood method, which selects the k neighboring nodes with the highest similarity metrics of the input node to connect the bidirectional Top-k edges. The similarity metric can be calculated in the following three ways: (1) Dot product similarity: ; Its meaning is to measure the projection similarity of two vectors, that is, how much they are in the same direction; (2) Cosine similarity: ; Its meaning is to measure the similarity of the angle between two vectors, which has nothing to do with the modulus; (3) Gaussian kernel similarity: ; Its meaning is the heat kernel constructed based on the Euclidean distance. The closer the distance, the closer the similarity is to 1.

[0012] Furthermore, in step S4, the graph isomorphic network is: ; ; in: is a brain network constructed using Pearson correlation. For the Layer Node Initialize the eigenvector of is a fully connected layer; For learnable parameters or set a fixed scalar; Is with the node A set of adjacent nodes, It is an aggregation node Neighbor node information.

[0013] Furthermore, in step S1, the DPARSF toolbox of Matlab was used to preprocess the brain functional map imaging to remove the front At each time point, the image is corrected for head motion artifacts and time layers, and the effects of ventricular and white matter signals and head motion high-order effects are removed. The corrected image is registered to the standard space. Temporal bandpass filtering to reduce the effects of heartbeat and respiration.

[0014] Beneficial effects of the present invention: This paper introduces two encoders, Mamba and GIN, combining the former's ability to efficiently capture temporal features in long-sequence modeling with the latter's ability to extract spatial correlation features from graph structures, significantly improving computational efficiency while ensuring modeling performance. By using a spatiotemporal dual-stream model with different feature encoders to capture temporal and spatial features in parallel, a label that integrates spatiotemporal features is obtained. This label is then used to construct a population graph consisting of two clusters, enabling the classification and recognition of diseased and healthy images in brain functional images. Continuous training and replacement can effectively improve the accuracy of recognition and classification.

[0015] This invention also innovatively incorporates population prior information. By constructing a population graph, different subject samples are linked based on similarity metrics, enabling cross-subject information fusion and structural induction, thereby embedding population-level structural priors during model training. This design effectively improves the stability and robustness of the model, particularly when the number of samples is limited or the data quality is uneven. It significantly enhances the model's generalization and practical application potential with multi-site and multimodal data.

[0016] By introducing the Mamba encoder, the two major advantages of GIN, which can effectively capture spatial features, and Mamba, which is suitable for long sequence modeling and can efficiently process large-scale data, are combined, and group priors are embedded. Unlike existing methods that mostly only build independent brain functional networks for a single subject, relevant connections are established between subjects based on similarity, achieving recognition and classification more effectively and accurately.

[0017] The present invention proposes a dual-stream spatiotemporal brain network analysis method embedded with group priors, which can use the dot product similarity between subject labels as a similarity measure suitable for group-based classification. Compared with the existing brain disease image recognition in brain functional images, which predicts the labels of individual subjects without identifying the information of adjacent subjects in the subject group, it can achieve more accurate classification and recognition, and effectively improve accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Flowchart of the present invention.

[0019] Figure 2 This is a group diagram in the present invention.

[0020] Figure 3 This is a table of model evaluation indicators on the ABIDE dataset in the present invention.

[0021] Figure 4 This is a table of model evaluation indicators on the MDD dataset in the present invention. DETAILED DESCRIPTION

[0022] In order to clearly illustrate the technical features of this solution, this solution is described below through specific implementation methods.

[0023] In this embodiment, a public Autism Brain Imaging Data Exchange (ABIDE) dataset is used. The dataset contains 184 subjects, including 79 autistic patients and 105 normal subjects.

[0024] like Figure 1 As shown, a two-stream spatiotemporal brain network analysis method embedded with group priors includes the following steps: S1. The brain functional image acquired by the functional magnetic resonance imaging device was preprocessed using the DPARSF toolbox of Matlab, including format reading and conversion. In order to keep the time series stable, the first 10 time points of each subject's magnetic resonance image were removed; the image was then corrected for head motion artifacts and time layers; the effects of ventricular, white matter signals and head motion high-order effects were removed; the corrected image was aligned to the standard space; the subject image was used Temporal bandpass filtering to reduce the effects of heartbeat and respiration.

[0025] S2. Select a standardized brain partitioning template. In this example, the brain is divided into 116 brain regions based on the AAL template, including 90 cerebral regions and 16 cerebellar regions. The 116 brain regions represent 116 nodes of the brain functional connectivity network.

[0026] The time series average of all voxels in each brain region in the magnetic resonance image is calculated, and the time series corresponding to different standard partitions are extracted, and the extracted time series are expressed as ,in Indicates the The time series of nodes, represents the number of times in the time series, Indicates the number of nodes.

[0027] S3. The connection weights are calculated by Pearson correlation coefficient to construct the edges of the brain map as follows: For brain areas and , and its Pearson correlation coefficient: ; in It is the brain area The time series is z-score standardized to make the mean 0 and the standard deviation 1: ; in For brain area At the time point The signal value of and brain regions The mean and standard deviation of . , a total of Secondary correlation coefficient, output symmetrical correlation matrix .

[0028] S4. Define a graph isomorphism network under spatial features, use the correlation matrix R output in step S3 as the input of the graph isomorphism network, and output the label Z1 about the spatial features: The spatial feature learning module is as follows: ; ; in: Represents the input BOLD signal, which captures the spatial characteristics between brain regions through GIN, making it convenient for subsequent aggregation of brain regions with similar functions.

[0029] S5. Collect BOLD signal input of pre-processed brain functional images Give Model, obtain the label Z2 about the time feature; The temporal feature learning module is as follows: ; ; in: , , is the identity matrix.

[0030] S6, respectively perform dimensionality reduction aggregation on the spatial feature label Z1 obtained in step S4 and the temporal feature label Z2 obtained in step S5 to obtain a same-dimensional label Z; S7. Use the fusion of same-dimensional labels Z to establish a group-based attraction graph , using group attraction graph to realize the classification and recognition of brain functional images.

[0031] The edges of the group graph G_p are constructed using the Top-k neighborhood method, which selects the k neighbor nodes with the highest similarity measure of the input node to connect the bidirectional Top-k edges.

[0032] The similarity measure can be calculated in one of the following three ways: (1) Dot product similarity: ; Its meaning is to measure the projection similarity of two vectors, that is, how much they are in the same direction; (2) Cosine similarity: ; Its meaning is to measure the similarity of the angle between two vectors, which has nothing to do with the modulus; (3) Gaussian kernel similarity: ; Its meaning is the heat kernel constructed based on the Euclidean distance. The closer the distance, the closer the similarity is to 1.

[0033] In this embodiment, the dot product similarity is selected as the calculation method of the similarity metric.

[0034] In steps S4 and S5, the regularization term is added to the total loss function during training: ; ; in: is the cross entropy loss, is the number of categories; is the real label, represented by one-hot encoding, where only the position corresponding to the target category is 1 and the other positions are 0; is the predicted probability of the model, i.e. The output value of It is a learnable parameter used to control the influence of the regularization term; is the regularized loss defined by the regularizer, which should be in vector form ,in It is thermonuclear rights.

[0035] The learning rate is set to 0.0001 because the BOLD signal value is relatively small. The batch size is set to 16. The number of subjects is small. A suitable batch size and a small learning rate can ensure the stability of training.

[0036] When the training reaches the last epoch, the training is stopped and the test accuracy and label of this round are output. The labels of all patient nodes in the group graph are embedded in turn. Each patient in the healthy group is regarded as a positive sample to make them gather and pull together in the feature space, while each patient in the patient group is regarded as a negative sample to make them sparse and push away in the feature space; each patient in the patient group is regarded as a positive sample to make them gather and pull together in the feature space, while each patient in the healthy group is regarded as a negative sample to make them sparse and push away in the feature space, and finally a group graph is obtained. .

[0037] To verify the effectiveness of the model proposed in this paper, the performance indicators of the model constructed by a two-stream spatiotemporal brain network analysis method with embedded population priors were tested and evaluated on two public neuroimaging classification datasets, ABIDE and MDD. Figure 3 and Figure 4 As shown in the data, on the ABIDE dataset, the model of the present invention achieved a classification accuracy (ACC) of 70.64% and 75.00% on the NYU116 and Leuven subsets, respectively, and an AUC of 70.17% and 75.42%, respectively. The balanced accuracy (BAC) also exceeded 69%, and the F1-score was also stable, indicating that the model has good recognition performance for autism samples.

[0038] On the MDD dataset, this model showed good classification accuracy and generalization ability in data from multiple sites. For example, on the SITE9 dataset, the model achieved an AUC of 85.39% and a BAC of 78.99%, with an F1-score of up to 79.47%, indicating that the model also has good discriminative ability in identifying depression samples.

[0039] ABIDE and MDD are datasets from two different fields. The model invented in this paper achieved good performance on both datasets, which shows that the model structure design has good versatility and adaptability, and has significant practical value in actual clinical prediction tasks.

[0040] Of course, the above description is not limited to the above examples. Technical features not described in the present invention can be achieved by or by adopting existing technologies, which will not be described here. The above embodiments and drawings are only used to illustrate the technical solutions of the present invention and are not limitations of the present invention. The present invention is described in detail with reference to the preferred implementation methods. Ordinary technicians in this field should understand that changes, modifications, additions or substitutions made by ordinary technicians in this technical field within the essential scope of the present invention do not depart from the purpose of the present invention and should also fall within the scope of protection of the claims of the present invention.

Claims

1. A two-stream spatiotemporal brain network analysis method embedded with group priors, characterized by: The following steps are involved: S1. Preprocessing of acquired brain functional images; S2. Based on the automatic anatomical labeling atlas, the brain is divided into several brain regions, and the average time series of each brain region is extracted and expressed as ,in Indicates the Time series of brain regions, represents the length of the time series, Indicates the number of brain regions; S3. The connection weights are calculated by Pearson correlation coefficient to construct the edges of the brain map as follows: For brain areas and , and its Pearson correlation coefficient: ； in: It is the brain area The time series is z-score standardized to make the mean 0 and the standard deviation 1: ； in: For brain area At the time point The signal value of and brain regions The mean and standard deviation of , a total of Secondary correlation coefficient, output symmetrical correlation matrix ; S4. Define a graph isomorphism network under spatial features. Use the correlation matrix R output in step S3 and the original data X as the input of the graph isomorphism network, and output the label Z1 that has learned the spatial features: S5. Collect BOLD signal input of pre-processed brain functional images Give Model, obtains the label Z2 that learns the time feature; S6, respectively perform dimensionality reduction aggregation on the label Z1 of the learned spatial features obtained in step S4 and the label Z2 of the learned temporal features obtained in step S5 to obtain a label Z that integrates the spatial features and temporal features in the original data; S7. Build a group diagram , input the fused label Z as a node into the group graph , the group attraction graph is used to update the features of node Z to obtain a new label Z_cat embedded with group priors, thereby realizing the classification and recognition of brain functional images.

2. The dual-stream spatiotemporal brain network analysis method embedded with population priors according to claim 1 is characterized by: In step S5, the selective mechanism mapping is used to obtain 、 and , with the state space model SSM formula, using parameters Through zero-order hold and After discretization, we get and input Multiply and discretize to get and original state Multiply and add these two terms to get the new state , and finally the new state and Multiply to get the output : ； ； in: , , is the identity matrix, and the discretized SSM is equivalent to the following convolution: ； 。 3. The dual-stream spatiotemporal brain network analysis method embedded with population priors according to claim 1 is characterized by: In steps S5 and S6, the regularization term is added to the total loss function during training to define a regularizer: ； During training, the regularization term is added to the overall loss function: ； ； in: is the standard cross entropy loss, is the number of categories; is the real label, represented by one-hot encoding, where only the position corresponding to the target category is 1 and the other positions are 0; is the label probability predicted by the model, i.e. The output value of It is a learnable parameter used to control the influence of the regularization term; is the regularized loss defined by the regularizer, which should be in vector form ,in It is thermonuclear rights.

4. The dual-stream spatiotemporal brain network analysis method embedded with population priors according to claim 1 is characterized by: In step S7, the label Z obtained according to the spatiotemporal dual-stream model is embedded into the patient node. Each patient in the healthy group is regarded as a positive sample, so that they are clustered together in the feature space. At the same time, each patient in the patient group is regarded as a negative sample, so that they are dispersed and pushed away in the feature space. Each patient in the patient group is regarded as a positive sample, so that they are clustered together in the feature space. At the same time, each patient in the healthy group is regarded as a negative sample, so that they are sparse and pushed away in the feature space. Finally, a group graph G consisting of two groups of clusters is obtained. p .

5. The dual-stream spatiotemporal brain network analysis method embedded with population priors according to claim 1 is characterized by: In step S7, the group graph The edge is constructed as the Top-k neighborhood method, which selects the k neighbor nodes with the highest similarity measure of the input node to connect the bidirectional Top-k edges.

6. The dual-stream spatiotemporal brain network analysis method embedded with population priors according to claim 1, characterized in that: In step S4, the graph isomorphism network is: ； ； in: is a brain network constructed using Pearson correlation. For the Layer Node Initialize the eigenvector of is a fully connected layer; For learnable parameters or set a fixed scalar; Is with the node A set of adjacent nodes, It is an aggregation node Neighbor node information.

7. The dual-stream spatiotemporal brain network analysis method embedded with population priors according to claim 1 is characterized by: In step S1, the DPARSF toolbox of Matlab was used to preprocess the brain functional map imaging to remove the front At each time point, the image is corrected for head motion artifacts and time layers, and the effects of ventricular and white matter signals and head motion high-order effects are removed. The corrected image is registered to the standard space. Temporal bandpass filtering to reduce the effects of heartbeat and respiration.

Images