A transient co-activation pattern analysis method and system

Abstract

The application discloses a transient co-activation pattern analysis method and system, and relates to the field of functional magnetic resonance technology. fMRI data of natural stimulus processing of a plurality of subjects is acquired; the fMRI data of each subject is preprocessed to obtain preprocessed fMRI data; a subject-to-subject analysis method is used to calculate the preprocessed fMRI data to obtain stimulus-induced signals; spontaneous activity signals and non-neural signals in the stimulus-induced signals are filtered out by a linear regression method to obtain task-induced signals; a co-activation pattern analysis method is used to analyze the task-induced signals frame by frame to obtain brain co-activation patterns; a K-means clustering method is used to identify and extract repeatedly appearing subject-to-subject co-activation states in the brain co-activation patterns; the subject-to-subject co-activation states are reordered and visually displayed to obtain visualized transient brain activation patterns. The application improves the accuracy and sensitivity of a dynamic co-activation pattern analysis method of the brain under continuous stimulation.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of functional magnetic resonance technology, in particular to a transient co-activation pattern analysis method and system. BACKGROUND

[0002] Traditional cognitive neuroscience research relies on carefully designed parametric tasks (such as block design or event-related design) to study the relationship between brain and behavior. However, the ecological validity of traditional experimental design is low, which cannot match the real-life real stimulus and complex behavior. Usually, the stimulus (picture, story, movie) is not isolated, it is dynamically linked with other modal information. In addition, the meaning of these complex stimuli also depends on the context information on different time scales. Because the traditional experimental design ignores these factors, it leads to low repeatability of traditional experimental research results, causing the reproducibility crisis. In order to get rid of this predicament, in the past five years, more and more researchers have begun to use natural stimuli for cognitive neuroscience research. Researchers call this trend the third wave of stimulus paradigm revolution (compared to traditional experimental design and resting state research). The complexity of natural stimuli also increases the difficulty of analysis.

[0003] Traditional functional magnetic resonance imaging (fMRI) activation analysis methods (such as general linear model) have strict control conditions, which are highly structured and have specific time scales, and are usually limited to models that require parametric activation, so the general linear model activation analysis method is difficult to apply to natural stimuli. Natural stimulus fMRI signals include three parts: stimulus-induced signals, spontaneous fluctuations, and irrelevant noise. Through inter-subject analysis method, the influence of the latter two parts on task-induced signals can be eliminated. However, inter-subject analysis method can only study the static brain activation pattern of the brain, and cannot reveal the dynamic co-activation pattern of the brain under continuous stimulation.

[0004] Co-activation pattern analysis (CAP) is an analysis method that evaluates the whole brain activation pattern frame by frame. However, CAP was originally used to process resting state fMRI signals. The resting state fMRI signal is single in composition, while for the fMRI signal of the natural stimulus paradigm, because it contains three components at the same time, it is difficult to determine which component causes a certain co-activation pattern, resulting in low accuracy and sensitivity. SUMMARY

[0005] The embodiment of the present application aims to provide a transient co-activation pattern analysis method and system to improve the accuracy and sensitivity of the dynamic co-activation pattern of the brain under continuous stimulation.

[0006] To achieve the above object, the embodiment of the present application provides the following scheme.

[0007] A transient co-activation pattern analysis method comprises the following steps.

[0008] fMRI data of natural stimulation processing of a plurality of subjects is acquired; the fMRI data of one subject comprises a plurality of fMRI images;

[0009] The fMRI data of each subject is preprocessed to obtain preprocessed fMRI data; the preprocessing comprises time correction, head motion correction, structural image segmentation, spatial registration, Gaussian smoothing and filtering;

[0010] The preprocessed fMRI data is calculated by using an inter-subject analysis method to obtain a stimulation-induced signal; spontaneous activity signals and non-neural signals in the stimulation-induced signal are filtered out by using a linear regression method to obtain a task-induced signal;

[0011] The task-induced signal is analyzed frame by frame by using a co-activation pattern analysis method to obtain a brain co-activation pattern; an inter-subject co-activation state repeatedly appearing in the brain co-activation pattern is identified and extracted by using a K-means clustering method;

[0012] The inter-subject co-activation state is reordered and visually displayed to obtain a visualized transient brain activation pattern.

[0013] Optionally, the calculation of the preprocessed fMRI data by using the inter-subject analysis method to obtain the stimulation-induced signal specifically comprises the following steps.

[0014] One target subject is selected from N subjects each time, and the preprocessed fMRI data of the remaining N-1 subjects is averaged and calculated; the inter-subject consistent signal is obtained after traversing the N subjects;

[0015] The spontaneous activity signals and the non-neural signals in the inter-subject consistent signal are calculated by using a linear regression method to obtain the task-induced signal.

[0016] Optionally, the frame-by-frame analysis of the task-induced signal by using the co-activation pattern analysis method to obtain the brain co-activation pattern; the inter-subject co-activation state repeatedly appearing in the brain co-activation pattern is identified and extracted by using the K-means clustering method specifically comprises the following steps.

[0017] Filter the task-induced signal by using a head movement control method to obtain a filtered inter-subject consistent signal;

[0018] Compare all the voxel activation data of the subject with a preset activation threshold T according to the filtered inter-subject consistent signal to determine a co-activation frame; or, compare all the voxel activation data of the subject with a preset percentage range P of the voxel activation degree according to the filtered inter-subject consistent signal to determine a co-activation frame;

[0019] Perform optimal cluster number estimation on a new time sequence composed of the co-activation frames to obtain an optimal cluster number k;

[0020] According to the optimal cluster number k, the co-activation frames are assigned to the corresponding brain co-activation patterns frame by frame;

[0021] The k clusters are reordered to obtain the inter-subject co-activation state.

[0022] Optionally, the inter-subject co-activation state is reordered and visually displayed to obtain a visualized transient brain activation pattern, which specifically includes:

[0023] Perform spatiotemporal analysis on the inter-subject co-activation state to obtain the spatial and temporal characteristics of the inter-subject co-activation state;

[0024] According to the spatial and temporal characteristics, a visualized transient brain activation pattern of the subject individual in natural stimulation is obtained.

[0025] Optionally, according to the filtered inter-subject consistent signal, the co-activation frame is determined by comparing all the voxel activation data of the subject with a preset activation threshold T; or, according to the filtered inter-subject consistent signal, the co-activation frame is determined by comparing all the voxel activation data of the subject with a preset percentage range P of the voxel activation degree, which specifically includes:

[0026] Obtain the voxel activation data corresponding to the whole brain region;

[0027] If the voxel activation data is greater than the preset activation threshold T, it is determined that the filtered inter-subject consistent signal corresponding to the voxel activation data is the co-activation frame;

[0028] If the voxel activation data is less than or equal to the preset activation threshold T, it is determined that the filtered inter-subject consistent signal corresponding to the voxel activation data is not the co-activation frame;

[0029] Or,

[0030] If the voxel activation data is greater than the preset percentage range P of the voxel activation level, it is determined that the inter-subject consistent signal corresponding to the voxel activation data is the co-activation frame;

[0031] If the voxel activation data is less than or equal to the preset percentage range P of the voxel activation level, it is determined that the inter-subject consistent signal corresponding to the voxel activation data is not the co-activation frame.

[0032] A transient co-activation pattern analysis system, comprising:

[0033] A time series acquisition module for acquiring fMRI data of natural stimulus processing of a plurality of subjects; the fMRI data of one of the subjects comprises a plurality of fMRI images;

[0034] A preprocessing module for preprocessing the fMRI data of each subject to obtain preprocessed fMRI data; the preprocessing includes time correction, head motion correction, structural image segmentation, spatial registration, Gaussian smoothing and filtering;

[0035] An inter-subject consistent signal acquisition module for calculating the preprocessed fMRI data using an inter-subject analysis method to obtain a stimulus-induced signal; filtering out spontaneous activity signals and non-neural signals in the stimulus-induced signal by a linear regression method to obtain a task-induced signal;

[0036] An inter-subject co-activation state determination module for analyzing the task-induced signal frame by frame using a co-activation pattern analysis method to obtain a brain co-activation pattern; identifying and extracting a repeated inter-subject co-activation state in the brain co-activation pattern using a K-means clustering method;

[0037] A visualization module for reordering and visualizing the inter-subject co-activation state to obtain a visualized transient brain activation pattern.

[0038] Optionally, the inter-subject consistent signal acquisition module comprises:

[0039] An average calculation unit for selecting one target subject from N subjects each time, and averaging the processed fMRI data of the remaining N-1 subjects to obtain the inter-subject consistent signal after traversing the N subjects;

[0040] A regression calculation unit for performing regression calculation on spontaneous activity signals and non-neural signals in the inter-subject consistent signal using a linear regression method to obtain the task-induced signal.

[0041] Optionally, the inter-subject co-activation state determination module comprises:

[0042] a head motion control unit configured to filter the task-evoked signals using a head motion control method to obtain a filtered inter-subject consistent signal;

[0043] a co-activation frame determination unit configured to determine a co-activation frame by comparing all voxel activation data of the whole brain of the subject to a preset activation threshold T according to the filtered inter-subject consistent signal, or by comparing all voxel activation data of the whole brain of the subject to a preset percentage range P of the degree of voxel activation according to the filtered inter-subject consistent signal;

[0044] a clustering analysis unit configured to:

[0045] perform optimal cluster number estimation on the new time series composed of the co-activation frames to obtain an optimal cluster number k;

[0046] assign the co-activation frames to corresponding brain co-activation patterns frame by frame according to the optimal cluster number k;

[0047] reorder the k clusters to obtain the inter-subject co-activation state.

[0048] Optionally, the visualization module comprises:

[0049] a spatio-temporal feature extraction unit configured to perform spatio-temporal analysis on the inter-subject co-activation state to obtain spatial features and temporal features of the inter-subject co-activation state;

[0050] a display unit configured to obtain a transient brain activation pattern of the subject in natural stimulation for visualization according to the spatial features and the temporal features.

[0051] Optionally, the co-activation frame determination unit comprises:

[0052] a data extraction unit configured to obtain voxel activation data corresponding to a whole brain region;

[0053] a comparison unit configured to:

[0054] if the voxel activation data is greater than the preset activation threshold T, determine that the filtered inter-subject consistent signal corresponding to the voxel activation data is the co-activation frame;

[0055] if the voxel activation data is less than or equal to the preset activation threshold T, determine that the filtered inter-subject consistent signal corresponding to the voxel activation data is not the co-activation frame;

[0056] or,

[0057] If the voxel activation data is greater than the preset percentage range P of the voxel activation degree, it is determined that the inter-subject consistent signal corresponding to the voxel activation data is the co-activation frame;

[0058] If the voxel activation data is less than or equal to the preset percentage range P of the voxel activation degree, it is determined that the inter-subject consistent signal corresponding to the voxel activation data is not the co-activation frame.

[0059] In the embodiment of the present application, the preprocessed fMRI data of the target subject is analyzed by using the inter-subject analysis method to obtain the inter-subject consistent signal, and the fMRI data of the target subject having inter-subject consistency under the natural stimulus induced signal is the inter-subject consistent signal. Compared with the co-activation pattern analysis, the inter-subject consistent signal obtained by the inter-subject co-activation pattern analysis algorithm has a stronger corresponding relationship between the brain function subsystems, the inter-subject consistency is greatly improved, and it is significantly related to the processing degree of the natural stimulus of the subject, thereby improving the accuracy and sensitivity of the dynamic co-activation pattern analysis method of the brain under continuous stimulation. BRIEF DESCRIPTION OF DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0061] Figure 1 The flowchart of the transient co-activation pattern analysis method provided by the embodiment of the present application;

[0062] Figure 2 The inter-subject analysis method provided by the embodiment of the present application;

[0063] Figure 3 The co-activation pattern analysis method provided by the embodiment of the present application;

[0064] Figure 4 The multi-cognitive process of natural understanding provided by the embodiment of the present application;

[0065] Figure 5 The structure diagram of the transient co-activation pattern analysis system provided by the embodiment of the present application.

[0066] Symbol explanation:

[0067] Time series acquisition module-1, preprocessing module-2, inter-subject consistent signal acquisition module-3, inter-subject co-activation state determination module-4, visualization module-5. DETAILED DESCRIPTION

[0068] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the protection scope of the present application.

[0069] The present application aims to provide a transient co-activation pattern analysis method and system to solve the problem of low accuracy and sensitivity of the dynamic co-activation pattern of the brain under continuous stimulation.

[0070] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0071] Figure 1 An exemplary flow of the transient co-activation pattern analysis method is shown. Each step will be described in detail below.

[0072] Step 1: Obtain fMRI data of natural stimulus processing of a plurality of subjects; the fMRI data corresponding to one subject includes a plurality of fMRI images.

[0073] Step 1 can be performed by a time sequence acquisition module 1.

[0074] In one example, the time sequence acquisition module 1 can be a magnetic resonance scanner. The magnetic resonance scanner collects the brain fMRI signals of the subjects processing natural stimuli (such as listening to stories, watching movies, etc.).

[0075] Step 2: Preprocess the fMRI data of each subject to obtain preprocessed fMRI data; the preprocessing includes time correction, head motion correction, structural image segmentation, spatial registration, Gaussian smoothing and filtering.

[0076] Step 2 can be performed by a preprocessing module 2.

[0077] In one example, since the fMRI data contains machine, environmental and physiological noise, the fMRI data needs to be preprocessed.

[0078] The preprocessing steps include layer acquisition time correction, head motion correction, structural image segmentation, spatial registration, Gaussian smoothing and filtering, etc. The purpose of preprocessing is to denoise the fMRI data and realize uniform analysis between subjects, which is a necessary operation for fMRI data.

[0079] The pre-processing step can be operated according to specific conditions, and the more rigorous the fMRI data processing is, the more the processed fMRI data can reflect the activity of the brain receiving natural stimulation, and the more advantageous the data analysis is.

[0080] Step 3: The pre-processed fMRI data is calculated by an inter-subject analysis method to obtain a stimulus-induced signal; and a spontaneous activity signal and a non-neural signal in the stimulus-induced signal are filtered out by a linear regression method to obtain a task-induced signal.

[0081] Step 3 can be performed by a consistent inter-subject signal acquisition module 3.

[0082] The pre-processed fMRI data is calculated by an inter-subject analysis method to obtain a stimulus-induced signal; and a spontaneous activity signal and a non-neural signal in the stimulus-induced signal are filtered out by a linear regression method to obtain a task-induced signal specifically includes:

[0083] Step 31: Select one target subject from N subjects each time, and average the pre-processed fMRI data of the remaining N-1 subjects to obtain a consistent inter-subject signal after traversing the N subjects.

[0084] Step 31 can be performed by an average calculation unit.

[0085] Step 32: A spontaneous activity signal and a non-neural signal in the consistent inter-subject signal are calculated by a linear regression method to obtain the task-induced signal.

[0086] Step 32 can be performed by a regression calculation unit.

[0087] In one example, the leave-one-out method is used to calculate the consistent inter-subject signal. The basic principle is that all subjects have received the same natural stimulation material, so the activity induced by natural stimulation has consistency among subjects, while the spontaneous activity specific to the subject and the noise do not have consistency among subjects, and therefore the consistent inter-subject signal can be calculated by the leave-one-out method. For example, there are n subjects, and one subject is selected each time, and the pre-processed fMRI data of the remaining n-1 subjects is averaged to obtain a consistent inter-subject signal, which is regarded as a task-induced signal. Then, the spontaneous activity signal and the non-neural signal are regressed from the consistent inter-subject signal by a linear regression method, please refer to Figure 2 .

[0088] Step 4: The task-induced signal is analyzed frame by frame by a co-activation pattern analysis method to obtain a brain co-activation pattern; and a K-means clustering method is used to identify and extract a repeatedly appearing consistent inter-subject co-activation state in the brain co-activation pattern.

[0089] Step 4 can be performed by the inter-subject co-activation state determination module 4. Specifically, it includes:

[0090] Step 41: filtering the task-evoked signals using the head motion control method to obtain inter-subject consistent signals after filtering.

[0091] Step 42: comparing all the voxel activation data of the subject's whole brain with the preset activation threshold T according to the inter-subject consistent signals after filtering to determine the co-activation frame; or comparing all the voxel activation data of the subject's whole brain with the preset percentage range P of the voxel activation level according to the inter-subject consistent signals after filtering to determine the co-activation frame. Specifically, it includes:

[0092] Step 421: obtaining the voxel activation data corresponding to the whole brain region.

[0093] Step 422: if the voxel activation data is greater than the preset activation threshold T, it is determined that the inter-subject consistent signal after filtering corresponding to the voxel activation data is the co-activation frame;

[0094] if the voxel activation data is less than or equal to the preset activation threshold T, it is determined that the inter-subject consistent signal after filtering corresponding to the voxel activation data is not the co-activation frame;

[0095] or,

[0096] if the voxel activation data is greater than the preset percentage range P of the voxel activation level, it is determined that the inter-subject consistent signal after filtering corresponding to the voxel activation data is the co-activation frame;

[0097] if the voxel activation data is less than or equal to the preset percentage range P of the voxel activation level, it is determined that the inter-subject consistent signal after filtering corresponding to the voxel activation data is not the co-activation frame.

[0098] In one example, the inter-subject consistent signal after filtering corresponding to the voxel activation data greater than the preset activation threshold T is the co-activation frame, which can also be called the super-threshold frame, and is the state transition vector in the data calculation process. Those skilled in the art can flexibly design the value of T, for example, set T = 1.

[0099] In another example, those skilled in the art can set a percentage range P of the activation level of the co-activation frame to determine the super-threshold frame. Those skilled in the art can flexibly design the value of P, for example, set P = 90%.

[0100] In addition, the co-activation frame can also be obtained based on the seed region. The seed region is generally selected by some prior knowledge. One method is to select the strongest brain region in the brain partition related to the predetermined task; another method is to have a strong hypothesis on a specific brain anatomical region, and then compare the co-activation degree of the signals of other regions of the brain with the signals of the seed region.

[0101] There are three methods to determine the co-activation with the seed region. The first is to extract the co-activation time points by calculating the correlation between all voxels activation data and the seed region at each time point. The second is to set a pre-defined activation threshold T, as described above. The third is to set a percentage range P of the co-activation frame activation degree to determine the super-threshold frame. Regardless of which method, the super-threshold frames are grouped into a new time series for the next step of analysis. The co-activation pattern analysis can have multiple groups of subjects, which can be selected according to the research purpose. It should be noted that when analyzing the differences between multiple groups of subjects, the frame-by-frame analysis is performed according to the first group of subjects input into the TbCAPs toolbox. The co-activation pattern analysis can either select all super-threshold frames for analysis or select one or more seed regions for state comparison.

[0102] Step 43: Group the co-activation frames into a new time series for optimal cluster number estimation, and obtain the optimal cluster number k.

[0103] According to the optimal cluster number k, the co-activation frames are assigned to the corresponding brain co-activation patterns frame by frame. Please refer to Figure 3 ;

[0104] The k clusters are reordered to obtain the inter-subject co-activation state.

[0105] Step 43 can be performed by a cluster analysis unit.

[0106] In one example, the cluster analysis unit uses the K-means clustering method to determine the optimal cluster number. Since the K-means clustering method is an iterative process, it cannot guarantee convergence to the global optimum, so the algorithm is run 50 times from K = 2 to K max (e.g. K max is set to 8) to determine the optimal cluster number K. The cluster analysis unit can automatically estimate the number of clusters, or the number of clusters can be pre-set.

[0107] The consensus clustering measures the stability of the super-threshold frame assignment to a certain inter-subject co-activation state, so by setting the size of the consensus clustering (e.g. 90%) for frame-by-frame analysis, it is determined whether each frame can be retained and stably assigned to a certain inter-subject co-activation pattern. The percentage of positive and negative voxels used for clustering in each frame is 100% and 100%, respectively. Once the inter-subject co-activation pattern is determined, all retained frames are assigned to a certain cluster.

[0108] The k clusters need to be reordered for more intuitive observation and analysis. The k clusters can be reordered in different ways, a common way is to divide them according to the functional network components involved or the cognitive processes involved. Please refer to Figure 4 After the k clusters are reordered, the labels in the state transition matrix and the temporal attributes of the clusters also need to be relabeled to match the reordered states.

[0109] Step 5: Reorder and visualize the inter-subject co-activation states to obtain the visualized transient brain activation patterns. Specifically, it includes:

[0110] Step 51: Perform spatiotemporal analysis on the inter-subject co-activation states to obtain the spatial and temporal characteristics of the inter-subject co-activation states.

[0111] Step 52: According to the spatial and temporal characteristics, obtain the visualized transient brain activation patterns of the individual subjects under natural stimulation.

[0112] Step 5 can be performed by the visualization module 5.

[0113] In one example, the visualization module 5 can specifically be a network visualization tool (such as BrainNet Viewer) for visualizing the inter-subject co-activation states.

[0114] The spatial characteristics of the inter-subject co-activation patterns include the functional network components and cognitive processes involved in the co-activated brain regions, including the auditory and sensorimotor networks related to perception, the language network related to representation, the frontoparietal control network, the dorsal and ventral attention networks related to control and attention, and the default network related to integration. Please refer to Figure 4 .

[0115] The time characteristics of the inter-subject co-activation pattern include: 1) transition vectors, i.e., a probability matrix of transition of each state to several other states; 2) raw counts of different states, for example, state 1 appears 147 times in natural stimulus processing, etc., which can evaluate the occurrence of each state; 3) the possibility of transition from a given ISCAP to another given ISCAP at time t+1 (number of entries), for example, the number of times of transition from state 1 to state 2 can be evaluated, so as to determine which states have stronger transition conditions; 4) the possibility of remaining in the same state from time t to t+1 (resilience), for example, the possibility of remaining in state 1 from time t to t+1 is 20%, so as to determine the duration of each state; 5) betweenness, which calculates all the shortest paths of any two nodes in the network, and if many of these shortest paths pass through a certain node, it is considered that the betweenness of the node is high; 6) the possibility of transition to the state from other states (in-degree), for example, the possibility of transition to state 1 from other states is 10%, so as to determine the transition-in condition of each state; 7) the possibility of transition to other states from the state (out-degree), for example, the possibility of transition to other states from state 1 is 20%, so as to determine the transition-out condition of each state.

[0116] Based on the obtained spatial characteristics and time characteristics of the inter-subject activation pattern, the transient brain activation pattern visualized by the subject in the natural stimulus can be obtained.

[0117] In summary, the inter-subject analysis method is used to analyze the preprocessed fMRI data of the target subject, and the inter-subject consistent signal is obtained, and the fMRI data of the target subject under the natural stimulus induced signal is the inter-subject consistent signal. Compared with the co-activation pattern analysis, the inter-subject consistent signal obtained by the inter-subject co-activation pattern analysis algorithm has a stronger corresponding relationship with the brain function subsystem, the inter-subject consistency is greatly improved, and is significantly related to the processing degree of the natural stimulus of the subject, thereby improving the accuracy and sensitivity of the dynamic co-activation pattern analysis method of the brain under continuous stimulation.

[0118] To achieve the above object, the embodiment of the present application further provides the following scheme:

[0119] A transient co-activation pattern analysis system, please refer to Figure 5 , at least comprising: a time sequence acquisition module 1, a preprocessing module 2, an inter-subject consistent signal acquisition module 3, an inter-subject co-activation state determination module 4 and a visualization module 5.

[0120] The time series acquisition module 1 is configured to acquire fMRI data of natural stimulus processing of a plurality of subjects; the fMRI data of one subject comprises a plurality of fMRI images.

[0121] The description of the time series acquisition module 1 can refer to the foregoing, and will not be repeated here.

[0122] The preprocessing module 2 is configured to preprocess the fMRI data of each subject to obtain preprocessed fMRI data; the preprocessing comprises time correction, head motion correction, structural image segmentation, spatial registration, Gaussian smoothing and filtering.

[0123] The description of the preprocessing module 2 can refer to the foregoing, and will not be repeated here.

[0124] The inter-subject consistent signal acquisition module 3 is configured to calculate the preprocessed fMRI data by using an inter-subject analysis method to obtain stimulus-induced signals; and filter spontaneous activity signals and non-neural signals in the stimulus-induced signals by using a linear regression method to obtain task-induced signals.

[0125] The inter-subject consistent signal acquisition module 3 at least comprises an average calculation unit and a regression calculation unit.

[0126] The average calculation unit is configured to select one target subject from N subjects each time, average the preprocessed fMRI data of the remaining N-1 subjects, and obtain the task-induced signals after traversing the N subjects; the task-induced signals comprise inter-subject consistent signals and inter-subject inconsistent signals.

[0127] The regression calculation unit is configured to perform regression calculation on spontaneous activity signals and non-neural signals in the inter-subject consistent signals by using a linear regression method to obtain the task-induced signals.

[0128] The description of the inter-subject consistent signal acquisition module 3 can refer to the foregoing, and will not be repeated here.

[0129] The inter-subject co-activation state determination module 4 is configured to perform inter-subject co-activation pattern analysis on the inter-subject consistent signals to obtain inter-subject co-activation states.

[0130] The inter-subject co-activation state determination module 4 at least comprises a head motion control unit, a co-activation frame determination unit and a clustering analysis unit.

[0131] The head motion control unit is configured to filter the task-induced signals by using a head motion control method to obtain filtered inter-subject consistent signals.

[0132] The co-activation frame determination unit is configured to compare all the voxel activation data of the whole brain of the subject with a preset activation threshold T according to the filtered inter-subject consistent signals to determine the co-activation frame; or compare all the voxel activation data of the whole brain of the subject with a preset percentage range P of the voxel activation degree according to the filtered inter-subject consistent signals to determine the co-activation frame.

[0133] The co-activation frame determination unit at least includes a data extraction unit and a comparison unit.

[0134] The data extraction unit is configured to obtain the voxel activation data corresponding to the whole brain region;

[0135] The comparison unit is configured to determine that the filtered inter-subject consistent signal corresponding to the voxel activation data is the co-activation frame if the voxel activation data is greater than the preset activation threshold T;

[0136] determine that the filtered inter-subject consistent signal corresponding to the voxel activation data is not the co-activation frame if the voxel activation data is less than or equal to the preset activation threshold T;

[0137] Or,

[0138] determine that the filtered inter-subject consistent signal corresponding to the voxel activation data is the co-activation frame if the voxel activation data is greater than the preset percentage range P of the voxel activation degree;

[0139] determine that the filtered inter-subject consistent signal corresponding to the voxel activation data is not the co-activation frame if the voxel activation data is less than or equal to the preset percentage range P of the voxel activation degree.

[0140] The clustering analysis unit is configured to:

[0141] perform optimal cluster number estimation on the new time series composed of the co-activation frames to obtain the optimal cluster number k;

[0142] According to the optimal cluster number k, the co-activation frames are assigned to the corresponding brain co-activation patterns frame by frame;

[0143] reorder the k clusters to obtain the inter-subject co-activation state.

[0144] The description of the inter-subject co-activation state determination module 4 can refer to the foregoing description and will not be repeated here.

[0145] The visualization module 5 is configured to visually display the inter-subject co-activation state to obtain a visual transient brain activation pattern.

[0146] The visualization module 5 at least includes a spatiotemporal feature extraction unit and a display.

[0147] The space-time feature extraction unit is configured to perform space-time analysis on the inter-subject co-activation state to obtain spatial features and temporal features of the inter-subject co-activation state.

[0148] The display unit is configured to obtain a transient brain activation pattern of the individual subject in the natural stimulation according to the spatial features and the temporal features.

[0149] The description of the visualization module 5 can refer to the foregoing description, and will not be repeated here.

[0150] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between the embodiments can be referred to each other.

[0151] The principles and implementation manners of the embodiments of the present application are described by using specific examples. The above description of the embodiments is only used to help understand the method and the core idea of the embodiments of the present application. Meanwhile, for the general skilled in the art, the specific implementation manners and application ranges can be changed according to the idea of the embodiments of the present application. In conclusion, the content of the specification should not be understood as the limitation of the embodiments of the present application.

Claims

1. A transient co-activation mode analysis method, characterized in that, include: fMRI data on natural stimulus processing were acquired from multiple subjects; The fMRI data for one subject includes multiple fMRI images; The fMRI data of each subject were preprocessed to obtain preprocessed fMRI data; the preprocessing included: time correction, head motion correction, structural image segmentation, spatial registration, Gaussian smoothing and filtering; The preprocessed fMRI data were analyzed using a between-subjects analysis method to obtain stimulus-evoked signals. Spontaneous activity signals and non-neural signals in the stimulus-evoked signals were then filtered out using a linear regression method to obtain task-evoked signals. The task-evoked signals were analyzed frame-by-frame using a co-activation pattern analysis method to obtain brain co-activation patterns. K-means clustering was used to identify and extract recurring inter-subject co-activation states within these patterns, specifically including: The task-induced signals were filtered using a head movement control method to obtain filtered inter-subject consistent signals. Based on the filtered inter-subject consistent signal, the voxel activation data of the whole brain of the subject is compared with a preset activation threshold T to determine co-activated frames; or, based on the filtered inter-subject consistent signal, the voxel activation data of the whole brain of the subject is compared with a preset percentage range P of voxel activation level to determine co-activated frames. The co-activated frames are combined into a new time series for optimal cluster number estimation to obtain the optimal cluster number k; Based on the optimal clustering number k, the co-activation frames are assigned frame by frame to the corresponding brain co-activation modes. The k clusters are reordered to obtain the inter-subject co-activation state; The co-activation states among the subjects were reordered and visualized to obtain a visualized transient brain activation pattern.

2. The transient co-activation mode analysis method according to claim 1, characterized in that, The calculation of the stimulus-evoked signal from the preprocessed fMRI data using the inter-subject analysis method specifically includes: Each time, one target subject is selected from N subjects, and the average of the processed fMRI data of the remaining N-1 subjects is calculated. After traversing through N subjects, the inter-subject consistent signal is obtained. The task-induced signal was obtained by performing regression calculations on the spontaneous activity signal and non-neural signal in the inter-subject consistent signal using a linear regression method.

3. The transient co-activation mode analysis method according to claim 1, characterized in that, The co-activation states among the subjects were reordered and visualized to obtain a visualized transient brain activation pattern, which specifically includes: Spatiotemporal analysis was performed on the inter-subject co-activation state to obtain the spatial and temporal characteristics of the inter-subject co-activation state. Based on the spatial and temporal characteristics, the transient brain activation patterns of individual subjects in response to natural stimuli were obtained.

4. The transient co-activation mode analysis method according to claim 1, characterized in that, Based on the filtered inter-subject consistent signal, the voxel activation data of the whole brain of the subject are compared with the preset activation threshold T to determine the co-activated frames; Alternatively, based on the filtered inter-subject consistent signal, the voxel activation data of the subject's whole brain are compared with the preset percentage range P of voxel activation levels to determine the co-activated frames, specifically including: Obtain voxel activation data corresponding to all brain regions; If the voxel activation data is greater than the preset activation threshold T, then the filtered inter-subject consistent signal corresponding to the voxel activation data is determined to be the co-activation frame. If the voxel activation data is less than or equal to the preset activation threshold T, then the filtered inter-subject consistent signal corresponding to the voxel activation data is determined to be a co-activation frame. or, If the voxel activation data is greater than the preset percentage range P of the voxel activation level, then the filtered inter-subject consistent signal corresponding to the voxel activation data is determined to be the co-activation frame. If the voxel activation data is less than or equal to the percentage range P of the preset voxel activation level, then the filtered inter-subject consistent signal corresponding to the voxel activation data is determined not to be the co-activation frame.

5. A transient co-activation mode analysis system, characterized in that, include: The time-series acquisition module is used to acquire fMRI data on natural stimulus processing from multiple subjects; The fMRI data for one subject includes multiple fMRI images; The preprocessing module is used to preprocess the fMRI data of each subject to obtain preprocessed fMRI data; the preprocessing includes: time correction, head motion correction, structural image segmentation, spatial registration, Gaussian smoothing and filtering. The inter-subject consistency signal acquisition module is used to calculate the preprocessed fMRI data using inter-subject analysis methods to obtain stimulus-evoked signals; and to filter out spontaneous activity signals and non-neural signals from the stimulus-evoked signals using linear regression methods to obtain task-evoked signals. A co-activation state determination module is used to analyze the task-evoked signals frame-by-frame using a co-activation pattern analysis method to obtain brain co-activation patterns; and to identify and extract recurring inter-subject co-activation states from the brain co-activation patterns using a K-means clustering method. The co-activation state determination module includes: The head movement control unit is used to filter the task-induced signals using a head movement control method to obtain filtered inter-subject consistent signals. The co-activation frame determination unit is used to determine co-activation frames by comparing all voxel activation data of the subject's whole brain with a preset activation threshold T based on the filtered inter-subject consistent signal; or, by comparing all voxel activation data of the subject's whole brain with a preset percentage range P of voxel activation level based on the filtered inter-subject consistent signal. Cluster analysis unit, used for: The co-activated frames are combined into a new time series for optimal cluster number estimation to obtain the optimal cluster number k; Based on the optimal clustering number k, the co-activation frames are assigned frame by frame to the corresponding brain co-activation modes. The k clusters are reordered to obtain the inter-subject co-activation state; The visualization module is used to reorder and visualize the co-activation states among the subjects to obtain a visualized transient brain activation pattern.

6. The transient co-activation mode analysis system according to claim 5, characterized in that, The inter-subject consistency signal acquisition module includes: The averaging unit is used to select one target subject from N subjects each time, average the processed fMRI data of the remaining N-1 subjects, and obtain the inter-subject consistent signal after traversing N subjects. The regression calculation unit is used to perform regression calculations on the spontaneous activity signals and non-neural signals in the inter-subject consistent signals using a linear regression method to obtain the task-induced signals.

7. The transient co-activation mode analysis system according to claim 5, characterized in that, The visualization module includes: The spatiotemporal feature extraction unit is used to perform spatiotemporal analysis on the inter-subject co-activation state to obtain the spatial and temporal features of the inter-subject co-activation state. The display unit is used to obtain the transient brain activation patterns of an individual subject in response to natural stimuli, based on the spatial and temporal characteristics.

8. The transient co-activation mode analysis system according to claim 5, characterized in that, The co-activation frame determination unit includes: The data extraction unit is used to acquire voxel activation data corresponding to whole brain regions. Comparison unit, used for: If the voxel activation data is greater than the preset activation threshold T, then the filtered inter-subject consistent signal corresponding to the voxel activation data is determined to be the co-activation frame. If the voxel activation data is less than or equal to the preset activation threshold T, then the filtered inter-subject consistent signal corresponding to the voxel activation data is determined to be a co-activation frame. or, If the voxel activation data is greater than the preset percentage range P of the voxel activation level, then the filtered inter-subject consistent signal corresponding to the voxel activation data is determined to be the co-activation frame. If the voxel activation data is less than or equal to the percentage range P of the preset voxel activation level, then the filtered inter-subject consistent signal corresponding to the voxel activation data is determined not to be the co-activation frame.

Images