A computational method for integrating task-evoked and intrinsic spontaneous brain function activities
By integrating task-induced and intrinsic spontaneous brain activity at the individual level, and utilizing low-dimensional state space and Euclidean distance calculation methods, this method solves the problem of the ineffective integration of task-state and resting-state brain activity in existing technologies, and achieves a deeper understanding of the prediction of cognitive performance and the mechanisms of brain function.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- EAST CHINA NORMAL UNIV
- Filing Date
- 2025-12-22
- Publication Date
- 2026-07-21
AI Technical Summary
Existing technologies have failed to effectively integrate task-induced and intrinsic spontaneous brain functional activities, making it difficult to quantify the relationship between the two activity modes at the individual level. Furthermore, traditional methods are inadequate for assessing critical brain states and functional representation mechanisms.
By constructing an individualized low-dimensional state space, task-induced activation and spontaneous neural avalanche activity are projected into the same space, and Euclidean distance is calculated. A multiple linear regression model is then used to predict cognitive performance.
This study effectively integrates task-oriented and spontaneous brain activity, revealing the relationship between the cascade propagation pattern of spontaneous brain activity and the brain response pattern in task-oriented mode, and providing new perspectives and tools for understanding brain function mechanisms.
Smart Images
Figure CN121747829B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of functional magnetic resonance imaging and cognitive neuroscience research, specifically a computational method that integrates task-induced and intrinsic spontaneous brain functional activities. Background Technology
[0002] The human brain is one of the most complex systems in nature. As the core regulatory center, it supports humans in flexibly adapting to the environment and performing various cognitive and behavioral tasks. In-depth exploration of the brain's functional operation is not only crucial for unlocking its potential and maintaining brain health, but has also become a cutting-edge topic of common interest across multiple interdisciplinary fields. Among numerous research paradigms, spontaneous neural activity of the brain is considered a vital window into understanding its functional organization. Task-based functional magnetic resonance imaging (fMRI) can reveal the activation patterns of the brain during specific cognitive tasks, while resting-state fMRI reflects the characteristics of spontaneous brain activity in a task-free state. Previous studies have largely separated task-based and resting-state brain functional activity, failing to effectively integrate the relationship between the two to comprehensively understand the mechanisms of brain function.
[0003] In recent years, some studies have attempted to jointly analyze task-state and resting-state data, for example, by calculating the correlation between task activation patterns and resting-state functional connectivity, or by using resting-state functional connectivity to predict the responses of task-state brain regions. However, these methods do not take into account the cascading propagation patterns of spontaneous brain activity.
[0004] Neural avalanche, as an important phenomenon reflecting the critical state of the brain, offers a new perspective on understanding spontaneous brain activity due to its spatiotemporal dynamics in the resting state. Previous studies have shown that the size and duration distribution of neural avalanches follow a power-law distribution, suggesting that the brain is in a critical state, which is considered to be related to optimal information transmission and processing capabilities. The traditional neural avalanche framework defines spontaneous avalanches as continuous activity frames separated by inactive time. However, this definition has significant limitations: multiple spatiotemporally distinguishable independent avalanche events may be merged into a single avalanche, thus ignoring their spatial and temporal distinguishability and affecting the assessment of critical states of brain activity. Furthermore, limited by the recording time of macroscopic neuroimaging (such as functional magnetic resonance imaging), traditional methods often struggle to effectively assess brain criticality at the individual level, and its functional representation mechanisms remain unclear.
[0005] Therefore, there is an urgent need in this field for a computational method that can effectively integrate task-induced and intrinsic spontaneous brain functional activities, quantify the relationship between the two activity modes at the individual level, and provide new ways to understand brain functional mechanisms and their connection with cognitive performance. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a computational method that integrates task-induced and intrinsic spontaneous brain functional activities. By constructing an individualized low-dimensional state space, task-induced activation and spontaneous neural avalanche activity are projected into the same space, and the Euclidean distance between the two is calculated, thereby predicting an individual's cognitive performance.
[0007] To achieve the above objectives, the present invention adopts the following technical solution:
[0008] A computational method integrating task-induced and intrinsic spontaneous brain functional activities includes the following steps:
[0009] Step 1: Preprocess the resting-state and task-state functional magnetic resonance imaging data;
[0010] Step 2: Based on resting-state functional magnetic resonance imaging data, identify individual large-scale spontaneous neural avalanche activity patterns with spatial continuity;
[0011] Step 3: Perform principal component analysis on the individual's resting-state functional magnetic resonance imaging data, and extract the three principal components with the top three cumulative explanatory power to construct a low-dimensional state space;
[0012] Step 4: Project the individual's task-oriented brain activation pattern and spontaneous neural avalanche pattern, i.e., the temporal collapse pattern, onto the aforementioned low-dimensional state space.
[0013] Step 5: Calculate the average Euclidean distance between the task-state brain activation projection point and all spontaneous neural avalanche pattern projection points; where the Euclidean distance between two points is defined as... :
[0014] =
[0015] In the formula, The set of projection points for task-oriented brain activation is: , Indicates the first The coordinates of a spontaneous neural avalanche in a low-dimensional state space, where M represents the number of projection points of the spontaneous neural avalanche mode;
[0016] Mean Euclidean distance is defined as
[0017] .
[0018] Step 6: Analyze the relationship between the average Euclidean distance and individual task performance using a regression model to assess its predictive power for cognitive abilities.
[0019] The regression model is a multiple linear regression model, and its form is as follows:
[0020] In the formula, Indicates a score for a specific cognitive task. This represents a distance metric for a specific cognitive task. , and These represent the regression terms: age, gender, and head movement, respectively. This is the error term.
[0021] Furthermore, the preprocessing described in step one involves removing the first 10 time points from the resting-state data and regressing motor, cerebrospinal fluid, and white matter signals. After smoothing with a 4mm Gaussian kernel, the resting-state brain data is divided into 400 brain regions according to the Schaefer partitioning template. Based on the task-oriented functional magnetic resonance imaging data, the brain region activation vectors of individuals performing cognitive tasks are extracted using the general linear model GLM.
[0022] Step two, which involves identifying spontaneous neural avalanche patterns, specifically includes: Z-score normalization of the blood-oxygen-level-dependent signal (BOLD signal) for each brain region; identifying BOLD signal time points exceeding a preset threshold (+1.7SD) as spike events and binarizing the BOLD signal into "0" and "1"; determining that a time point is a spike event is necessary because the signal value at that time point is a local maximum of the signal values at the previous and next time points; based on the brain region template, determining whether there is a neighbor relationship between brain regions binarized to "1" in each time frame, and identifying the combination of brain regions with neighbor relationships as a cluster; defining the set of clusters with cascading propagation relationships within consecutive time frames as a neural avalanche; the size of the avalanche is the number of brain regions involved, and the duration of the avalanche is the number of consecutive time frames.
[0023] Step 3, which involves constructing a low-dimensional state space, specifically involves: standardizing and merging the resting-state time series of the individuals; performing principal component analysis on the merged time series data; and extracting the three principal components with the highest cumulative explanatory power to construct a low-dimensional state space.
[0024] The beneficial effects of this invention are as follows: by constructing an individualized low-dimensional state space, it unifies the representation of task-induced brain functional activities and intrinsic spontaneous neural avalanche activities, achieving effective integration of these two basic brain functional activities. In particular, it can reveal the relationship between the cascade propagation pattern of spontaneous brain activity and the task-oriented brain response pattern. The Euclidean distance is used to quantify the relationship between the projection points of task-oriented brain activation and the projection points of the spontaneous neural avalanche pattern, providing a novel measure of brain functional integration. Regression analysis verifies the predictive ability of this distance indicator for individual cognitive performance, providing a new perspective and tool for understanding brain functional mechanisms.
[0025] The method of this invention has been verified on real datasets and has good application prospects and promotion value. Attached Figure Description
[0026] Figure 1 This is a schematic diagram of the projection of spontaneous neural avalanche patterns and different cognitive task brain activation patterns in the individual's resting state low-dimensional state space in an embodiment of the present invention.
[0027] Figure 2 This is a schematic diagram illustrating the relationship between the average Euclidean distance between the spontaneous neural avalanche pattern projection point and the brain activation projection point for working memory tasks and task reaction time in an embodiment of the present invention.
[0028] Figure 3 This is a schematic diagram illustrating the relationship between the average Euclidean distance between the projection point of the spontaneous neural avalanche pattern and the brain activation projection point of the reasoning task in an embodiment of the present invention and the task accuracy. Detailed Implementation
[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will now be described in further detail with reference to the accompanying drawings and embodiments.
[0030] The present invention proposes a computational method that integrates task-induced and intrinsic spontaneous brain functional activities, which mainly includes the following steps:
[0031] Step 1: Preprocess the resting-state functional magnetic resonance imaging data, and calculate brain activation for the task-oriented functional imaging data according to different task paradigms;
[0032] Step 2: Based on resting-state functional magnetic resonance imaging data, identify large-scale spontaneous neural avalanche activity patterns with spatiotemporal continuity;
[0033] Step 3: Perform principal component analysis on the individual's resting-state functional magnetic resonance imaging data, and extract the first three principal components to construct a low-dimensional state space;
[0034] Step 4: Project the individual's task-oriented brain activation pattern and spontaneous neural avalanche pattern onto the three-dimensional state space described above.
[0035] Step 5: Calculate the average Euclidean distance between the task-state brain activation projection point and the spontaneous neural avalanche pattern projection point;
[0036] Step 6: Analyze the relationship between the average Euclidean distance and individual task performance using a regression model to assess its predictive power for cognitive abilities.
[0037] In step one, the resting-state imaging data of each subject were standardized and preprocessed. Based on the Schaefer partitioning template, the resting-state brain data was divided into 400 brain regions of interest, and the corresponding time series were obtained. Using task-oriented functional magnetic resonance imaging data, a general linear model was used to extract brain region activations when individuals performed different cognitive task paradigms: emotional tasks, language tasks, game-theoretic tasks, reasoning tasks, working memory tasks, and social tasks. These activations were represented as vectors corresponding to the number of brain regions.
[0038] In step two, the time series obtained in step one is further processed. After z-score normalization of the BOLD signal for each brain region, time points with BOLD signals exceeding a threshold (+1.7SD) are identified as spike events, and the BOLD signals are then binarized into "0" and "1". A necessary condition for a time point to be identified as a spike event is that its signal value must be higher than the signal values of the previous and next time points, that is, the signal value of that time point should be a local maximum within three consecutive time points. Regarding the definition of the neural avalanche cascade propagation mode, it is first determined whether there is a neighbor relationship between brain regions that are binarized to "1" in each frame of the time series of 400 brain regions. The determination method is: whether there is a neighbor relationship between the edge voxels of two brain regions is determined by whether they are connected on at least one face. The voxel is a cube with 6 faces. A cluster is defined as a group of brain regions that are neighbors. That is, each frame contains only a single brain region marked "1" or a cluster consisting of two or more neighboring brain regions marked "1". We define an avalanche as the set of clusters that have a cascading propagation relationship within consecutive frames. The size of the avalanche is the number of brain regions involved, and the duration of the avalanche is the number of consecutive time frames.
[0039] In step three, the resting-state time series is first standardized. For each subject sample, the standardized time series from four trials (1190 time points per trial, totaling 4760 time points) are first merged, and then transposed into session trials.
[0040]
[0041]
[0042] The merged time series were globally standardized, and based on this, principal component analysis was applied to reduce the dimensionality of all resting-state brain activity patterns. The first three principal components were retained to construct a three-dimensional state space to characterize the main directions of variation in spontaneous brain activity.
[0043] In step four, each subject's spontaneous neural avalanche pattern point was determined by different avalanche activity patterns. Transpose and merge into Projecting all avalanche patterns onto a low-dimensional state space, such as... Figure 1 The projection points of all spontaneous neural avalanche patterns and brain activation patterns for different cognitive tasks in the individual's resting-state low-dimensional state space are shown, including working memory tasks, reasoning tasks, game-theoretic tasks, social tasks, emotional tasks, and language tasks.
[0044] In step five, the average Euclidean distance between the task-state brain activation projection point and all spontaneous neural avalanche pattern projection points is calculated; whereby the Euclidean distance between two points is defined as... :
[0045] =
[0046] In the formula, The set of projection points for task-oriented brain activation is: , Indicates the first The coordinates of a spontaneous neural avalanche pattern in a low-dimensional state space. The average Euclidean distance is defined as...
[0047]
[0048] In step six, the regression model controls for covariates such as age, gender, and head movements to test the predictive power of the mean Euclidean distance on individual task performance (such as reaction time and accuracy). The regression model is as follows:
[0049]
[0050] In the formula, Indicates a score for a specific cognitive task. This represents a distance metric for a specific cognitive task. , and These represent the regression terms: age, gender, and head movement, respectively. This is the error term.
[0051] By representing spontaneous resting-state brain activity as avalanche patterns with spatiotemporal structure, and then projecting these patterns into an individual's low-dimensional state space, distance metrics can be calculated to predict performance on different cognitive tasks. Experiments were conducted on the real-world, publicly available Human Connectome Dataset (HCP) to validate the effectiveness of this method.
[0052] The technical solution of this invention will be further described in detail below with reference to application examples:
[0053] A specific example of this invention illustrates the evaluation of the effectiveness of the proposed method on a publicly available fMRI functional magnetic resonance imaging dataset. Table 1 provides basic information about these datasets.
[0054] Table 1: Statistical information of subjects in the dataset
[0055]
[0056] The fMRI data used in the experiment were obtained from the Human Connectome Project (HCP, https: / / db.humanconnectome.org / data / projects / HCP_1200) and minimally preprocessed. Subsequently, the resting and task-oriented fMRI data were further preprocessed using the SPM12 toolkit (http: / / www.fil.ion.ucl.ac.uk / spm) and REST software (https: / / www.nitrc.org / projects / rest). For the resting fMRI data, the first 10 time points of each session were discarded to account for the influence of participants' adaptation to the experimental environment. Linear regression was used to remove head movements, cerebrospinal fluid, and white matter signals. Then, linear trend removal, bandpass filtering (0.01–0.08 Hz), and spatial smoothing were performed. The brain was divided into 400 regions using the Schaefer400 template, and time series were extracted. For the task-oriented fMRI data, linear regression was used to remove head movements, followed by spatial smoothing using a 4 mm Gaussian filter. For each task, a general linear model (GLM) is used to obtain a voxel-level activation map of the whole brain. Then, based on the Schaefer400 template, the average activation of each brain region is calculated, generating a 400 × 1 activation vector. This dataset contains 6 tasks: emotion task, reasoning task, social task, game task, language task, and working memory task, thus generating 6 task-state activation vectors.
[0057] Spontaneous neural avalanche patterns were identified based on the spatiotemporal cascade propagation rules of resting-state brain activity. Principal component analysis was performed on all resting-state frame activity patterns, retaining the first three principal components to construct a low-dimensional state space. Individual spontaneous neural avalanche patterns and task-state brain activation patterns were projected one by one into this low-dimensional state space. The average Euclidean distance between the projection points of all individual spontaneous neural avalanche patterns and the projection points of each task-state brain activation pattern was calculated. A linear regression model of distance and corresponding cognitive task performance was constructed, and the effects of age, head movement, and gender were regressed to predict cognitive task performance.
[0058] Figure 2 The mean Euclidean distance between the spontaneous neural avalanche pattern projection point and the brain activation projection point for the working memory task can significantly predict the reaction time of the working memory task; the longer the distance, the longer the working memory reaction time.
[0059] Figure 3 The mean Euclidean distance between the projection point of the spontaneous neural avalanche pattern and the brain activation projection point of the reasoning task can significantly predict the accuracy of the reasoning memory task; the shorter the distance, the higher the accuracy of the reasoning task.
[0060] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.
Claims
1. A computational method integrating task-induced and intrinsic spontaneous brain functional activities, characterized in that, Includes the following steps: Step 1: Preprocess the resting-state and task-state functional magnetic resonance imaging data; Step 2: Based on resting-state functional magnetic resonance imaging data, identify large-scale spontaneous neural avalanche activity patterns with spatial continuity in individuals; Step 3: Perform principal component analysis on the individual's resting-state functional magnetic resonance imaging data to construct a low-dimensional state space; Step 4: Project the individual's task-oriented brain activation pattern and spontaneous neural avalanche pattern onto the low-dimensional state space, respectively; Step 5: Calculate the average Euclidean distance between the task-state activation projection point and all spontaneous neural avalanche pattern projection points; Step Six: Analyze the relationship between the mean Euclidean distance and individual task performance using a regression model to assess its predictive power for cognitive abilities.
2. The computational method for integrating task-induced and intrinsic spontaneous brain functional activities as described in claim 1, characterized in that, The preprocessing in step one is as follows: the resting-state brain data is divided into 400 brain regions of interest according to the Schaefer-Yeo7 map; the task-oriented functional magnetic resonance imaging data is preprocessed, and the brain region activation vectors of individuals when performing cognitive tasks are extracted using the general linear model GLM.
3. The computational method for integrating task-induced and intrinsic spontaneous brain functional activities as described in claim 1, characterized in that, Step two, identifying spontaneous neural avalanche activity patterns, includes: Z-score normalization of the oxygenation level dependent signal BOLD for each brain region; identifying BOLD signal time points exceeding a preset threshold + 1.7 SD as spike events, and binarizing the BOLD signal into "0" and "1"; determining, based on brain region templates, whether there are neighbor relationships between brain regions binarized to "1" in each time frame, and identifying combinations of brain regions with neighbor relationships as a cluster; defining a set of clusters with cascading propagation relationships within consecutive time frames as a neural avalanche; the size of the avalanche is the number of brain regions involved, and the duration of the avalanche is the number of consecutive time frames.
4. The computational method for integrating task-induced and intrinsic spontaneous brain functional activities as described in claim 3, characterized in that, A necessary condition for determining a point in time as a peak event is that the signal value at that point is a local maximum of the signal values at the previous and next time points.
5. The computational method for integrating task-induced and intrinsic spontaneous brain functional activities as described in claim 1, characterized in that, In step three, constructing a low-dimensional state space specifically involves: standardizing and merging the resting-state time series of individuals; performing principal component analysis on the merged time series data, and extracting the three principal components with the highest cumulative explanatory power to construct a low-dimensional state space.
6. The computational method for integrating task-induced and intrinsic spontaneous brain functional activities as described in claim 1, characterized in that, Step five involves calculating the average Euclidean distance between the task-state brain activation projection point and all spontaneous neural avalanche pattern projection points; where the Euclidean distance between two points is defined as... : = ； In the formula, The set of projection points for task-oriented brain activation is: , Indicates the first The coordinates of a spontaneous neural avalanche in a low-dimensional state space; the average Euclidean distance is defined as... : 。 7. The computational method for integrating task-induced and intrinsic spontaneous brain functional activities as described in claim 1, characterized in that, The regression model described in step six is a multiple linear regression model, which has the following form: ; In the formula, Indicates a score for a specific cognitive task. This represents a distance metric for a specific cognitive task. , and These represent the regression terms: age, gender, and head movement, respectively. This is the error term.