Transient brain network construction method and system

By processing fMRI data using the between-subject analysis method and the dynamic conditional correlation method, the problems of low accuracy and sensitivity of the sliding window technology in constructing transient brain networks were solved, and more accurate transient brain network construction was achieved.

CN116452533BActive Publication Date: 2025-10-21SOUTH CHINA NORMAL UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202310382532.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-04-12
Publication Date
2025-10-21
Estimated Expiration
2043-04-12

AI Technical Summary

Technical Problem

Traditional sliding window technology has problems with low accuracy and sensitivity when constructing transient brain networks, resulting in inaccurate calculation results.

Method used

The fMRI data were processed using the between-subject analysis method and the dynamic conditional correlation method. By obtaining natural stimulus processing data from multiple subjects, the original time series were extracted and standardized. The functional connectivity state data was determined using the K-means clustering algorithm, and the conditional covariance matrix was calculated frame by frame to construct the transient brain network.

Benefits of technology

It improves the accuracy of transient brain networks, overcomes the limitations of sliding window technology, and can more accurately construct transient brain networks, which is applicable to all natural stimulation paradigms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116452533B_ABST
    Figure CN116452533B_ABST
Patent Text Reader

Abstract

The application discloses a transient brain network construction method and system, and relates to the technical field of functional magnetic resonance. fMRI data of natural stimulus processing of a plurality of subjects is acquired; original time series of the subjects is extracted; the original time series is subjected to standardization processing to obtain standardized time series; the standardized time series of a target subject is analyzed by using an inter-subject analysis method to obtain stimulus-induced time series; frame-by-frame correlation calculation is performed on the stimulus-induced time series by using a dynamic conditional correlation method to obtain a conditional covariance matrix of the target subject at all time frames; an inter-subject correlation matrix is calculated from the conditional covariance matrix as a functional connection matrix of the target subject at all time frames; the functional connection state data of the target subject which can be repeatedly present is determined based on a K-means clustering algorithm; and the functional connection state data is visually displayed to obtain a visualized transient brain network. The application improves the precision and sensitivity of constructing the transient brain network.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of functional magnetic resonance technology, and in particular to a transient brain network construction method and system. Background Art

[0002] Traditional cognitive neuroscience research relies on carefully designed parametric tasks (such as block or event-related designs) to investigate the relationship between brain and behavior. However, these traditional experimental designs suffer from low ecological validity and are incomparable to the authentic stimuli and complex behaviors encountered in real life. In the past five years, an increasing number of researchers have begun using naturalistic stimuli in cognitive neuroscience research.

[0003] To track the dynamic changes in brain activity under natural stimulation, it is necessary to calculate the dynamic functional connectivity (dFC) between different brain regions in real time using functional magnetic resonance imaging (fMRI). fMRI signals under natural stimulation consist of three components: stimulus-induced signals, spontaneous fluctuations, and irrelevant noise. Spontaneous fluctuations and irrelevant noise are non-stimulus-induced signals.

[0004] Currently, the most commonly used strategy is the sliding window technique. Although the sliding window technique is widely used, it has major drawbacks:

[0005] 1. The sliding window only considers the values ​​within the window and ignores the values ​​outside the window.

[0006] 2. The size of the sliding window is arbitrary. A window that is too large can easily lead to low sensitivity, while a window that is too small can easily lead to fluctuations in correlation, resulting in inaccurate calculation results.

[0007] 3. The sliding window technique calculates the functional connectivity of whole-brain signals (stimulation-induced and non-stimuli-induced signals). The calculation results include the correlation of non-stimuli-induced signals, which has low accuracy.

[0008] Therefore, the sliding window technology has the problems of low accuracy and low sensitivity, and the constructed transient brain network is not accurate enough. Summary of the Invention

[0009] The purpose of the embodiments of the present invention is to provide a method and system for constructing a transient brain network to improve the accuracy of constructing a transient brain network.

[0010] To achieve the above objectives, the present invention provides the following solutions:

[0011] A method for constructing a transient brain network, comprising:

[0012] Acquiring fMRI data of natural stimulus processing from multiple subjects; the fMRI data corresponding to one subject includes multiple frames of fMRI images;

[0013] Extracting the original time series of each subject from the pre-processed fMRI data of the subject; the original time series includes the time series corresponding to different regions in the brain partition atlas; performing normalization processing on the original time series to obtain a normalized time series;

[0014] The standardized time series of the target subject is analyzed using an inter-subject analysis method to obtain a stimulus-induced time series; the stimulus-induced time series includes: the standardized time series of the target subject and an inter-subject average time series; the inter-subject average time series is obtained by averaging the time series of multiple subjects except the target subject; the target subject is any one of the multiple subjects;

[0015] A dynamic conditional correlation method is used to perform frame-by-frame correlation calculation on the stimulus-induced time series to obtain a conditional covariance matrix of the target subject over all time frames; an inter-subject correlation matrix is ​​calculated from the conditional covariance matrix as a functional connectivity matrix of the target subject over all time frames; the inter-subject correlation matrix represents the correlation coefficients between stimulus-induced signals in various brain regions of the target subject;

[0016] Determining reproducible functional connectivity state data of the target subject based on a K-means clustering algorithm; the functional connectivity state data is used to characterize transient brain network changes induced by natural stimuli;

[0017] The functional connectivity state data is visualized to obtain a visualized transient brain network.

[0018] Optionally,

[0019] The extracting of the original time series of each subject from the pre-processed fMRI data of the subject specifically includes:

[0020] The fMRI data of each subject were preprocessed to obtain processed fMRI data; the preprocessing specifically included: time correction, head motion correction, spatial registration to standard space, Gaussian smoothing and filtering;

[0021] Match the brain partition atlas with the preprocessed fMRI data to obtain the original time series corresponding to the regions in the brain atlas;

[0022] The original time series corresponding to the region in the brain atlas is normalized by using a zero-mean or Z-normalization method to obtain the normalized time series.

[0023] Optionally,

[0024] The standardized time series of the target subject is analyzed by the between-subject analysis method to obtain the stimulation-induced time series, which specifically includes:

[0025] Select the standardized time series of one target subject; average the standardized time series of the remaining N-1 subjects to obtain the average time series between subjects, where N is a positive integer; merge the average time series between subjects and the standardized time series of the target subject into a stimulus-evoked time series.

[0026] Optionally,

[0027] The dynamic conditional correlation method is used to perform frame-by-frame correlation calculation on the stimulus-induced time series to obtain the conditional covariance matrix of the target subject in all time frames, specifically including:

[0028] The stimulus-induced time series of each subject is calculated using a generalized autoregressive conditional heteroskedasticity model to obtain standard residuals; the standard residuals are calculated using an exponentially weighted moving average method to obtain a time-varying correlation matrix; and the conditional covariance matrix is ​​obtained based on the time-varying correlation matrix.

[0029] Optionally,

[0030] The step of calculating the inter-subject correlation matrix from the conditional covariance matrix as the functional connectivity matrix of the target subject at all time frames specifically includes:

[0031] Calculate the correlation coefficient of the time series corresponding to any pair of different regions of each subject based on the conditional covariance matrix;

[0032] The correlation matrix of each subject in all time frames is calculated based on the correlation coefficient; the correlation matrix includes the intra-subject correlation matrix and the inter-subject correlation matrix; the inter-subject correlation matrix extracted from the upper right corner of the correlation matrix is ​​used as the functional connectivity matrix of the target subject in all time frames.

[0033] Optionally,

[0034] The method of determining the reproducible functional connectivity state data of the target subject based on the K-means clustering algorithm specifically includes:

[0035] The functional connectivity matrix is ​​estimated using the elbow method to obtain the optimal cluster number k;

[0036] Perform cluster analysis based on the functional connectivity matrices of all subjects to obtain k clusters; each cluster corresponds to a number; a functional connectivity matrix of each subject corresponds to a number; the number sequence corresponding to the functional connectivity matrix of all frames of each subject is the state transfer vector of the subject; the state transfer vector is used to represent the functional connectivity state data of the subject; the state transfer vectors of all subjects constitute the state transfer matrix;

[0037] The k clusters are reordered to obtain the reproducible functional connectivity state data.

[0038] To achieve the above objectives, the present invention further provides the following solutions:

[0039] A transient brain network construction system, comprising:

[0040] Acquiring fMRI data of natural stimulus processing from multiple subjects; the fMRI data corresponding to one subject includes multiple frames of fMRI images;

[0041] Extracting the original time series of each subject from the pre-processed fMRI data of the subject; the original time series includes the time series corresponding to different regions in the brain partition atlas; performing normalization processing on the original time series to obtain a normalized time series;

[0042] Correlation calculation module, used for:

[0043] The standardized time series of the target subject is analyzed using an inter-subject analysis method to obtain a stimulus-induced time series; the stimulus-induced time series includes: the standardized time series of the target subject and an inter-subject average time series; the inter-subject average time series is obtained by averaging the time series of multiple subjects except the target subject; the target subject is any one of the multiple subjects;

[0044] A dynamic conditional correlation method is used to perform frame-by-frame correlation calculation on the stimulus-induced time series to obtain a conditional covariance matrix of the target subject over all time frames; an inter-subject correlation matrix is ​​calculated from the conditional covariance matrix as a functional connectivity matrix of the target subject over all time frames; the inter-subject correlation matrix represents the correlation coefficients between stimulus-induced signals in various brain regions of the target subject;

[0045] Functional connection state module, used to:

[0046] Determining reproducible functional connectivity state data of the target subject based on a K-means clustering algorithm; the functional connectivity state data is used to characterize transient brain network changes induced by natural stimuli;

[0047] Display module for:

[0048] The functional connectivity state data is visualized to obtain a visualized transient brain network.

[0049] Optionally, the time series acquisition module includes:

[0050] Pre-processing unit for:

[0051] The fMRI data of each subject were preprocessed to obtain processed fMRI data; the preprocessing specifically included: time correction, head motion correction, spatial registration to standard space, Gaussian smoothing and filtering;

[0052] Time series extraction unit, used to:

[0053] Match the brain partition atlas with the preprocessed fMRI data to obtain the original time series corresponding to the regions in the brain atlas;

[0054] Standardized units for:

[0055] The original time series corresponding to the region in the brain atlas is normalized by using a zero-mean or Z-normalization method to obtain the normalized time series.

[0056] Optionally, the correlation calculation module includes:

[0057] Average time series calculation unit, used for:

[0058] Select the standardized time series of one target subject; average the standardized time series of the remaining N-1 subjects to obtain the average time series between subjects, where N is a positive integer; merge the average time series between subjects and the standardized time series of the target subject into a stimulus-evoked time series;

[0059] Transient functional connectivity matrix calculation unit, used for:

[0060] The stimulus-induced time series of each subject is calculated using a generalized autoregressive conditional heteroskedasticity model to obtain a standard residual; the standard residual is calculated using an exponentially weighted moving average method to obtain a time-varying correlation matrix; and the conditional covariance matrix is ​​obtained based on the time-varying correlation matrix;

[0061] Calculate the correlation coefficient of the time series corresponding to any pair of different regions of each subject based on the conditional covariance matrix;

[0062] The correlation matrix of each subject in all time frames is calculated based on the correlation coefficient; the correlation matrix includes the intra-subject correlation matrix and the inter-subject correlation matrix; the inter-subject correlation matrix extracted from the upper right corner of the correlation matrix is ​​used as the functional connectivity matrix of the target subject in all time frames.

[0063] Optionally, the functional connection status module includes:

[0064] Cluster analysis unit, used to:

[0065] The functional connectivity matrix is ​​estimated using the elbow method to obtain the optimal cluster number k;

[0066] Perform cluster analysis based on the functional connectivity matrices of all subjects to obtain k clusters; each cluster corresponds to a number; a functional connectivity matrix of each subject corresponds to a number; the number sequence corresponding to the functional connectivity matrix of all frames of each subject is the state transfer vector of the subject; the state transfer vector is used to represent the functional connectivity state data of the subject; the state transfer vectors of all subjects constitute the state transfer matrix;

[0067] Sorting unit, used to:

[0068] The k clusters are reordered to obtain the reproducible functional connectivity state data.

[0069] In an embodiment of the present invention, an inter-subject analysis method is used to analyze the original time series of the target subject to obtain a stimulus-induced time series. The inter-subject average time series is a time series that is consistent across subjects under the stimulus-induced signal. Since non-stimulus-induced signals are removed, the accuracy of constructing transient brain networks can be improved. A dynamic conditional correlation method is used to calculate the frame-by-frame correlation of the stimulus-induced time series to obtain the conditional covariance matrix of the target subject on all time frames. The transient brain network is constructed frame by frame. Based on the conditional covariance matrix, the functional connectivity state data of each subject is obtained. The functional connectivity state data is visualized to obtain a transient brain network, overcoming the limitations of the sliding window technology.

[0070] Stimulus-evoked time series are highly consistent across subjects, while spontaneous fluctuations and irrelevant noise are uncorrelated across subjects. The inter-subject average time series is the time series that exhibits subject consistency under the stimulus-evoked time series. Inter-subject analysis and dynamic conditional correlation methods offer high accuracy and sensitivity, enabling precise reconstruction of transient brain networks. This method is applicable to all naturalistic stimulation paradigms.

[0071] The dynamic conditional correlation method effectively utilizes information from stimulus-evoked time series. Compared to the sliding window dynamic functional connectivity method, dynamic conditional correlation is calculated frame by frame, overcoming many limitations of the sliding window method and improving the accuracy of constructing transient brain networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0073] Figure 1 A schematic diagram of a process for constructing a transient brain network according to an embodiment of the present invention;

[0074] Figure 2 A schematic diagram of clustering status provided by an embodiment of the present invention;

[0075] Figure 3 A schematic diagram of a state transition vector provided by an embodiment of the present invention;

[0076] Figure 4 A schematic diagram of a state transition matrix provided in an embodiment of the present invention;

[0077] Figure 5 A schematic structural diagram of a transient brain network construction system provided in an embodiment of the present invention.

[0078] Explanation of symbols:

[0079] Time series acquisition module-1, correlation calculation module-2, functional connection status module-3, display module-4. DETAILED DESCRIPTION

[0080] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0081] The purpose of the present invention is to provide a transient brain network construction method and system to solve the problems of low accuracy and low sensitivity of the existing sliding window technology, and the problem that the constructed transient brain network is not accurate enough.

[0082] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0083] Figure 1 An exemplary process of the above-mentioned transient brain network construction method is shown. Each step is described in detail below.

[0084] Step 1: Acquire fMRI data of natural stimulus processing from multiple subjects; the fMRI data corresponding to one subject includes multiple frames of fMRI images; extract the original time series of each subject from the preprocessed fMRI data of the subject; the original time series includes the time series corresponding to different regions in the brain partition map; standardize the original time series to obtain a standardized time series.

[0085] Step 11: Preprocess the fMRI data of each subject to obtain processed fMRI data; the preprocessing specifically includes: time correction, head motion correction, spatial alignment to standard space, Gaussian smoothing and filtering.

[0086] In one example, the more accurately the processed fMRI data is, the more it can reflect the brain's activity in response to natural stimuli, which is more beneficial for data analysis.

[0087] Step 11 may be specifically performed by a pre-processing unit.

[0088] Step 12: Match the brain partition map with the preprocessed fMRI data to obtain the original time series corresponding to the regions in the brain map.

[0089] In one example, a brain parcellation atlas can be selected from public sources or customized based on needs. A specific brain parcellation atlas can retain signals (fMRI data) from brain regions of interest or intended study. For example, if one wants to study the dynamics of the default network, one can select brain regions associated with the default network and construct a brain parcellation atlas.

[0090] Step 12 may be specifically performed by a time series extraction unit.

[0091] Step 13: Using a zero-mean or Z-normalization method, the original time series corresponding to the region in the brain atlas is normalized to obtain the normalized time series.

[0092] In one example, zero-meaning or Z-standardization can be used to reduce the effects of volatility and noise in a time series. Z-standardization can take many forms, such as zero-meaning the data and dividing it by the individual standard deviation, or zero-meaning the data and dividing it by the overall standard deviation. Both methods produce very similar results.

[0093] Step 13 may be specifically performed by a standardization unit.

[0094] Step 2: Use the between-subject analysis method to analyze the standardized time series of the target subject to obtain the stimulus-induced time series; the stimulus-induced time series includes: the standardized time series of the target subject and the between-subject average time series; the between-subject average time series is obtained by averaging the time series of multiple subjects except the target subject; the target subject is any one of the multiple subjects.

[0095] Step 2 can be specifically performed by the correlation calculation module 2.

[0096] The standardized time series of the target subjects were analyzed using the between-subjects analysis method, and the stimulus-induced time series obtained specifically included:

[0097] Step 21: Select the standardized time series of one target subject; average the standardized time series of the remaining N-1 subjects to obtain the average time series between subjects, where N is a positive integer; merge the average time series between subjects and the standardized time series of the target subject into a stimulus-evoked time series.

[0098] In one example, the leave-one-out method is used to calculate the average time series between subjects. All subjects participating in the experiment received exactly the same natural stimulus materials, so the brain activities of all subjects induced by natural stimulation are consistent between subjects, while the subject-specific spontaneous activities and noise are not consistent between subjects. Therefore, the leave-one-out method is used to calculate the average time series between subjects, representing the consistent task-induced activities of these subjects. The horizontal axis of the average time series between subjects represents different moments (in frames, T), and the vertical axis represents different brain regions (N). Each time a subject is selected, the standardized time series of all remaining subjects are averaged to obtain the average time series between subjects, and the average time series between subjects (NxT) and the original time series (NxT) of the target subject are merged into a stimulus-induced time series (2NxT).

[0099] Step 21 may be specifically performed by an average time series calculation unit.

[0100] Step 3: Use the dynamic conditional correlation method to calculate the frame-by-frame correlation of the stimulus-induced time series to obtain the conditional covariance matrix of the target subject in all time frames; calculate the inter-subject correlation matrix from the conditional covariance matrix as the functional connectivity matrix of the target subject in all time frames; the inter-subject correlation matrix represents the correlation coefficient between the stimulus-induced signals of each brain region of the target subject.

[0101] Step 3 may be specifically performed by the correlation calculation module 2 .

[0102] The dynamic conditional correlation method is used to calculate the frame-by-frame correlation of the stimulus-induced time series, and the conditional covariance matrix of the target subject in all time frames is obtained, which specifically includes:

[0103] Step 31: Use the generalized autoregressive conditional heteroskedasticity model to calculate the stimulus-evoked time series of each subject to obtain the standard residual; use the exponentially weighted moving average method to calculate the standard residual to obtain the time-varying correlation matrix; based on the time-varying correlation matrix, obtain the conditional covariance matrix.

[0104] Step 31 may be specifically performed by a transient functional connectivity matrix calculation unit.

[0105] Calculating the inter-subject correlation matrix from the conditional covariance matrix as the functional connectivity matrix of the target subject at all time frames specifically includes:

[0106] Step 32: Calculate the correlation coefficient of the time series corresponding to any pair of different regions for each subject based on the conditional covariance matrix.

[0107] Step 32 may be specifically performed by a transient functional connectivity matrix calculation unit.

[0108] Step 33: Calculate the correlation matrix of each subject in all time frames based on the correlation coefficient; the correlation matrix includes the intra-subject correlation matrix and the inter-subject correlation matrix; the inter-subject correlation matrix extracted from the upper right corner of the correlation matrix is ​​used as the functional connectivity matrix of the target subject in all time frames.

[0109] Step 33 may be specifically performed by a transient functional connectivity matrix calculation unit.

[0110] In one example, a subject was scanned with 8 minutes of fMRI image data, with each frame scanning time of 2 seconds (TR = 2 seconds). Assuming that there are connections between 10 brain regions, the calculated transient functional connectivity matrix is ​​10x10x240.

[0111] Step 4: Determine the reproducible functional connectivity state data of the target subject based on the K-means clustering algorithm; the functional connectivity state data is used to characterize transient brain network changes induced by natural stimulation.

[0112] Step 4 may be specifically performed by the functional connection status module 3 .

[0113] The K-means clustering algorithm is used to determine the reproducible functional connectivity data of the target subjects, including:

[0114] Step 41: Use the elbow method to estimate the functional connectivity matrix to obtain the optimal clustering number k.

[0115] Cluster analysis is performed based on the functional connectivity matrices of all subjects to obtain k clusters; each cluster corresponds to a number; a functional connectivity matrix of each subject corresponds to a number; the number sequence corresponding to the functional connectivity matrix of all frames of each subject is the state transfer vector of the subject; the state transfer vector is used to represent the functional connectivity state data of the subject; the state transfer vectors of all subjects constitute the state transfer matrix.

[0116] Step 41 may be specifically performed by a cluster analysis unit.

[0117] Step 42: Reorder the k clusters to obtain reproducible functional connectivity state data.

[0118] Step 42 may be specifically performed by a sorting unit.

[0119] In one example, a K-means clustering method is used for cluster analysis to decompose the functional connectivity matrix into k non-overlapping clusters. Based on spatial similarity, the functional connectivity matrix on each frame is assigned to different clusters so that the Manhattan distance between the matrix and the cluster center is minimized. Figure 2 , the horizontal axis is the brain region number, and the vertical axis is the brain region number; each functional connectivity state data (cluster state) corresponds to a number. All subjects receive a corresponding number according to the cluster they belong to at each time frame. Each subject has a corresponding number sequence, which is called the state transition vector of the subject. For state transition vector (functional connectivity state data), please refer to Figure 3 , the horizontal axis is the time, the vertical axis is the state number; the state transfer vectors of all subjects constitute the state transfer matrix, the state transfer matrix can be found in Figure 4 , the horizontal axis is the time, and the vertical axis is the subject number.

[0120] Step 5: Visualize the functional connectivity state data to obtain a visualized transient brain network.

[0121] Step 5 may be specifically executed by the display module 4 .

[0122] In an example, the display module 4 can be displayed using the imagesc function in Matlab.

[0123] In summary, in an embodiment of the present invention, an inter-subject analysis method is used to analyze the standardized time series of the target subject to obtain a stimulus-induced time series. Since the non-stimulation-induced signals are removed, the accuracy of constructing a transient brain network can be improved. A dynamic conditional correlation method is used to calculate the frame-by-frame correlation of the stimulus-induced time series to obtain the conditional covariance matrix of the target subject on all time frames. The inter-subject correlation matrix is ​​calculated from the conditional covariance matrix as the functional connectivity matrix of the target subject on all time frames. The transient brain network is constructed frame by frame. According to the transient functional connectivity matrix, the functional connectivity state data of each subject is obtained. The functional connectivity state data is visualized to obtain a transient brain network, which overcomes the limitations of the sliding window technology.

[0124] Stimulus-evoked signals are highly consistent across subjects, while spontaneous fluctuations and irrelevant noise are uncorrelated across subjects. The inter-subject average time series represents the time series that exhibits consistency across subjects under stimulus-evoked signals. Inter-subject analysis and dynamic conditional correlation methods offer high accuracy and sensitivity, enabling precise reconstruction of transient brain networks. This method is applicable to all naturalistic stimulation paradigms.

[0125] Dynamic conditional correlation methods effectively utilize information in time series. Compared to sliding window dynamic functional connectivity methods, dynamic conditional correlation is calculated frame by frame, overcoming many limitations of sliding window methods and improving the accuracy of constructing transient brain networks.

[0126] In addition, the dynamic conditional correlation method does not rely on prior parameter selection. This method parameterizes the conditional correlation and all parameters are effectively estimated using maximum likelihood estimation, which is more convenient and more efficient.

[0127] See Figure 5 A transient brain network construction system includes at least: a time series acquisition module 1, a correlation calculation module 2, a functional connection state module 3, and a display module 4.

[0128] The time series acquisition module 1 is used to acquire fMRI data of natural stimulus processing of multiple subjects; the fMRI data corresponding to one subject includes multiple frames of fMRI images.

[0129] The time series acquisition module 1 is also used to extract the original time series of each subject from the preprocessed fMRI data of the subject; the original time series includes the time series corresponding to different areas in the brain partition map; the original time series is standardized to obtain a standardized time series.

[0130] The time series acquisition module 1 includes: a preprocessing unit, a time series extraction unit, and a standardization unit.

[0131] The preprocessing unit is used to preprocess the fMRI data of each subject to obtain processed fMRI data; the preprocessing specifically includes: time correction, head motion correction, spatial alignment to standard space, Gaussian smoothing and filtering.

[0132] The time series unit is used to match the brain partition map with the preprocessed fMRI data to obtain the original time series corresponding to the regions in the brain map.

[0133] The standardization unit is used to standardize the original time series corresponding to the region in the brain atlas using the zero mean or Z-standardization method to obtain a standardized time series.

[0134] For the description of time series acquisition 1 and the preprocessing unit, time series extraction unit, and normalization unit, please refer to the above and will not be repeated here.

[0135] The correlation calculation module 2 is used to analyze the standardized time series of the target subject using the inter-subject analysis method to obtain the stimulus-induced time series; the stimulus-induced time series includes: the standardized time series of the target subject and the inter-subject average time series; the inter-subject average time series is obtained by averaging the time series of multiple subjects except the target subject; the target subject is any one of the multiple subjects.

[0136] The correlation calculation module 2 is also used to perform frame-by-frame correlation calculation on the stimulus-induced time series using a dynamic conditional correlation method to obtain the conditional covariance matrix of the target subject in all time frames; the inter-subject correlation matrix is ​​calculated from the conditional covariance matrix as the functional connection matrix of the target subject in all time frames; the inter-subject correlation matrix represents the correlation coefficient between the stimulus-induced signals of each brain region of the target subject.

[0137] The correlation calculation module 2 includes: an average time series calculation unit and a transient functional connectivity matrix calculation unit.

[0138] The average time series calculation unit is used to select the standardized time series of one target subject; average the standardized time series of the remaining N-1 subjects to obtain the average time series between subjects, where N is a positive integer; and merge the average time series between subjects and the standardized time series of the target subject into a stimulus-induced time series.

[0139] The transient functional connectivity matrix calculation unit is used to calculate the stimulus-induced time series of each subject using a generalized autoregressive conditional heteroskedasticity model to obtain a standard residual; the standard residual is calculated using an exponentially weighted moving average method to obtain a time-varying correlation matrix; and the conditional covariance matrix is ​​obtained based on the time-varying correlation matrix.

[0140] The correlation coefficient of the time series corresponding to any pair of different regions of each subject is calculated based on the conditional covariance matrix.

[0141] The correlation matrix of each subject in all time frames was calculated based on the correlation coefficient; the correlation matrix included the intra-subject correlation matrix and the inter-subject correlation matrix; the inter-subject correlation matrix extracted from the upper right corner of the correlation matrix was used as the functional connectivity matrix of the target subject in all time frames.

[0142] Regarding the description of the correlation calculation module 2, the average time series calculation unit, and the transient functional connectivity matrix calculation unit, please refer to the above and will not be repeated here.

[0143] Functional connectivity state module 3 is used to determine the reproducible functional connectivity state data of the target subject based on the K-means clustering algorithm; the functional connectivity state data is used to characterize transient brain network changes induced by natural stimulation.

[0144] Functional connectivity status module 3 includes: a cluster analysis unit and a sorting unit.

[0145] The cluster analysis unit is used to estimate the functional connectivity matrix using the elbow method to obtain the optimal cluster number k.

[0146] The cluster analysis unit is also used to perform cluster analysis based on the functional connectivity matrices of all subjects to obtain k clusters; each cluster corresponds to a number; a functional connectivity matrix of each subject corresponds to a number; the number sequence corresponding to the functional connectivity matrix of all frames of each subject is the state transfer vector of the subject; the state transfer vector is used to represent the functional connectivity state data of the subject; the state transfer vectors of all subjects constitute the state transfer matrix.

[0147] The sorting unit is used to reorder the k clusters to obtain reproducible functional connectivity state data.

[0148] Regarding the description of the functional connectivity state module 3, the cluster analysis unit, and the sorting unit, please refer to the above and will not be repeated here.

[0149] The display module 4 is used to visualize the functional connectivity state data to obtain a visualized transient brain network.

[0150] Please refer to the above for the description of the display module 4, which will not be repeated here.

[0151] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0152] This document uses specific examples to illustrate the principles and implementation methods of the embodiments of the present invention. The description of the above embodiments is only intended to help understand the methods and core concepts of the embodiments of the present invention. At the same time, for those skilled in the art, based on the concepts of the embodiments of the present invention, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the embodiments of the present invention.

Claims

1. A method for constructing a transient brain network, characterized in that: include: Acquiring fMRI data of natural stimulus processing from multiple subjects; the fMRI data corresponding to one subject includes multiple frames of fMRI images; Extracting the original time series of each subject from the pre-processed fMRI data of the subject; the original time series includes the time series corresponding to different regions in the brain partition atlas; performing normalization processing on the original time series to obtain a normalized time series; The standardized time series of the target subject is analyzed using an inter-subject analysis method to obtain a stimulus-induced time series; the stimulus-induced time series includes: the standardized time series of the target subject and an inter-subject average time series; the inter-subject average time series is obtained by averaging the time series of multiple subjects except the target subject; the target subject is any one of the multiple subjects; A dynamic conditional correlation method is used to perform frame-by-frame correlation calculation on the stimulus-induced time series to obtain a conditional covariance matrix of the target subject over all time frames; an inter-subject correlation matrix is ​​calculated from the conditional covariance matrix as a functional connectivity matrix of the target subject over all time frames; the inter-subject correlation matrix represents the correlation coefficients between stimulus-induced signals in various brain regions of the target subject; Determining reproducible functional connectivity state data of the target subject based on a K-means clustering algorithm; the functional connectivity state data is used to characterize transient brain network changes induced by natural stimuli; The functional connectivity state data is visualized to obtain a visualized transient brain network.

2. The transient brain network construction method according to claim 1, characterized in that: The extracting of the original time series of each subject from the pre-processed fMRI data of the subject specifically includes: The fMRI data of each subject were preprocessed to obtain processed fMRI data; the preprocessing specifically included: time correction, head motion correction, spatial registration to standard space, Gaussian smoothing and filtering; Match the brain partition atlas with the preprocessed fMRI data to obtain the original time series corresponding to the regions in the brain atlas; The original time series corresponding to the region in the brain atlas is normalized by using a zero-mean or Z-normalization method to obtain the normalized time series.

3. The transient brain network construction method according to claim 1, characterized in that: The standardized time series of the target subject is analyzed by the between-subject analysis method to obtain the stimulation-induced time series, which specifically includes: Select the standardized time series of one target subject; average the standardized time series of the remaining N-1 subjects to obtain the average time series between subjects, where N is a positive integer; merge the average time series between subjects and the standardized time series of the target subject into a stimulus-evoked time series.

4. The transient brain network construction method according to claim 1, characterized in that: The dynamic conditional correlation method is used to perform frame-by-frame correlation calculation on the stimulus-induced time series to obtain the conditional covariance matrix of the target subject in all time frames, specifically including: The stimulus-induced time series of each subject is calculated using a generalized autoregressive conditional heteroskedasticity model to obtain standard residuals; the standard residuals are calculated using an exponentially weighted moving average method to obtain a time-varying correlation matrix; and the conditional covariance matrix is ​​obtained based on the time-varying correlation matrix.

5. The transient brain network construction method according to claim 1, characterized in that: The step of calculating the inter-subject correlation matrix from the conditional covariance matrix as the functional connectivity matrix of the target subject at all time frames specifically includes: Calculating the correlation coefficient of the time series corresponding to any pair of different regions in the brain partition map of each subject based on the conditional covariance matrix; The correlation matrix of each subject in all time frames is calculated based on the correlation coefficient; the correlation matrix includes the intra-subject correlation matrix and the inter-subject correlation matrix; the inter-subject correlation matrix extracted from the upper right corner of the correlation matrix is ​​used as the functional connectivity matrix of the target subject in all time frames.

6. The transient brain network construction method according to claim 1, characterized in that: The method of determining the reproducible functional connectivity state data of the target subject based on the K-means clustering algorithm specifically includes: The functional connectivity matrix is ​​estimated using the elbow method to obtain the optimal cluster number k; Perform cluster analysis based on the functional connectivity matrices of all subjects to obtain k clusters; each cluster corresponds to a number; a functional connectivity matrix of each subject corresponds to a number; the number sequence corresponding to the functional connectivity matrix of all frames of each subject is the state transfer vector of the subject; the state transfer vector is used to represent the functional connectivity state data of the subject; the state transfer vectors of all subjects constitute the state transfer matrix; The k clusters are reordered to obtain the reproducible functional connectivity state data.

7. A transient brain network construction system, characterized in that: include: Time series acquisition module, used for: Acquiring fMRI data of natural stimulus processing from multiple subjects; the fMRI data corresponding to one subject includes multiple frames of fMRI images; Extracting the original time series of each subject from the pre-processed fMRI data of the subject; the original time series includes the time series corresponding to different regions in the brain partition atlas; performing normalization processing on the original time series to obtain a normalized time series; Correlation calculation module, used for: The standardized time series of the target subject is analyzed using an inter-subject analysis method to obtain a stimulus-induced time series; the stimulus-induced time series includes: the standardized time series of the target subject and an inter-subject average time series; the inter-subject average time series is obtained by averaging the time series of multiple subjects except the target subject; the target subject is any one of the multiple subjects; A dynamic conditional correlation method is used to perform frame-by-frame correlation calculation on the stimulus-induced time series to obtain a conditional covariance matrix of the target subject over all time frames; an inter-subject correlation matrix is ​​calculated from the conditional covariance matrix as a functional connectivity matrix of the target subject over all time frames; the inter-subject correlation matrix represents the correlation coefficients between stimulus-induced signals in various brain regions of the target subject; Functional connection state module, used to: Determining reproducible functional connectivity state data of the target subject based on a K-means clustering algorithm; the functional connectivity state data is used to characterize transient brain network changes induced by natural stimuli; Display module for: The functional connectivity state data is visualized to obtain a visualized transient brain network.

8. The transient brain network construction system according to claim 7, characterized in that: The time series acquisition module includes: Pre-processing unit for: The fMRI data of each subject were preprocessed to obtain processed fMRI data; the preprocessing specifically included: time correction, head motion correction, spatial registration to standard space, Gaussian smoothing and filtering; Time series extraction unit, used to: Match the brain partition atlas with the preprocessed fMRI data to obtain the original time series corresponding to the regions in the brain atlas; Standardized units for: The original time series corresponding to the region in the brain atlas is normalized by using a zero-mean or Z-normalization method to obtain the normalized time series.

9. The transient brain network construction system according to claim 7, characterized in that: The correlation calculation module includes: Average time series calculation unit, used for: Select the standardized time series of one target subject; average the standardized time series of the remaining N-1 subjects to obtain the average time series between subjects, where N is a positive integer; merge the average time series between subjects and the standardized time series of the target subject into a stimulus-evoked time series; Transient functional connectivity matrix calculation unit, used for: The stimulus-induced time series of each subject is calculated using a generalized autoregressive conditional heteroskedasticity model to obtain a standard residual; the standard residual is calculated using an exponentially weighted moving average method to obtain a time-varying correlation matrix; and the conditional covariance matrix is ​​obtained based on the time-varying correlation matrix; Calculating the correlation coefficient of the time series corresponding to any pair of different regions in the brain partition map of each subject based on the conditional covariance matrix; The correlation matrix of each subject in all time frames is calculated based on the correlation coefficient; the correlation matrix includes the intra-subject correlation matrix and the inter-subject correlation matrix; the inter-subject correlation matrix extracted from the upper right corner of the correlation matrix is ​​used as the functional connectivity matrix of the target subject in all time frames.

10. The transient brain network construction system according to claim 7, characterized in that: The functional connection status module includes: Cluster analysis unit, used to: The functional connectivity matrix is ​​estimated using the elbow method to obtain the optimal cluster number k; Perform cluster analysis based on the functional connectivity matrices of all subjects to obtain k clusters; each cluster corresponds to a number; a functional connectivity matrix of each subject corresponds to a number; the number sequence corresponding to the functional connectivity matrix of all frames of each subject is the state transfer vector of the subject; the state transfer vector is used to represent the functional connectivity state data of the subject; the state transfer vectors of all subjects constitute the state transfer matrix; Sorting unit, used to: The k clusters are reordered to obtain the reproducible functional connectivity state data.

Citation Information

Patent Citations

  • method for detecting fMRI brain network dynamic co-variations

    CN109522894A

  • fMRI data classification and identification method and device based on brain area function connection

    CN112233086A