A method and system for formulating individualized control strategy for task-based functional magnetic resonance imaging

Abstract

The present invention discloses a method and system for formulating an individualized regulation strategy for task-state functional magnetic resonance imaging, and relates to the technical field of individualized regulation strategy formulation. The present invention utilizes functional magnetic resonance imaging with high temporal resolution to obtain the task activation state of the individual subject, extract the corresponding brain network and brain region, construct an individual effective connection model, and use the sum of the weighted brain region activity states as the optimization target, and provide an individualized intervention plan based on the simulation results of the generated model. When group studies of specific diseases show that there are multiple possible target brain regions and intervention strategies, the present invention can provide an objective, feasible, and reliable method for formulating individualized regulation plans, and can achieve comprehensive optimization of the overall characteristics of the brain network.

Description

Technical Field

[0001] The present invention relates to the technical field of individualized regulation strategy formulation, and in particular to a method and system for formulating an individualized regulation strategy for task-state functional magnetic resonance imaging. Background Art

[0002] In recent years, repetitive transcranial magnetic stimulation (rTMS), as a new non-invasive treatment, has gradually been used for functional recovery of cerebrovascular disease, and its safety and effectiveness have been verified. Some applications have been included in the category A recommendations of the guidelines, and can be used as an effective supplement to traditional rehabilitation methods. However, the difficulty in its clinical application lies in the fact that its intervention effect is affected by the stimulation frequency, especially the selection of stimulation targets. In existing clinical applications, general plans are often given based on the disease mechanisms obtained from group studies or the experience of evidence-based medicine. However, for motor or cognitive dysfunction after cerebrovascular disease, the variability of neural circuits between individuals leads to differences in the intervention effect of rTMS in the population. The intervention strategies driven by evidence-based medicine and group mechanisms are not universal, and even the various strategies given are contradictory and controversial. Taking post-stroke movement disorders as an example, rTMS inhibits the healthy side, promotes the affected side, and promotes the healthy side, which are all possible effective means, so there is a lack of a method for formulating individualized intervention strategies.

[0003] The current strategy for individualized transcranial magnetic stimulation is mainly to select voxel-level targets through functional connectivity analysis. When neuroimaging is not used, early studies proposed to use the motor hot spots measured by electrophysiological measurements as a benchmark, or to combine the click position of the EEG cap to empirically locate the brain region; when combined with imaging, specific anatomical structures can generally be extracted as representations of brain regions. In recent years, studies have proposed that based on task-state functional magnetic resonance imaging (fMRI), active surface brain regions can be used as regions of interest, and the functional connections between each voxel and the deep activated brain region can be calculated, and the voxels close to the scalp with the strongest connection can be selected as targets. Based on resting-state fMRI, the brain regions with enhanced spontaneous activity in the whole brain can be used as regions of interest for subsequent processing through inter-group comparison; or the target brain region of interest can be directly given, and the individualized target can be selected based on the strength of its functional connection with the deep brain region corresponding to the damaged cognition. In short, the existing methods first determine a broad region of interest based on some analysis results or prior knowledge, and then perform functional connectivity analysis on all voxels in the region to determine the target.

[0004] Based on the above description, the optimization method of the existing method has shortcomings: functional connectivity analysis is performed on all voxels of interest at the same time to select targets, without considering the range of effects of rTMS intervention. rTMS produces an effect on a brain region (area) rather than one voxel. The optimal target position obtained by the existing method may have significantly effective voxels, but it may not be optimal from the perspective of brain regions.

[0005] The selection basis of existing methods cannot guarantee the optimization purpose: using undirected functional connections as the basis for target selection, without considering the directionality of brain connections, it is impossible to ensure that the intervention effect is transmitted to deep brain regions. The connection between brain regions is actually a directed connection (effective connection), and functional connectivity analysis regards all brain region effects as bidirectional and non-causal. If the deep brain region is only unidirectionally connected to the surface, the existing methods cannot achieve the effect of affecting the deep through surface stimulation.

[0006] Existing methods cannot provide individualized strategy guidance beyond the target: they only compare the functional connectivity strength of the region of interest and cannot provide additional information; they select a single intervention frequency by default and then look for possible brain regions. Existing inventions generally first assume that the abnormality caused by a disease is an abnormal increase (or decrease) in brain region activity, and then analyze and obtain representative voxels that show the corresponding state for voxel selection. Since no other meaningful physiological information is provided, the mechanism of simultaneous changes in the activities of multiple brain regions and their significance to rehabilitation are unclear, and it is impossible to select the best of multiple possible intervention frequency options. Summary of the invention

[0007] In order to solve the above problems existing in the prior art, the present invention provides a method and system for formulating an individualized control strategy for task-state functional magnetic resonance imaging.

[0008] To achieve the above object, the present invention provides the following solutions:

[0009] A method for formulating an individualized control strategy for task-based functional magnetic resonance imaging, comprising:

[0010] Acquire multi-band task-state functional magnetic resonance imaging of the subjects;

[0011] Constructing a functional activation map of the subject based on the multi-band task-state functional magnetic resonance imaging;

[0012] Extracting individual brain region time series based on the functional activation map;

[0013] Perform dynamic causal modeling and fitting based on the individual brain region time series to obtain a dynamic causal model;

[0014] Performing Bayesian model selection on the dynamic causal model to obtain an individual optimal model;

[0015] Constructing a generative model of a dynamic causal model based on the individual optimal model and the individualized model parameters corresponding to the individual optimal model;

[0016] Mapping external regulation into parameters of the generative model to simulate the dynamic activities of brain networks under different regulation strategies;

[0017] A multivariate linear model was constructed with the motor scale score at the group level as the dependent variable and the activation of each brain region in the functional activation map as the independent variable;

[0018] Based on the output value of the multivariate linear model as the optimization target, the dynamic activities of the brain network under different control strategies are compared, and the control strategy that maximizes the motor function output value is determined as the individual optimal strategy.

[0019] Optionally, constructing a functional activation map of the subject based on the multi-band task-state functional magnetic resonance imaging specifically includes:

[0020] Preprocessing the multi-band task-state functional magnetic resonance images;

[0021] A linear model is established; the parameters of the linear model include: an experimental task design matrix, a time series extracted in a standard space, and a head motion parameter; the time series and the head motion parameter are extracted from the preprocessed multi-band task-state functional magnetic resonance imaging; the time series includes the white matter time series, cerebrospinal fluid time series, and global signal time series of the individual subject;

[0022] Using the linear model, a single sample test of the task state compared with the resting state is defined on the experimental task design matrix to obtain activation in the brain region;

[0023] The functional activation map is generated based on the brain area activation.

[0024] Optionally, extracting individual brain region time series based on the functional activation map specifically includes:

[0025] Taking the region of interest as the group-level brain region coordinates, taking the group-level brain region coordinates as the sphere center and taking the first preset value as the radius, determine a first spherical region;

[0026] The peak points of the motor task activation in the functional activation map in the first spherical region are used as the coordinates of the individual brain regions;

[0027] Determine a second spherical area with the coordinates of the individual brain region as the center of the sphere and a second preset value as the radius;

[0028] The first feature vectors of all voxels in the second spherical region are used as the individual brain region time series.

[0029] Optionally, performing dynamic causal modeling and fitting based on the individual brain region time series to obtain a dynamic causal model specifically includes:

[0030] Using a Bayesian approach to fit the individual brain region time series data within the framework of an initial dynamic causal model; the initial dynamic causal model is used to characterize the emergent interhemispheric connectivity patterns;

[0031] During the data fitting process, the effective connection parameters are adjusted so that the error between the data generated by the initial dynamic causal model and the true value meets the set conditions, and the effective connection parameters between brain regions under each initial dynamic causal model are obtained to form the dynamic causal model.

[0032] Optionally, performing Bayesian model selection on the dynamic causal model to obtain an individual optimal model specifically includes:

[0033] Fixed-effects analysis was used to determine the level of evidence for the dynamic causal model;

[0034] The dynamic causal model with the highest level of evidence is taken as the optimal model; and the effective connection parameters corresponding to the optimal model are taken as the individualized model parameters.

[0035] Optionally, external regulation is mapped to parameters of the generative model to simulate the dynamic activities of the brain network under different regulation strategies, including:

[0036] The effect of high-frequency transcranial stimulation is used to weaken the self-inhibition strength of the brain areas in the generative model, while the effect of low-frequency transcranial stimulation is simplified to enhance the self-inhibition strength of the brain areas in the generative model. The results of all brain areas in the target brain network under different regulation strategies are simulated to obtain the dynamic activities of the brain network under different regulation strategies.

[0037] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0038] The method for formulating task-state functional magnetic resonance individualized regulation strategies provided by the present invention uses high-temporal-resolution functional magnetic resonance images to obtain the task activation state of the individual subject, extract the corresponding network and brain region, construct an individual effective connection model, and use the sum of the weighted brain region activity states as the optimization target. Based on the simulation results of the generative model, an individualized intervention plan is given. When group studies of specific diseases show that there are multiple possible target brain regions and intervention strategies, the present invention can provide an objective, feasible, and reliable method for formulating individualized regulation plans, which can achieve comprehensive optimization of the overall characteristics of the brain network.

[0039] In addition, the present invention also provides a task-state functional magnetic resonance imaging individualized regulation strategy formulation system, the system comprising:

[0040] An image collector, used for collecting multi-band task-state functional magnetic resonance images of the subject;

[0041] Memory for storing computer software programs;

[0042] A processor is connected to the image collector and the memory, respectively, and is used to call and execute the computer software program to implement the above-mentioned task-state functional magnetic resonance individualized regulation strategy formulation method based on the multi-band task-state functional magnetic resonance images of the subject acquired by the image collector.

[0043] Optionally, the processor includes:

[0044] An image acquisition unit, used for acquiring multi-band task-state functional magnetic resonance images of the subject;

[0045] A map construction unit, used for constructing a functional activation map of the subject based on the multi-band task-state functional magnetic resonance imaging;

[0046] A sequence extraction unit, used for extracting individual brain region time series based on the functional activation map;

[0047] A model fitting unit, used for performing dynamic causal modeling fitting based on the individual brain region time series to obtain a dynamic causal model;

[0048] A model selection unit, used for performing Bayesian model selection on the dynamic causal model to obtain an individual optimal model;

[0049] A first model building unit, configured to build a generation model of a dynamic causal model based on the individual optimal model and the individualized model parameters corresponding to the individual optimal model;

[0050] The activity simulation unit is used to map external regulation into parameters of the generative model to simulate the dynamic activities of the brain network under different regulation strategies;

[0051] The second model building unit is used to build a multivariate linear model with the motor scale score at the group level as the dependent variable and the activation of each brain region in the functional activation map as the independent variable;

[0052] A strategy determination unit is used to compare the dynamic activities of the brain network under different control strategies based on the output value of the multivariate linear model as the optimization target, and determine the control strategy that maximizes the motor function output value as the individual optimal strategy.

[0053] Optionally, the memory is a computer-readable storage medium.

[0054] Since the technical effect achieved by the system provided by the present invention is the same as the technical effect achieved by the above-mentioned method for formulating individualized control strategies of task-state functional magnetic resonance imaging, it will not be described in detail here. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. 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 creative work.

[0056] Figure 1 A flow chart of a method for formulating an individualized control strategy for task-based functional magnetic resonance imaging provided in Example 1 of the present invention;

[0057] Figure 2 This is a main workflow diagram provided for Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0059] The purpose of the present invention is to provide a method and system for formulating an individualized control strategy for task-state functional magnetic resonance imaging, which can provide an objective, feasible and reliable method for formulating individualized control plans to achieve comprehensive optimization of the overall characteristics of the network.

[0060] 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.

[0061] Embodiment 1

[0062] like Figure 1 As shown, the method for formulating an individualized control strategy for task-state functional magnetic resonance imaging provided by the present invention comprises:

[0063] Step 100: Acquire multi-band task-state functional magnetic resonance imaging of the subject.

[0064] Step 101: construct a functional activation map of the subject based on multi-band task-state functional magnetic resonance imaging. The implementation process of this step may include:

[0065] Step 1010: Preprocess the multi-band task-state functional magnetic resonance images.

[0066] Step 1011: Establish a linear model. The parameters of the linear model include: the experimental task design matrix, the time series extracted in the standard space, and the head motion parameters. The time series and the head motion parameters are extracted from the preprocessed multi-band task-state functional magnetic resonance images. The time series includes the white matter time series, cerebrospinal fluid time series, and global signal time series of the individual subjects.

[0067] Step 1012: Using a linear model, a single sample test of the task state compared to the resting state is defined on the experimental task design matrix to obtain activation in the brain region.

[0068] Step 1013: Generate a functional activation map based on brain region activation.

[0069] Step 102: Extract individual brain region time series based on the functional activation map. The implementation process of this step can be:

[0070] Step 1020: Determine a first spherical region with the region of interest as the group-level brain region coordinates, the group-level brain region coordinates as the sphere center, and a first preset value as the radius.

[0071] Step 1021: The peak point of the motor task activation in the functional activation map in the first spherical region is used as the individual brain region coordinates.

[0072] Step 1022: Determine a second spherical region with the individual brain region coordinates as the sphere center and a second preset value as the radius.

[0073] Step 1023: The first feature vectors of all voxels in the second spherical region are used as individual brain region time series.

[0074] Step 103: Perform dynamic causal modeling and fitting based on the individual brain region time series to obtain a dynamic causal model. The implementation process of this step may include:

[0075] Step 1030: Using the Bayesian method to fit the individual brain region time series data within the framework of the initial dynamic causal model. The initial dynamic causal model is used to characterize the emergent inter-hemispheric connectivity pattern.

[0076] Step 1031: During the data fitting process, the effective connection parameters are adjusted so that the error between the data generated by the initial dynamic causal model and the true value meets the set conditions, and the effective connection parameters between brain regions under each initial dynamic causal model are obtained to form a dynamic causal model.

[0077] Step 104: Perform Bayesian model selection on the dynamic causal model to obtain the individual optimal model. The implementation process of this step can be:

[0078] Step 1040: Use fixed effects analysis to determine the level of evidence for the dynamic causal model.

[0079] Step 1041: The dynamic causal model with the highest level of evidence is taken as the optimal model, and the effective connection parameters corresponding to the optimal model are taken as the individualized model parameters.

[0080] Step 105: Construct a generative model of the dynamic causal model based on the individual optimal model and the individualized model parameters corresponding to the individual optimal model.

[0081] Step 106: Mapping external regulation to the parameters of the generative model to simulate the dynamic activities of the brain network under different regulation strategies. Specifically, the high-frequency transcranial stimulation is used to weaken the self-inhibition strength of the brain region of the generative model, and the low-frequency transcranial stimulation is simplified to enhance the self-inhibition strength of the brain region of the generative model. The results of all brain regions in the target brain network under different regulation strategies are simulated to obtain the dynamic activities of the brain network under different regulation strategies.

[0082] Step 107: Construct a multivariate linear model with the motor scale score at the group level as the dependent variable and the activation of each brain region in the functional activation map as the independent variable.

[0083] Step 108: The output value of the multivariate linear model constructed based on the independent variables is used as the optimization target to compare the dynamic activities of the brain network under different control strategies, and the control strategy that maximizes the motor function output value is determined as the individual optimal strategy.

[0084] Embodiment 2

[0085] This embodiment is based on the method for formulating individualized regulation strategies for task-state functional MRI provided in Embodiment 1, and its overall technical route is: 1. Collect high-temporal-resolution task-state functional MRI images of the subjects. 2. Construct an individual activation map and extract the peak activation voxels of each brain region in the network of interest. 3. Construct a dynamic causal model for the above-mentioned network nodes and calculate the effective connection parameters between brain regions. 4. Use the Bayesian model selection method to screen out the optimal model and parameters. 5. Construct a generative model to simulate the dynamic activity of the brain network under different regulation strategies. 6. Based on group data, a multivariate linear model is constructed with the motor scale score as the dependent variable and the activation of each brain region as the independent variable. 7. At the individual level, the output value of the multivariate linear model is used as the optimization target, and the optimal strategy that maximizes the fitting value is selected. Based on this, Figure 2 As shown, in this embodiment, the specific implementation process of the method provided in Embodiment 1 includes:

[0086] S1. Collect multi-band (high temporal resolution) task-state fMRI of the subjects.

[0087] In this embodiment, the subject completes the motor task as required during the scanning process, and finally obtains the task-state fMRI of the subject. Among them, the multiband factor Multibandfactor = 8, and the repetition time TR = 400ms. The task involves 10 blocks of left-hand tapping (5 blocks) and right-hand tapping (5 blocks) alternating, and each block contains 3 trials. Within the trial, the subject taps the index finger of the corresponding hand at the maximum speed according to the prompts on the screen. Rest time is set between trials and blocks. The purpose of the task is to present the brain network distribution corresponding to the impaired function in fMRI and to record the corresponding brain network dynamics.

[0088] S2. Construct a functional activation map of the individual subjects.

[0089] S21. Preprocess the acquired multi-band (high temporal resolution) task-state fMRI. The specific process includes: DICOM to NII image format conversion, removal of unstable time points, head motion correction, field correction, image segmentation and registration, spatial smoothing, and standardization to MNI space.

[0090] S22. Establish a general linear model. The parameters in the linear model include: ① Experimental task design parameters (i.e., experimental task design matrix). ② In the standard space, extract the time series of the subject's white matter, cerebrospinal fluid, and global signal. ③ Head movement parameters obtained by preprocessing.

[0091] S23. In the general linear model obtained in step S22 and on the experimental task design matrix, a single-sample t-test of the task state versus the resting state is defined to obtain a statistical parameter map of brain functional activation in the task state relative to the non-task state (i.e., a functional activation map).

[0092] S3. Extract individual brain region time series.

[0093] S31. Determine the coordinates of individual brain regions. Locate the region of interest as the coordinates of the group-level brain region, make a spherical region with a radius of 8 mm with the group-level coordinates as the center of the sphere, and select the peak point of the motor task activation in the functional activation map within the spherical region as the coordinates of the individual brain region.

[0094] S32, time series extraction: Extract the first eigenvectors of all voxels in a sphere with the coordinates of the individual brain region as the center and a radius of 6 mm as the individual brain region time series.

[0095] S4. Dynamic causal modeling.

[0096] S41. Model assumptions about the possible inter-hemispheric connections within the brain network, mainly including four situations: only left-to-right, only right-to-left, both left and right, and irrelevant left and right.

[0097] S42. Perform dynamic causal modeling for each model hypothesis and calculate effective connection parameters. Specifically: Under the framework of the dynamic causal model, use the individual brain region time series obtained by the Bayesian method to fit the data, adjust the effective connection parameters to make the model-generated data close to the true value, and finally obtain the effective connection parameters between brain regions under each model hypothesis, that is, the intensity of the directional promotion or inhibition between brain regions.

[0098] S5. Bayesian model selection

[0099] The fixed effects analysis (FFX) method is used to calculate the evidence level (i.e., model accuracy-complexity) of the dynamic causal model obtained in step S4, and the dynamic causal model with the highest evidence level is selected as the individual optimal model. The effective connection parameters corresponding to this dynamic causal model are used as individualized model parameters.

[0100] S6. Intervention outcome simulation

[0101] The individual optimal model obtained in the above steps and the corresponding individualized model parameters are used to construct a generative model of the dynamic causal model, and the external regulation is concretized as the change of the generative model parameters to simulate the dynamic activities of the brain network under different regulation strategies (targets and frequencies). In this embodiment, the high-frequency rTMS effect is simplified to the weakening of the self-inhibition strength of the brain area, and the low-frequency rTMS effect is simplified to the enhancement of the self-inhibition strength of the brain area. The possible results of all possible brain areas in the target network under the two strategies are simulated, and the activation intensity of each brain area after "regulation" in the task state is recorded.

[0102] S7. Optimization indicator construction

[0103] Based on the group data of the same disease and symptoms, in addition to collecting task-state fMRI, the motor function of the subjects was assessed using the motor function scale tool. Then, the motor scale score at the group level was used as the dependent variable, and the activation of each brain region of the subject was used as the independent variable to construct a multivariate linear model, that is, fitting the motor output with brain region activation.

[0104] S8. Development of individual intervention strategies

[0105] At the individual level, the output value of the multivariate linear model constructed by S7 is used as the optimization target, the output values ​​of all possible regulatory strategies (i.e., brain network dynamic activities) are compared, and the regulatory strategy that maximizes the fitted output value is selected as the individual optimal strategy.

[0106] Based on the above description of Embodiment 1 and Embodiment 2, compared with the prior art, the technical solution provided by the present invention has the following advantages:

[0107] 1) The present invention innovatively combines brain activation analysis and brain effective connectivity analysis, and proposes an individualized regulatory strategy formulation method, including target determination and frequency selection strategies at the brain region level and voxel level.

[0108] 2) The present invention proposes a method of combining multi-band fMRI scanning with individual neuroimaging analysis, which can improve the reliability of individual network modeling with the advantage of high temporal resolution compared to traditional scanning parameters.

[0109] 3) The present invention innovatively divides the determination of individual targets into two independent steps: brain region level and voxel level, so as to better correspond to the scope of action of common regulatory means and avoid the interference of abnormal voxels in individual strategy formulation.

[0110] 4) The present invention innovatively applies the method of dynamic causal modeling to neural regulation, which can provide information on the direction of interaction between brain regions and the promotion-inhibition relationship, provide a basis for the formulation of regulatory strategies, and simulate various possible strategies based on the generative model. It can comprehensively compare the effects of different strategies and provide the optimal regulatory strategy solution.

[0111] 5) The present invention proposes to use the network integrity index as the optimization target, model the fitting relationship between brain network parameters and functional scales with group data, and predict the actual functional output of the network based on the model. Compared with the traditional brain area activation index as the optimization purpose, the idea of ​​the present invention is to consider the common contribution of each brain area in the network to the functional output, which is more reasonable.

[0112] 6) The individualized regulation strategy formulation method proposed in the present invention is an objective method based on individual data analysis under the condition of considering the classic intervention strategy. It does not rely on the subjective judgment of medical staff and is superior to experience-based selection.

[0113] 7) The present invention regards each brain region as a whole, combines the prior knowledge of network assumptions and the calculated brain activation values, first obtains the individualized coordinates of each brain region in the network, and simplifies it into a region centered on a representative voxel, and then selects the possible target brain region. The present invention divides the target selection at the brain region level and the voxel level into two steps, thereby taking into account the overall effect of the brain region, better resisting the interference of voxel noise, and making the selected target conform to the actual regulation scenario.

[0114] 8) The present invention uses a dynamic causal modeling method to analyze the directed connections at the level of neural groups between brain regions. This method combines modeling at the neural activity level and the hemodynamic level, which can clarify the direction of information conduction between brain regions and ensure effective stimulation conduction.

[0115] 9) Based on effective connection analysis, the present invention not only clarifies the connection direction between brain regions, but also clarifies the promotion or inhibition of brain regions, compared with the traditional method that only provides the strength of the correlation between brain regions. In addition, the generative model provided by the present invention allows simulation of different states, and can comprehensively compare the possible effects of different strategies, and can use parameters at a higher level than the activity of a single brain region as the optimization target of the regulation strategy.

[0116] In addition, the present invention also provides a task-state functional magnetic resonance imaging individualized regulation strategy formulation system, the system comprising:

[0117] The image collector is used to collect multi-band task-state functional magnetic resonance images of the subjects.

[0118] Memory used to store computer software programs.

[0119] The processor is connected to the image collector and the memory, respectively, and is used to call and execute a computer software program to implement the above-mentioned task-state functional magnetic resonance individualized regulation strategy formulation method based on the multi-band task-state functional magnetic resonance images of the subjects acquired by the image collector.

[0120] Furthermore, the processor used above may include: an image acquisition unit, a map construction unit, a sequence extraction unit, a model fitting unit, a model selection unit, a first model construction unit, an activity simulation unit, a second model construction unit and a strategy determination unit.

[0121] The image acquisition unit is used to acquire multi-band task-state functional magnetic resonance images of the subject.

[0122] The atlas construction unit is used to construct a functional activation atlas of the subject based on multi-band task-state functional magnetic resonance imaging.

[0123] The sequence extraction unit is used to extract individual brain region time series based on the functional activation map.

[0124] The model fitting unit is used to perform dynamic causal modeling and fitting based on the individual brain region time series to obtain a dynamic causal model.

[0125] The model selection unit is used to perform Bayesian model selection on the dynamic causal model to obtain the individual optimal model.

[0126] The first model building unit is used to build a generation model of the dynamic causal model based on the individual optimal model and the individualized model parameters corresponding to the individual optimal model.

[0127] The activity simulation unit is used to map external regulation into parameters of the generative model to simulate the dynamic activity of the brain network under different regulation strategies.

[0128] The second model building unit is used to build a multivariate linear model with the motor scale score at the group level as the dependent variable and the activation of each brain region in the functional activation map as the independent variable.

[0129] The strategy determination unit is used to construct the output value of the multivariate linear model based on the independent variables, compare the dynamic activities of the brain network under different control strategies as the optimization target, and determine the control strategy that maximizes the motor function output value as the individual optimal strategy.

[0130] In addition, when the computer program in the above-mentioned memory is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk.

[0131] Although the image preprocessing, brain activation map analysis, dynamic causal modeling and statistical test methods used in the present invention are all mainstream methods in this field, the present invention has made innovative combinations and improvements on this basis and proposed a strategy formulation method for individualized neural regulation. It should be noted that:

[0132] 1. The purpose of high temporal resolution imaging in the present invention is to improve the reliability of subsequent analysis results. An alternative method may also be a combination of multimodal imaging based on task-based fMRI.

[0133] 2. The purpose of using dynamic causal modeling in the present invention is to obtain an effective connection model and perform data simulation generation. Alternative methods may also be Granger causal models, graph theory models, models constructed by deep learning methods, etc.

[0134] 3. The main purpose of using the Bayesian model selection method in the present invention is to obtain the optimal effective connection model for the subjects. An alternative solution may also be Bayesian model averaging, that is, weighted averaging the modeling results of multiple possible models according to the model evidence level.

[0135] 4. The present invention exemplifies the application of the method in the scenario of "optimizing the activity of the affected brain area by combining the relationship between the bilateral hemispheres", but the individual regulation strategy formulation method proposed in the present invention can also be extended to a network composed of any two or more brain areas, and the scope of applicable symptoms is not limited to post-stroke movement disorders.

[0136] 5. In the embodiments of the present invention, the optimization index construction is completed using a multivariate linear model, and its main purpose is to construct a fitting relationship between brain activation and functional output. Alternative solutions may use other complex linear and nonlinear equations, machine learning models, etc.

[0137] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0138] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for formulating an individualized control strategy for task-based functional magnetic resonance imaging, characterized in that: include: Acquire multi-band task-state functional magnetic resonance imaging of the subjects; Constructing a functional activation map of the subject based on the multi-band task-state functional magnetic resonance imaging; Extracting individual brain region time series based on the functional activation map; Perform dynamic causal modeling and fitting based on the individual brain region time series to obtain a dynamic causal model; Performing Bayesian model selection on the dynamic causal model to obtain an individual optimal model; Constructing a generative model of a dynamic causal model based on the individual optimal model and the individualized model parameters corresponding to the individual optimal model; Mapping external regulation into parameters of the generative model to simulate the dynamic activities of brain networks under different regulation strategies; A multivariate linear model was constructed with the motor scale score at the group level as the dependent variable and the activation of each brain region in the functional activation map as the independent variable; Comparing the dynamic activities of the brain network under different control strategies based on the output value of the multivariate linear model as the optimization target, and determining the control strategy that maximizes the motor function output value as the individual optimal strategy; Constructing a functional activation map of the subject based on the multi-band task-state functional magnetic resonance imaging specifically includes: Preprocessing the multi-band task-state functional magnetic resonance images; A linear model is established; the parameters of the linear model include: an experimental task design matrix, a time series extracted in a standard space, and a head motion parameter; the time series and the head motion parameter are extracted from the preprocessed multi-band task-state functional magnetic resonance imaging; the time series includes the white matter time series, cerebrospinal fluid time series, and global signal time series of the individual subject; Using the linear model, a single sample test of the task state compared with the resting state is defined on the experimental task design matrix to obtain activation in the brain region; generating the functional activation map based on the brain region activation; The dynamic causal model is obtained by performing dynamic causal modeling and fitting based on the individual brain region time series, specifically including: Using a Bayesian approach to fit the individual brain region time series data within the framework of an initial dynamic causal model; the initial dynamic causal model is used to characterize the emergent interhemispheric connectivity patterns; During the data fitting process, the effective connection parameters are adjusted so that the error between the data generated by the initial dynamic causal model and the true value meets the set conditions, and the effective connection parameters between brain regions under each initial dynamic causal model are obtained to form the dynamic causal model; Mapping external regulation into parameters of the generative model to simulate the dynamic activities of brain networks under different regulation strategies, including: The effect of high-frequency transcranial stimulation is used to weaken the self-inhibition strength of the brain areas in the generative model, while the effect of low-frequency transcranial stimulation is simplified to enhance the self-inhibition strength of the brain areas in the generative model. The results of all brain areas in the target brain network under different regulation strategies are simulated to obtain the dynamic activities of the brain network under different regulation strategies.

2. The method for formulating task-based functional magnetic resonance imaging individualized control strategy according to claim 1, characterized in that: Extracting individual brain region time series based on the functional activation map specifically includes: Taking the region of interest as the group-level brain region coordinates, taking the group-level brain region coordinates as the sphere center and taking the first preset value as the radius, determine a first spherical region; The peak points of the motor task activation in the functional activation map in the first spherical region are used as the coordinates of the individual brain regions; Determine a second spherical area with the coordinates of the individual brain region as the center of the sphere and a second preset value as the radius; The first feature vectors of all voxels in the second spherical region are used as the individual brain region time series.

3. The method for formulating task-based functional magnetic resonance imaging individualized control strategy according to claim 1, characterized in that: The Bayesian model selection is performed on the dynamic causal model to obtain an individual optimal model, specifically including: Fixed-effects analysis was used to determine the level of evidence for the dynamic causal model; The dynamic causal model with the highest level of evidence is taken as the optimal model; and the effective connection parameters corresponding to the optimal model are taken as the individualized model parameters.

4. A task-based functional magnetic resonance imaging individualized control strategy formulation system, characterized in that: include: An image collector, used for collecting multi-band task-state functional magnetic resonance images of the subject; Memory for storing computer software programs; A processor is connected to the image collector and the memory, respectively, and is used to call and execute the computer software program to implement the method for formulating an individualized task-state functional magnetic resonance regulation strategy as described in any one of claims 1 to 3 based on the multi-band task-state functional magnetic resonance images of the subject acquired by the image collector.

5. The task-based functional magnetic resonance imaging individualized control strategy formulation system according to claim 4, characterized in that: The processor comprises: An image acquisition unit, used for acquiring multi-band task-state functional magnetic resonance images of the subject; A map construction unit, used for constructing a functional activation map of the subject based on the multi-band task-state functional magnetic resonance imaging; A sequence extraction unit, used for extracting individual brain region time series based on the functional activation map; A model fitting unit, used for performing dynamic causal modeling fitting based on the individual brain region time series to obtain a dynamic causal model; A model selection unit, used for performing Bayesian model selection on the dynamic causal model to obtain an individual optimal model; A first model building unit, configured to build a generation model of a dynamic causal model based on the individual optimal model and the individualized model parameters corresponding to the individual optimal model; The activity simulation unit is used to map external regulation into parameters of the generative model to simulate the dynamic activities of the brain network under different regulation strategies; A second model building unit is used to build a multivariate linear model with the motor scale score at the group level as the dependent variable and the activation of each brain region in the functional activation map as the independent variable; A strategy determination unit is used to compare the dynamic activities of the brain network under different control strategies based on the output value of the multivariate linear model as the optimization target, and determine the control strategy that maximizes the motor function output value as the individual optimal strategy.

6. The task-based functional magnetic resonance imaging individualized control strategy formulation system according to claim 4, characterized in that: The memory is a computer-readable storage medium.

Images