A Brain Function Dynamic Prediction Device, Equipment and Medium Based on rTMS

Abstract

The present invention provides a brain function dynamic prediction device, equipment and medium based on rTMS, relating to the technical field of brain data processing. The device is used to acquire multi-modal magnetic resonance imaging data, perform preprocessing, and generate a BOLD time series; perform Hilbert transform on the BOLD time series to determine the phase coherence matrix at the same time point between different brain regions; perform weighted eigenvector dynamics analysis on the phase coherence matrix to generate weighted eigenvectors; perform clustering analysis on the weighted eigenvectors to generate functional connectivity brain state clustering data; determine the similarity between the functional connectivity brain state clustering data and the Yeo resting state network, and generate a functional connectivity brain network state; determine the state change data between the functional connectivity brain network states to generate a state transition probability matrix; generate an rTMS prediction result according to the transition probability matrix and brain indicators. The present invention can improve the prediction efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of brain data processing, and more particularly, to a brain function dynamic prediction device, equipment and medium based on rTMS. Background Art

[0002] Repetitive transcranial magnetic stimulation (rTMS) is a non-invasive neuromodulation technique that induces an induced current in brain tissue through a time-varying magnetic field to regulate the excitability of neurons in specific brain regions and can be applied to the dynamic regulation research of neural function activities.

[0003] In the related art, to optimize the targeting and parameter adaptability of neuromodulation, rTMS prediction is generally based on the static functional connection characteristics of resting-state functional magnetic resonance, and the regulation strategy is adjusted according to the prediction results. However, brain function activities have significant dynamic characteristics, and their functional connection patterns will continuously change over time and states, rather than a single stable state. The single-time-point analysis based on static functional connection cannot comprehensively characterize the dynamic change characteristics of the brain network, resulting in low accuracy of rTMS prediction. Summary of the Invention

[0004] The problem solved by the present invention is how to improve the accuracy of rTMS prediction.

[0005] To solve the above problems, the present invention provides a brain function dynamic prediction device, equipment and medium based on rTMS.

[0006] In a first aspect, the present invention provides a brain function dynamic prediction device based on rTMS, including:

[0007] An acquisition module, configured to acquire multi-modal magnetic resonance imaging data of an rTMS user, where the multi-modal magnetic resonance imaging data includes structural MRI data and resting-state fMRI data at multiple time points;

[0008] A processing module, configured to preprocess the multi-modal magnetic resonance imaging data to generate a BOLD time series;

[0009] A phase module, configured to perform a Hilbert transform on the BOLD time series to determine the BOLD signal phase of each brain region of the rTMS user, and determine a phase coherence matrix at the same time point between different brain regions according to the BOLD signal phase;

[0010] A weighting module, configured to perform weighted eigenvector dynamics analysis on the phase coherence matrix to generate weighted eigenvectors;

[0011] A clustering module, configured to perform clustering analysis on the weighted eigenvectors to generate functional connection brain state clustering data;

[0012] A similarity module, configured to determine the similarity between the functional connectivity brain state clustering data and the Yeo resting state network, and generate a functional connectivity brain network state based on the similarity;

[0013] A change module, configured to determine state change data between the functional connectivity brain network states at two adjacent time points, and generate a state transition probability matrix based on the state change data;

[0014] A prediction module, configured to generate an rTMS prediction result according to the transition probability matrix and brain metrics.

[0015] Optionally, the weighting module includes a formula unit;

[0016] The formula unit is configured to perform weighted eigenvector dynamics analysis on the phase coherence matrix by using a weighted eigenvector dynamics analysis formula to generate the weighted eigenvector. The weighted eigenvector dynamics analysis formula includes:

[0017] ;

[0018] Wherein, is the weighted eigenvector, and are respectively two non-zero eigenvalues in the phase coherence matrix, and are respectively and corresponding eigenvectors.

[0019] Optionally, the state change data includes a state change manner and a change count;

[0020] The change module includes a change determination unit and a normalization unit;

[0021] The change determination unit is configured to determine the state change manner between the functional connectivity brain network states at two adjacent time points, and determine the change count of each state change manner, wherein the state change manner includes multiple types;

[0022] The normalization unit is configured to normalize the change count to generate a state transition probability matrix.

[0023] Optionally, the similarity module includes a Pearson unit, a similarity unit, and an overlap degree unit;

[0024] The Pearson unit is configured to extract the weighted eigenvector in the functional connectivity brain state clustering data, and determine the Pearson correlation coefficient between the weighted eigenvector in the functional connectivity brain state clustering data and the Yeo resting state network;

[0025] The similarity unit is used to generate the similarity according to the Pearson correlation coefficient;

[0026] The overlap degree unit is used to determine the overlap degree between the functional connection brain state clustering data and the Yeo resting state network according to the similarity, and generate the functional connection brain network state according to the overlap degree.

[0027] Optionally, generating the rTMS prediction result according to the transition probability matrix and the brain index includes:

[0028] Generating the rTMS prediction result by using the trained support vector regression model according to the transition probability matrix and the brain index.

[0029] Optionally, the processing module includes a structural preprocessing unit and a resting state preprocessing unit;

[0030] The structural preprocessing unit is used to sequentially perform skull stripping, tissue segmentation, and spatial normalization operations on the structural MRI data to generate preprocessed structural MRI data;

[0031] The resting state preprocessing unit is used to preprocess the resting state fMRI data according to the preprocessed structural MRI data to generate the BOLD time series.

[0032] Optionally, the resting state preprocessing unit includes a primary preprocessing subunit, a final preprocessing subunit, a postprocessing subunit, and a mapping subunit;

[0033] The primary preprocessing subunit is used to sequentially perform motion correction, scan layer time correction, and susceptibility distortion correction operations on the resting state fMRI data to generate primary preprocessed resting state fMRI data;

[0034] The final preprocessing subunit is used to perform structural-functional image registration and MNI space normalization operations on the primary preprocessed resting state fMRI data based on the preprocessed structural MRI data to generate final preprocessed resting state fMRI data;

[0035] The postprocessing subunit is used to postprocess the final preprocessed resting state fMRI data to generate the four-dimensional image of the BOLD time series;

[0036] The mapping subunit is used to map the four-dimensional image of the BOLD time series to the Schaefer atlas in the MNI space to generate the BOLD time series of each brain region.

[0037] Optionally, the clustering module includes a centroid unit, a label unit, a new centroid unit, and a loop unit;

[0038] The centroid unit is configured to generate a plurality of initial centroids according to the weighted feature vectors;

[0039] The label unit is configured to determine the clustering label of each weighted feature vector according to the initial centroids;

[0040] The new centroid unit is configured to generate new centroids according to the clustering labels and replace the corresponding initial centroids;

[0041] The loop unit is configured to return and execute the step of determining the clustering label of each weighted feature vector according to the initial centroids according to the new centroids, and loop until a convergence condition is reached, and generate the functional connectivity brain state clustering data according to the final new centroids and the corresponding weighted feature vectors.

[0042] In a second aspect, the present invention provides an electronic device, including a memory and a processor;

[0043] The memory is configured to store a computer program;

[0044] The processor is configured to, when executing the computer program, implement the following steps:

[0045] Obtain multi-modal magnetic resonance imaging data of an rTMS user, where the multi-modal magnetic resonance imaging data includes structural MRI data and resting-state fMRI data at multiple time points;

[0046] Preprocess the multi-modal magnetic resonance imaging data to generate a BOLD time series;

[0047] Perform a Hilbert transform on the BOLD time series to determine the BOLD signal phase of each brain region of the rTMS user, and determine a phase coherence matrix between different brain regions at the same time point according to the BOLD signal phase;

[0048] Perform weighted feature vector dynamics analysis on the phase coherence matrix to generate weighted feature vectors;

[0049] Perform clustering analysis on the weighted feature vectors to generate functional connectivity brain state clustering data;

[0050] Determine the similarity between the functional connectivity brain state clustering data and the Yeo resting-state network, and generate a functional connectivity brain network state based on the similarity;

[0051] Determine the state change data between the functional connectivity brain network states at two adjacent time points, and generate a state transition probability matrix based on the state change data;

[0052] Generate an rTMS prediction result according to the transition probability matrix and brain metrics.

[0053] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0054] Obtain multi-modal magnetic resonance imaging data of an rTMS user, wherein the multi-modal magnetic resonance imaging data includes structural MRI data and resting-state fMRI data at multiple time points;

[0055] Preprocess the multi-modal magnetic resonance imaging data to generate a BOLD time series;

[0056] Perform a Hilbert transform on the BOLD time series to determine the BOLD signal phase of each brain region of the rTMS user, and determine the phase coherence matrix between different brain regions at the same time point according to the BOLD signal phase;

[0057] Perform weighted eigenvector dynamics analysis on the phase coherence matrix to generate weighted eigenvectors;

[0058] Perform clustering analysis on the weighted eigenvectors to generate functional connectivity brain state clustering data;

[0059] Determine the similarity between the functional connectivity brain state clustering data and the Yeo resting-state network, and generate a functional connectivity brain network state based on the similarity;

[0060] Determine the state change data between the functional connectivity brain network states at two adjacent time points, and generate a state transition probability matrix based on the state change data;

[0061] Generate an rTMS prediction result according to the transition probability matrix and brain metrics.

[0062] The beneficial effects of the rTMS-based brain function dynamic prediction device, equipment and medium of the present invention are:

[0063] By using an acquisition module to obtain multi-modal magnetic resonance imaging data of rTMS users, and the multi-modal magnetic resonance imaging data includes structural MRI data and resting-state fMRI data at multiple time points, the dynamic change characteristics of the brain structure and function of rTMS users within a certain period of time can be understood, thereby providing a data basis for subsequent analysis, and further obtaining accurate prediction results. By using a processing module to preprocess the multi-modal magnetic resonance imaging data, noise and other interference factors can be removed, ensuring the accuracy of the BOLD time series, thus more truly reflecting the brain activity. Then, through a phase module, Hilbert transform is performed on the BOLD time series. Hilbert transform allows the extraction of the instantaneous phase information of the BOLD signal, capturing the instantaneous synchronization changes between brain regions, and further calculating the phase coherence matrix, providing a data basis for understanding the brain network dynamics. Then, through a weighting module, weighted eigenvector dynamics analysis is performed on the phase coherence matrix to generate weighted eigenvectors, which more comprehensively describe the subtle dynamic changes of the brain state, and can improve the understanding accuracy of the complex dynamic patterns of the brain. Then, through a clustering module, clustering analysis is performed on the weighted eigenvectors to generate functional connectivity brain state clustering data, which can classify the functional connectivity patterns of the brain at different time points into different stable states. Then, through a similarity module, the functional connectivity brain state clustering data is compared with the known Yeo resting-state network to generate a functional connectivity brain network state, which can enhance the biological interpretability of the results. Then, through a change module, the state change data between the functional connectivity brain network states at two adjacent time points is determined, and based on the state change data, a state transition probability matrix is generated, which can reveal the transition rules between brain network states. Finally, through a prediction module, the transition probability matrix is analyzed with various brain metrics to obtain accurate rTMS prediction results with scientific basis. Description of the Drawings

[0064] Figure 1 FIG. is a schematic structural diagram of a brain function dynamic prediction device based on rTMS provided by an embodiment of the present invention;

[0065] Figure 2 FIG. is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. Detailed Embodiments

[0066] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention will be given with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments described herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present invention. It should be understood that the drawings and embodiments of the present invention are only for exemplary purposes and are not used to limit the protection scope of the present invention.

[0067] It should be understood that the various steps described in the method embodiments of the present invention may be executed in different orders and / or executed in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this regard.

[0068] As used herein, the term "including" and its variations are open-ended, that is, "including but not limited to"; the term "based on" is "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiment". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts such as "first" and "second" mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order of the functions performed by these devices, modules or units or their interdependent relationships.

[0069] It should be noted that the modifications of "one" and "a plurality of" mentioned in the present invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise clearly specified in the context, it should be understood as "one or more".

[0070] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only for illustrative purposes and are not used to limit the scope of these messages or information.

[0071] In view of the problems existing in the above related technologies, this embodiment provides a brain function dynamic prediction device, device and medium based on rTMS.

[0072] As Figure 1 shown, a brain function dynamic prediction device based on rTMS provided by an embodiment of the present invention includes:

[0073] An acquisition module, configured to acquire multi-modal magnetic resonance imaging data of an rTMS user, where the multi-modal magnetic resonance imaging data includes structural MRI data and resting-state fMRI data at multiple time points.

[0074] Specifically, an rTMS user refers to a user who receives repetitive transcranial magnetic stimulation (rTMS) treatment. The acquisition module collects multimodal magnetic resonance imaging data at different time points, such as at baseline, after the acute phase, and during the follow-up phase, in order to comprehensively evaluate the long-term dynamic effects of rTMS on the user's brain structure and function. Among them, MRI refers to magnetic resonance imaging. The structural MRI data is obtained by scanning the brain structure with a high-resolution T1-weighted structural MRI, and the structural MRI data helps to detect morphological changes and possible morphological changes in specific brain structures, so as to evaluate whether the rTMS treatment will exacerbate or alleviate the changes in the brain state network. The resting-state fMRI data is obtained by recording the changes in the BOLD signal of the brain's spontaneous activity when the subject is in a resting state using an MRI device, so that the intrinsic functional connectivity pattern of the brain can be studied through the resting-state fMRI data.

[0075] A processing module, configured to preprocess the multimodal magnetic resonance imaging data to generate a BOLD time series.

[0076] Specifically, the processing module preprocesses the multimodal magnetic resonance imaging data to improve the data quality, ensure the consistency and reliability of the data, reduce the influence of noise and other interference factors, generate a four-dimensional BOLD time series image, and map the four-dimensional BOLD time series image to the Schaefer atlas in the MNI standard space. The Schaefer atlas in the MNI standard space includes 400 brain regions, that is, the rTMS user's brain is divided into 400 brain regions, and each brain region has a corresponding BOLD time series. Among them, BOLD is the blood oxygenation level-dependent functional signal, and the four dimensions of the four-dimensional BOLD time series image refer to the three-dimensional image plus the corresponding time dimension.

[0077] A phase module, configured to perform a Hilbert transform on the BOLD time series to determine the BOLD signal phase of each brain region of the rTMS user, and determine the phase coherence matrix at the same time point between different brain regions according to the BOLD signal phase.

[0078] Specifically, the phase module performs a Hilbert transform on the BOLD time series to generate the evolution of the BOLD time series for each brain region, thereby determining the BOLD signal phase of each brain region of the rTMS user. Furthermore, based on the BOLD signal phase, using the phase coherence formula, the phase coherence between each pair of brain regions n and p at time t is estimated by calculating the cosine of the phase difference, and the phase coherence matrix at the same time point t between different brain regions n and p is determined. The phase coherence matrix is a 3D matrix of size N×N×T, where N is the number of delimited brain regions and N = 400, and T is the total number of time frames. It should be noted that the phase coherence matrix for each time frame is undirected, so the phase coherence matrix is a symmetric matrix. The phase coherence formula includes:

[0079] ;

[0080] where, is the phase coherence matrix, n and p are brain regions respectively, t is the time point, is the angular value of the phase difference.

[0081] The weighting module is used to perform weighted eigenvector dynamics analysis on the phase coherence matrix to generate weighted eigenvectors.

[0082] Specifically, the weighted eigenvector dynamics analysis method (WEiDA) of the weighting module in this embodiment is modified based on the LEiDA method. The weighted eigenvector dynamics analysis method takes into account the information contained in other eigenvectors ignored by the LEiDA method. That is, when performing weighted eigenvector dynamics analysis on the phase coherence matrix, not only the main eigenvectors in the phase coherence matrix are considered, but also other eigenvectors are considered, and the generated weighted eigenvectors are the sum of the weighted main eigenvectors and other eigenvectors respectively.

[0083] The clustering module is used to perform clustering analysis on the weighted eigenvectors to generate functional connectivity brain state clustering data.

[0084] Specifically, the clustering module uses the K-means algorithm to perform clustering analysis on the weighted eigenvectors, classifies the functional connectivity patterns of the brain at different time points into different states, and generates functional connectivity brain state clustering data.

[0085] In one embodiment, the clustering module includes a centroid unit, a label unit, a new centroid unit, and a loop unit;

[0086] The centroid unit is used to generate multiple initial centroids according to the weighted eigenvectors;

[0087] The label unit is used to determine the clustering label of each weighted eigenvector according to the initial centroids;

[0088] The new centroid unit is used to generate a new centroid according to the clustering label and replace the corresponding initial centroid;

[0089] The loop unit is used to, according to the new centroid, return to execute the step of determining the clustering label of each weighted feature vector according to the initial centroid, loop and execute until a convergence condition is reached, and generate the functional connectivity brain state clustering data according to the final new centroid and the corresponding weighted feature vector.

[0090] Specifically, input the weighted feature vectors corresponding to each time point into the K-means algorithm. The K-means algorithm randomly selects k initial centroids, where k is multiple; for each weighted feature vector, calculate its distance from each centroid and assign it to the cluster represented by the nearest centroid to obtain the clustering label of each weighted feature vector; recalculate the average value of all data points within each cluster and use this new average value as the new centroid of the cluster; according to the new centroid, loop the above steps until the distribution of the clusters no longer changes significantly or reaches the convergence condition, and the convergence condition can be set according to the actual situation. For example, loop a preset number of times, output the finally determined k centroids and the corresponding weighted feature vectors, that is, k clusters and the weighted feature vectors in each cluster, and generate the functional connectivity brain state clustering data. All the weighted feature vectors are finally divided into k clusters, and each cluster represents a group of weighted feature vectors, that is, the functional connectivity brain state clustering data (FC Brain States). These functional connectivity brain state clustering data reflect the stable connection characteristics shown by the brain at different time points. By analyzing these states, the regulation mechanism and its dynamic changes of the brain network during the rTMS treatment process can be deeply understood.

[0091] A similarity module is used to determine the similarity between the functional connectivity brain state clustering data and the Yeo resting-state network, and generate a functional connectivity brain network state based on the similarity.

[0092] Specifically, the Yeo resting-state network is a method for partitioning brain functional networks based on the analysis of resting-state functional magnetic resonance imaging (rs-fMRI) data. It divides the brain into seven main resting-state networks, namely: the prefrontal network, the default mode network, the visual network, the motor network, the auditory network, the dorsal attention network, and the anterior cingulate cortex network. These seven networks may play different roles in different cognitive and behavioral tasks. By determining the similarity between the functional connectivity brain state clustering data and the Yeo resting-state network through similar modules, and based on the similarity, it can be determined which resting-state network the functional connectivity brain state clustering data corresponds to, thereby generating a functional connectivity brain network state. For example, if the functional connectivity brain state clustering data corresponding to the k-th cluster is similar to the brain regions of the visual network, then the functional connectivity brain network state generated corresponding to this functional connectivity brain state clustering data can be named the visual functional state.

[0093] A change module, configured to determine state change data between the functional connectivity brain network states at two adjacent time points, and generate a state transition probability matrix based on the state change data.

[0094] Specifically, the state change data includes the state change mode and the number of changes. The change module marks the functional connectivity brain network states at each time point, and counts the number of occurrences of each functional connectivity brain network state, as well as the proportion of each functional connectivity brain network state in the total time series, that is, the probability distribution of each functional connectivity brain network state, for subsequent prediction analysis. The change module then analyzes the two functional connectivity brain network states at two adjacent time points, records the state change mode, that is, the process of the i-th functional connectivity brain network state transitioning to other functional connectivity brain network states (including the i-th functional connectivity brain network state), and records the number of times of each state change mode of the i-th functional connectivity brain network state. It should be understood that each functional connectivity brain network state includes at least one state change mode. And the corresponding transition probability is determined. The transition probabilities of all functional connectivity brain network states constitute the final state transition probability matrix. The state transition probability matrix can not only intuitively display the transition rules between states, but also reveal the internal characteristics of the dynamic changes of the brain, such as the persistence, easy convertibility of a certain state, and the interconversion relationship with other states. As a quantitative state description index, this matrix provides a new perspective for subsequent clinical effect analysis, can capture the subtle differences in the brain state transition during the rTMS treatment process, and thus improve the accuracy of subsequent predictions.

[0095] A prediction module, configured to generate an rTMS prediction result according to the transition probability matrix and the brain index.

[0096] In one embodiment, the generating the rTMS prediction result according to the transition probability matrix and the brain index includes:

[0097] Based on the conversion probability matrix and the brain metrics, a trained support vector regression model is used to generate the rTMS prediction result.

[0098] Specifically, the brain metrics include structural metrics, functional metrics, and microstructural metrics. The structural metrics can be the brain anatomical features reflected by structural MRI data, and the functional metrics can be the BOLD signal intensity and functional connectivity strength of resting-state fMRI data. The prediction module inputs the conversion probability matrix and the brain metrics into the trained support vector regression model, thereby giving the rTMS prediction result. Through experiments, it is known that the conversion between certain specific states has a higher probability of occurring in users with significantly improved treatment effects. Then, the conversion between certain specific states can be used as a potential biomarker for predicting the rTMS treatment response. In the subsequent prediction process, if there is a high probability of conversion between certain specific states in an rTMS user, the rTMS prediction result is that the rTMS user has a better acceptance effect for rTMS. Exemplarily, the brain metrics can also include other statistical metrics, such as the average duration, occurrence frequency, and conversion frequency of each state.

[0099] In this embodiment, a multi-modal magnetic resonance imaging data of an rTMS user is obtained by an acquisition module, and the multi-modal magnetic resonance imaging data includes structural MRI data and resting-state fMRI data at multiple time points, which can understand the dynamic change characteristics of the brain structure and function of the rTMS user within a certain period of time, thereby providing a data basis for subsequent analysis, and then obtaining an accurate prediction result. The multi-modal magnetic resonance imaging data is preprocessed by a processing module to remove noise and other interference factors, ensuring the accuracy of the BOLD time series, and thus more truly reflecting the brain activity. Then, the Hilbert transform is performed on the BOLD time series by a phase module. The Hilbert transform allows the extraction of the instantaneous phase information of the BOLD signal, captures the instantaneous synchronization changes between brain regions, and then calculates the phase coherence matrix, providing a data basis for understanding the brain network dynamics. Then, a weighted eigenvector dynamics analysis is performed on the phase coherence matrix by a weighting module to generate a weighted eigenvector, which more comprehensively describes the subtle dynamic changes of the brain state and can improve the understanding accuracy of the complex dynamic patterns of the brain. Then, a clustering analysis is performed on the weighted eigenvector by a clustering module to generate functional connectivity brain state clustering data, which can classify the functional connectivity patterns of the brain at different time points into different stable states. Then, the functional connectivity brain state clustering data is compared with the known Yeo resting-state network by a similarity module to generate a functional connectivity brain network state, which can enhance the biological interpretability of the results. Then, a change module determines the state change data between the functional connectivity brain network states at two adjacent time points, and based on the state change data, generates a state transition probability matrix, which can reveal the transition rules between brain network states. Finally, a prediction module analyzes the transition probability matrix with various brain metrics to obtain an accurate rTMS prediction result with scientific basis.

[0100] Optionally, the weighting module includes a formula unit;

[0101] The formula unit is used to perform a weighted eigenvector dynamics analysis on the phase coherence matrix by using a weighted eigenvector dynamics analysis formula to generate the weighted eigenvector. The weighted eigenvector dynamics analysis formula includes:

[0102] ;

[0103] where is the weighted eigenvector, and are respectively two non-zero eigenvalues in the phase coherence matrix, and are respectively the eigenvectors corresponding to and .

[0104] Specifically, the weighted eigenvector dynamics analysis method (WEiDA) is modified based on the LEiDA method. The weighted eigenvector dynamics analysis method takes into account the information contained in other eigenvectors that the LEiDA method ignores, that is, performs weighted eigenvector dynamics analysis on the phase coherence matrix, considering not only the main eigenvectors in the phase coherence matrix , but also other eigenvectors , and the generated weighted eigenvector is the sum of the weighted main eigenvector and other eigenvectors respectively, that is .

[0105] Optionally, the state change data includes the state change mode and the number of changes;

[0106] The change module includes a change determination unit and a normalization unit;

[0107] The change determination unit is used to determine the state change mode between the functional connectivity brain network states at two adjacent time points, and determine the number of changes for each state change mode, where the state change mode includes multiple types;

[0108] The normalization unit is used to normalize the number of changes to generate a state transition probability matrix.

[0109] Specifically, the change determination unit analyzes the two functional connectivity brain network states at two adjacent time points, records the state change mode, that is, the process of the i-th functional connectivity brain network state transitioning to other functional connectivity brain network states (including the i-th functional connectivity brain network state), and records the number of times for each state change mode of the i-th functional connectivity brain network state, that is, the number of changes. It should be understood that each functional connectivity brain network state includes at least one state change mode. Then the normalization unit normalizes the number of changes to determine the transition probability corresponding to the i-th functional connectivity brain network state, and the transition probabilities of all functional connectivity brain network states constitute the final state transition probability matrix.

[0110] Optionally, the similarity module includes a Pearson unit, a similarity unit, and an overlap degree unit;

[0111] The Pearson unit is used to extract the weighted eigenvector in the functional connectivity brain state clustering data and determine the Pearson correlation coefficient between the weighted eigenvector in the functional connectivity brain state clustering data and the Yeo resting state network;

[0112] The similarity unit is used to generate the similarity according to the Pearson correlation coefficient;

[0113] The overlap degree unit is used to determine the overlap degree between the functional connectivity brain state clustering data and the Yeo resting-state network according to the similarity, and generate the functional connectivity brain network state according to the overlap degree.

[0114] Specifically, the Pearson unit extracts the weighted feature vector, i.e., the WEiDA vector, from the centroid of the phase-locked state of the functional connectivity brain state clustering data, compares the weighted feature vector in the functional connectivity brain state clustering data with the Yeo resting-state network, calculates the Pearson correlation coefficient, and thus evaluates the similarity between each centroid and each resting-state network through the similarity unit. Furthermore, the overlap degree unit quantifies the overlap degree between the functional connectivity brain state clustering data and the Yeo resting-state network, and determines which resting-state network the functional connectivity brain state clustering data corresponds to according to the overlap degree, so as to generate the functional connectivity brain network state. For example, if the functional connectivity brain state clustering data corresponding to the k-th clustering is similar to the brain regions of the visual network, the functional connectivity brain network state generated corresponding to this functional connectivity brain state clustering data can be named the visual functional state.

[0115] Optionally, the processing module includes a structural preprocessing unit and a resting-state preprocessing unit;

[0116] The structural preprocessing unit is used to sequentially perform skull stripping, tissue segmentation, and spatial normalization operations on the structural MRI data to generate preprocessed structural MRI data;

[0117] The resting-state preprocessing unit is used to preprocess the resting-state fMRI data according to the preprocessed structural MRI data to generate the BOLD time series.

[0118] Specifically, the structural preprocessing unit uses the fMRIPrep neuroimaging preprocessing application to sequentially perform skull stripping, tissue segmentation, and spatial normalization operations on the structural MRI data to generate preprocessed structural MRI data, and then the resting-state preprocessing unit preprocesses the resting-state fMRI data according to the preprocessed structural MRI data to generate the BOLD time series.

[0119] Optionally, the resting-state preprocessing unit includes a primary preprocessing subunit, a final preprocessing subunit, a post-processing subunit, and a mapping subunit;

[0120] The primary preprocessing subunit is used to sequentially perform motion correction, slice time correction, and susceptibility distortion correction operations on the resting-state fMRI data to generate primary preprocessed resting-state fMRI data;

[0121] The final preprocessing subunit is used to perform structural-functional image registration and MNI space standardization operations on the primary preprocessed resting-state fMRI data based on the preprocessed structural MRI data, and generate the final preprocessed resting-state fMRI data;

[0122] The postprocessing subunit is used to postprocess the final preprocessed resting-state fMRI data to generate the four-dimensional BOLD time series image;

[0123] The mapping subunit is used to map the four-dimensional BOLD time series image to the Schaefer atlas in the MNI space to generate the BOLD time series of each brain region.

[0124] Specifically, the primary preprocessing subunit first performs motion correction, slice time correction, and susceptibility distortion correction operations on the resting-state fMRI data in sequence to generate the primary preprocessed resting-state fMRI data. Since the primary preprocessed resting-state fMRI data cannot achieve accurate spatial standardization, the final preprocessing subunit needs to perform structural-functional image registration and MNI space standardization operations on the primary preprocessed resting-state fMRI data based on the preprocessed structural MRI data to generate the final preprocessed resting-state fMRI data. The postprocessing subunit then uses the XCP_D postprocessing noise regression tool to postprocess the final preprocessed resting-state fMRI data to generate the denoised four-dimensional BOLD time series image. The mapping subunit maps the four-dimensional BOLD time series image to the Schaefer atlas in the MNI space. Among them, the Schaefer atlas in the MNI standard space includes 400 brain regions. The voxel-level signals are averaged at the brain region level to obtain the BOLD time series of 400 brain regions. At the same time, band-pass filtering is performed, and the band-pass filtering frequency range can be 0.01~0.08 Hz to obtain the BOLD time series of each brain region.

[0125] As Figure 2 shown, an electronic device 200 provided by an embodiment of the present invention includes a memory 210 and a processor 220; the memory 210 is used to store a computer program; the processor 220 is used to implement the following steps when executing the computer program:

[0126] Obtain multi-modal magnetic resonance imaging data of the rTMS user, where the multi-modal magnetic resonance imaging data includes structural MRI data and resting-state fMRI data at multiple time points;

[0127] Preprocess the multi-modal magnetic resonance imaging data to generate a BOLD time series;

[0128] Perform Hilbert transform on the BOLD time series, determine the BOLD signal phase of each brain region of the rTMS user, and based on the BOLD signal phase, determine the phase coherence matrix between different brain regions at the same time point;

[0129] Perform weighted eigenvector dynamics analysis on the phase coherence matrix to generate weighted eigenvectors;

[0130] Perform clustering analysis on the weighted eigenvectors to generate functional connectivity brain state clustering data;

[0131] Determine the similarity between the functional connectivity brain state clustering data and the Yeo resting state network, and based on the similarity, generate a functional connectivity brain network state;

[0132] Determine the state change data between the functional connectivity brain network states at two adjacent time points, and based on the state change data, generate a state transition probability matrix;

[0133] Generate an rTMS prediction result according to the transition probability matrix and brain metrics.

[0134] A computer-readable storage medium provided by an embodiment of the present invention, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0135] Obtain multi-modal magnetic resonance imaging data of the rTMS user, where the multi-modal magnetic resonance imaging data includes structural MRI data and resting state fMRI data at multiple time points;

[0136] Preprocess the multi-modal magnetic resonance imaging data to generate a BOLD time series;

[0137] Perform Hilbert transform on the BOLD time series, determine the BOLD signal phase of each brain region of the rTMS user, and based on the BOLD signal phase, determine the phase coherence matrix between different brain regions at the same time point;

[0138] Perform weighted eigenvector dynamics analysis on the phase coherence matrix to generate weighted eigenvectors;

[0139] Perform clustering analysis on the weighted eigenvectors to generate functional connectivity brain state clustering data;

[0140] Determine the similarity between the functional connectivity brain state clustering data and the Yeo resting state network, and based on the similarity, generate a functional connectivity brain network state;

[0141] Determine the state change data between the functional connectivity brain network states at two adjacent time points, and generate a state transition probability matrix based on the state change data;

[0142] Generate an rTMS prediction result according to the transition probability matrix and brain metrics.

[0143] Now, an electronic device 200 that can be a server or a client of the present invention will be described. It is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device 200 is intended to represent various forms of digital electronic computer devices, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device 200 can also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0144] The electronic device 200 includes a computing unit that can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) or a computer program loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The computing unit, the ROM, and the RAM are connected to each other through a bus. An input / output (I / O) interface is also connected to the bus.

[0145] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc. In the present application, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiments of the present invention. In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0146] Although the present invention is disclosed as above, the scope of protection of the present invention is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will all fall within the scope of protection of the present invention.

Claims

1. A device for dynamic prediction of brain function based on rTMS, characterized in that: include: An acquisition module, used to acquire multimodal magnetic resonance imaging data of the rTMS user, wherein the multimodal magnetic resonance imaging data includes structural MRI data and resting state fMRI data at multiple time points; A processing module, used for preprocessing the multimodal magnetic resonance imaging data to generate a BOLD time series; A phase module, used to perform Hilbert transform on the BOLD time series, determine the BOLD signal phase of each brain region of the rTMS user, and determine the phase coherence matrix between different brain regions at the same time point based on the BOLD signal phase; A weighting module, used for performing weighted eigenvector dynamics analysis on the phase coherence matrix to generate a weighted eigenvector; A clustering module, used for performing cluster analysis on the weighted feature vectors to generate functional connectivity brain state clustering data; a similarity module, for determining the similarity between the functionally connected brain state clustering data and the Yeo resting state network, and generating a functionally connected brain network state based on the similarity; a change module, used to determine state change data between the functionally connected brain network states at two adjacent time points, and generate a state transition probability matrix based on the state change data; A prediction module, used for generating rTMS prediction results according to the conversion probability matrix and brain indicators; The weighting module includes a formula unit; The formula unit is used to use a weighted eigenvector dynamics analysis formula to perform a weighted eigenvector dynamics analysis on the phase coherence matrix to generate the weighted eigenvector. The weighted eigenvector dynamics analysis formula includes: ； in, is the weighted feature vector, and are the two non-zero eigenvalues ​​in the phase coherence matrix, and Respectively and The corresponding feature vector.

2. The device for dynamic prediction of brain function based on rTMS according to claim 1, characterized in that: The state change data includes the state change mode and the number of changes; The change module includes a change determination unit and a normalization unit; The change determination unit is used to determine the state change mode between the functional connection brain network states at two adjacent time points, and determine the number of changes of each state change mode, wherein the state change mode includes multiple modes; The normalization unit is used to normalize the number of changes to generate a state transition probability matrix.

3. The device for dynamic prediction of brain function based on rTMS according to claim 1, characterized in that: The similarity module includes a Pearson unit, a similarity unit and an overlap unit; The Pearson unit is used to extract the weighted feature vector in the functional connectivity brain state clustering data, and determine the Pearson correlation coefficient between the weighted feature vector in the functional connectivity brain state clustering data and the Yeo resting state network; The similarity unit is used to generate the similarity according to the Pearson correlation coefficient; The overlap unit is used to determine the overlap between the functionally connected brain state clustering data and the Yeo resting state network according to the similarity, and generate the functionally connected brain network state according to the overlap.

4. The device for dynamic prediction of brain function based on rTMS according to claim 1, characterized in that: Generating rTMS prediction results according to the conversion probability matrix and brain indicators includes: The rTMS prediction result is generated according to the conversion probability matrix and the brain index using a trained support vector regression model.

5. The device for dynamic prediction of brain function based on rTMS according to claim 1, characterized in that: The processing module includes a structure preprocessing unit and a resting state preprocessing unit; The structural preprocessing unit is used to perform skull stripping, tissue segmentation and space standardization operations on the structural MRI data in sequence to generate preprocessed structural MRI data; The resting state preprocessing unit is used to preprocess the resting state fMRI data according to the preprocessed structural MRI data to generate the BOLD time series.

6. The device for dynamic prediction of brain function based on rTMS according to claim 5, characterized in that: The resting state preprocessing unit includes a primary preprocessing subunit, a final preprocessing subunit, a post-processing subunit and a mapping subunit; The primary preprocessing subunit is used to perform motion correction, scanning layer time correction and magnetic susceptibility distortion correction operations on the resting state fMRI data in sequence to generate primary preprocessed resting state fMRI data; The final preprocessing subunit is used to perform structural functional image registration and MNI space standardization operations on the primary preprocessed resting-state fMRI data based on the preprocessed structural MRI data to generate final preprocessed resting-state fMRI data; The post-processing subunit is used to post-process the final pre-processed resting-state fMRI data to generate the BOLD time series four-dimensional image; The mapping subunit is used to map the BOLD time series four-dimensional image to the Schaefer atlas of the MNI space to generate the BOLD time series of each brain region.

7. The device for dynamic prediction of brain function based on rTMS according to claim 1, characterized in that: The clustering module includes a centroid unit, a label unit, a new centroid unit and a cycle unit; The centroid unit is used to generate a plurality of initial centroids according to the weighted feature vector; The label unit is used to determine the cluster label of each weighted feature vector according to the initial centroid; The new centroid unit is used to generate a new centroid according to the clustering label and replace the corresponding initial centroid; The loop unit is used to return to the step of determining the cluster label of each weighted feature vector based on the initial center of mass according to the new center of mass, and loop until the convergence condition is reached, and generate the functional connection brain state clustering data according to the final new center of mass and the corresponding weighted feature vector.

8. An electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is configured to implement the following steps when executing the computer program: Acquiring multimodal magnetic resonance imaging data of the rTMS user, wherein the multimodal magnetic resonance imaging data includes structural MRI data and resting state fMRI data at multiple time points; Preprocessing the multimodal magnetic resonance imaging data to generate a BOLD time series; Performing Hilbert transformation on the BOLD time series to determine the BOLD signal phase of each brain region of the rTMS user, and determining the phase coherence matrix between different brain regions at the same time point based on the BOLD signal phase; Performing weighted eigenvector dynamics analysis on the phase coherence matrix to generate weighted eigenvectors; Performing cluster analysis on the weighted feature vectors to generate functional connectivity brain state cluster data; Determining similarity between the functionally connected brain state clustering data and the Yeo resting state network, and generating a functionally connected brain network state based on the similarity; Determining state change data between the functionally connected brain network states at two adjacent time points, and generating a state transition probability matrix based on the state change data; generating rTMS prediction results according to the conversion probability matrix and brain indicators; The step of performing weighted eigenvector dynamics analysis on the phase coherence matrix to generate a weighted eigenvector comprises: The weighted eigenvector dynamics analysis formula is used to perform a weighted eigenvector dynamics analysis on the phase coherence matrix to generate the weighted eigenvector. The weighted eigenvector dynamics analysis formula includes: ； in, is the weighted feature vector, and are the two non-zero eigenvalues ​​in the phase coherence matrix, and Respectively and The corresponding feature vector.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program. When the computer program is executed by the processor, the following steps are implemented: Acquiring multimodal magnetic resonance imaging data of the rTMS user, wherein the multimodal magnetic resonance imaging data includes structural MRI data and resting state fMRI data at multiple time points; Preprocessing the multimodal magnetic resonance imaging data to generate a BOLD time series; Performing Hilbert transformation on the BOLD time series to determine the BOLD signal phase of each brain region of the rTMS user, and determining the phase coherence matrix between different brain regions at the same time point based on the BOLD signal phase; Performing weighted eigenvector dynamics analysis on the phase coherence matrix to generate weighted eigenvectors; Performing cluster analysis on the weighted feature vectors to generate functional connectivity brain state cluster data; Determining similarity between the functionally connected brain state clustering data and the Yeo resting state network, and generating a functionally connected brain network state based on the similarity; Determining state change data between the functionally connected brain network states at two adjacent time points, and generating a state transition probability matrix based on the state change data; generating rTMS prediction results according to the conversion probability matrix and brain indicators; The step of performing weighted eigenvector dynamics analysis on the phase coherence matrix to generate a weighted eigenvector comprises: The weighted eigenvector dynamics analysis formula is used to perform a weighted eigenvector dynamics analysis on the phase coherence matrix to generate the weighted eigenvector. The weighted eigenvector dynamics analysis formula includes: ； in, is the weighted feature vector, and are the two non-zero eigenvalues ​​in the phase coherence matrix, and Respectively and The corresponding feature vector.

Images