A construction method and system for a co-spatiotemporal brain computing model

Abstract

The present invention provides a method and system for constructing a co-spatiotemporal brain computing model, belonging to the technical fields of computational neuroscience and artificial intelligence; it solves the problem that the existing models have insufficient simulation accuracy for brain neural dynamics activities; it includes the following steps: continuously scanning the brain to obtain functional magnetic resonance imaging data of the brain; preprocessing the obtained functional magnetic resonance imaging data to obtain BOLD signals of each moment in different brain regions; constructing a brain computing model: dividing the BOLD signals into multiple transmission intensity matrices, and extracting a spatiotemporal weight matrix containing the intrinsic common information of the brain from these transmission intensity matrices; simulating brain activity signals according to the spatiotemporal weight matrix, and calculating the corresponding phase space similarity loss function; determining the parameters of the optimal brain computing model through the phase space similarity loss function to construct a personalized co-spatiotemporal brain computing model; the present invention is applied to the construction of brain computing models.

Description

Technical Field

[0001] The present invention provides a method and system for constructing a co - spatio - temporal brain computing model, belonging to the technical fields of computational neuroscience and artificial intelligence. Background Art

[0002] With the advent of the era of artificial intelligence, understanding human behavior and cognitive processes can build more intelligent and adaptable intelligent systems. Human cognition mainly comes from the process of the release of electrical and chemical signals between neurons in the brain system, which is also known as brain activity. With the development of neuroimaging technology, different recording techniques and computational analysis methods have provided some insights into the study of brain internal activities. Brain computing models based on neurodynamics theory can simulate the synchronous patterns between biologically realistic neuron populations, thereby directly generating spontaneous or stimulus - induced activities, reproducing the changes in dynamic activities in the brain network, and further explaining its mechanism. Therefore, using this technology to construct a virtual brain and simulate and analyze the internal activity state creates opportunities for better describing brain dynamics and its relationship with brain functions. However, the currently existing brain computing models are restricted by the observed brain imaging data. They are mainly constructed based on a symmetric and only positive anatomical structure, which is the connectivity parameterization estimated by the white matter integrity of diffusion imaging. There may be potential misinferences in this; in the loss function part of the later stage of the model, only the similarity between functional networks is considered, while ignoring the unique dynamic characteristics within the brain. This series of reasons leads to unsatisfactory current simulation results and the inability to obtain a reliable and interpretable simulated brain model. Summary of the Invention

[0003] To solve the problem that the existing models have insufficient simulation accuracy for brain neurodynamics activities, the present invention proposes a method and system for constructing a co - spatio - temporal brain computing model to accurately simulate brain nerve activities and functions.

[0004] The technical solution adopted by the present invention is as follows: A method for constructing a co - spatio - temporal brain computing model includes the following steps:

[0005] Step 1: Continuously scan the brain to obtain functional magnetic resonance imaging data of the brain;

[0006] Step 2: Pre - process the obtained functional magnetic resonance imaging data to obtain the BOLD signals of each moment in different brain regions;

[0007] Step 3: Construct a brain computing model, specifically including:

[0008] Step 3.1: Divide the BOLD signals into multiple transmission intensity matrices, and extract a spatio - temporal weight matrix containing the common information inherent in the brain from these transmission intensity matrices;

[0009] Step 3.2: According to the spatio-temporal weight matrix, simulate the brain activity signal by calculating the relationship between the spontaneous local activity of different brain regions and their outputs to other brain regions. The process of simulating brain nerve activity is as follows:

[0010] Based on the theory of neural dynamics, take the spatio-temporal weight matrix constructed above as the input of the brain computational model, and describe the evolution process between each brain region through the following differential equation

[0011]

[0012] where: η t is the noise received by the brain region, describes the spatio-temporal reconstruction information captured from the brain activity, is the relationship between the spontaneous local activity of a brain region and its output to other brain regions;

[0013] Convert the simulated neural activity into the simulated BOLD signal through the hemodynamic response Balloon-Windkessel model;

[0014] Step 4: Calculate the corresponding phase space similarity loss function according to the empirically collected data and the brain activities simulated by the brain computational model under different parameters;

[0015] Step 5: Determine the optimal brain computational model parameters through the phase space similarity loss function to construct a personalized co-spatio-temporal brain computational model.

[0016] Furthermore, the functional magnetic resonance imaging data is collected by a nuclear magnetic resonance machine.

[0017] Furthermore, the preprocessing of the obtained functional magnetic resonance imaging data includes: data deletion, image time slice correction, image head motion correction, spatial smoothing, filtering, and image covariate regression processing.

[0018] Furthermore, use the brain atlas segmentation template to partition the preprocessed functional magnetic resonance imaging data and extract the BOLD signal of each brain region.

[0019] Furthermore, the process of constructing the spatio-temporal weight matrix is as follows: divide the BOLD signal into multiple time blocks by the method of dividing sliding time windows, and calculate the Pearson correlation between brain regions within each time block respectively to construct the corresponding transmission intensity matrix; then extract the common information of these transmission intensity matrices through the fusion algorithm of augmented Lagrangian multiplier and principal component pursuit to construct the spatio-temporal weight matrix containing the intrinsic common information of the brain.

[0020] Further, the process of constructing the phase space similarity loss function is as follows:

[0021] Convert the brain activities simulated in step 3.2 into the corresponding phase spaces of each brain region respectively, and then use the following formula to measure the spatial similarity of the phase spaces between brain regions to construct a phase space similarity matrix:

[0022]

[0023] In the formula: PSA ij is the phase space similarity between the i-th and j-th regions of the brain, PS i is the phase space map of the i-th region of the brain, PS j is the phase space map of the j-th region of the brain, is the average value of the corresponding term, σ PS is the standard deviation of the corresponding term, c1 = (k 1 L) 2 and c2 = (k 2 L) 2 are both constants used to maintain stability, where L is the dynamic range of the input image, k 1 = 0.01, k 2 = 0.03;

[0024] Calculate the phase space similarity matrices of the brain activities collected according to experience and the brain activities simulated by the Balloon-Windkessel model respectively, and then compare the similarity between the two matrices through Pearson correlation to construct a phase space similarity loss function.

[0025] Further, the process of constructing the co-spatiotemporal brain calculation model is as follows: Solve the phase space similarity loss function values of the brain calculation model under different parameters through the genetic algorithm, find the parameters with the most similar empirical BOLD signal and simulated BOLD signal, and construct a personalized co-spatiotemporal brain calculation model.

[0026] Further, the empirical acquisition data in step 4 is functional magnetic resonance imaging data collected by a nuclear magnetic resonance machine.

[0027] A system for constructing a co-spatiotemporal brain calculation model includes:

[0028] An image data acquisition module for acquiring brain image data collected by a nuclear magnetic resonance machine;

[0029] A preprocessing module for preprocessing the brain image data to obtain the BOLD signal of each brain region;

[0030] A spatiotemporal weight matrix construction module for extracting the information common to the brain;

[0031] A neural activity simulation module for describing the evolution process between each brain region;

[0032] A loss function construction module for determining the optimal parameters of the brain computational model and evaluating the accuracy of simulating the BOLD signal;

[0033] A co - spatio - temporal brain computational model construction module for constructing a co - spatio - temporal brain computational model according to the determined optimal parameters to simulate accurate and personalized brain activities.

[0034] The beneficial effects of the present invention compared with the prior art are as follows: Compared with the existing methods for constructing brain computational models based on neurodynamics theory, the method and system for constructing a co - spatio - temporal brain computational model proposed by the present invention can simulate neural activities and neurophysiological characteristics closer to the real brain, accurately and comprehensively clarify the functional characteristics and neural activity mechanisms of the brain, and turn the sectional brain in anatomical biological research into a vivid dynamic brain, and study the brain working mechanism more systematically and dynamically. Brief Description of the Drawings

[0035] The following further describes the present invention with reference to the drawings:

[0036] Figure 1 is a schematic flow chart of the method for constructing a co - spatio - temporal brain computational model provided by the present invention.

[0037] Figure 2 is a schematic diagram of the construction principle and mathematical representation of the co - spatio - temporal brain computational model provided by the present invention.

[0038] Figure 3 is a schematic structural diagram of the system for constructing a co - spatio - temporal brain computational model provided by the present invention.

[0039] Figure 4 is a comparison chart of the simulated signals and empirical data obtained by using the method and system provided by the present invention based on 30 healthy subjects on multiple evaluation indexes. Detailed Embodiments

[0040] As Figures 1 to 4 shown, the present invention provides a method for constructing a co - spatio - temporal brain computational model, including the following steps:

[0041] Step 1: Obtain functional magnetic resonance imaging (fMRI) data of the brain by continuously scanning the brain;

[0042] Step 2: Pre - process the obtained fMRI data to obtain the BOLD signal of each moment in different brain regions. Specifically, the process of pre - processing the fMRI data is as follows:

[0043] Remove the first 10 time points of the collected data to ensure the image quality of the subjects;

[0044] Perform temporal slice correction on the images collected by the nuclear magnetic resonance machine to reduce the systematic error in subsequent data analysis caused by the differences in acquisition time;

[0045] Perform head motion correction on the images collected by the nuclear magnetic resonance machine so that each image can be aligned with the standard template;

[0046] Perform spatial smoothing on the images using a Gaussian kernel function to suppress the influence of noise and improve the signal-to-noise ratio of the images;

[0047] Perform noise reduction on the images using the ICA-AROMA strategy to effectively remove the motion artifacts in the functional magnetic resonance data and improve the sensitivity to brain signals at the same time;

[0048] Perform band-pass filtering (0.01Hz - 0.09Hz) on the images collected by the nuclear magnetic resonance machine to eliminate noise interference;

[0049] Perform covariate regression on the images collected by the nuclear magnetic resonance machine to remove the interference noise of white matter, cerebrospinal fluid, and whole-brain mean, etc., and reduce the interference of non-neuronal signals;

[0050] Use the brain atlas segmentation template to partition the images collected by the nuclear magnetic resonance machine and extract the BOLD signals of each brain region.

[0051] Step 3: Construct a brain computational model, specifically including:

[0052] Step 3.1: Divide the BOLD signals into multiple transmission intensity matrices by dividing sliding time windows, and extract the spatio-temporal weight matrix containing the intrinsic common information of the brain from these matrices. Specifically, the process of constructing the spatio-temporal weight matrix is as follows:

[0053] Divide the BOLD signals into multiple time blocks by dividing sliding time windows, and calculate the Pearson correlation between brain regions within each time block respectively to construct the corresponding transmission intensity matrices; then extract the common information of these matrices through the fusion algorithm of augmented Lagrangian multiplier and principal component pursuit to construct the spatio-temporal weight matrix containing the intrinsic common information of the brain. This process can be expressed by the following formula:

[0054]

[0055] In the formula: {X i (t), t = 1,..., T, i = 1,..., M} is the collected brain activity, N is the number of time blocks divided, L n = [l1,..., l Nis a matrix of rank , S n = [s1,..., s N is a sparse matrix with non-zero entries, and W L is a spatio-temporal weight matrix containing the intrinsic common information of the brain.

[0056] For the decomposition process, the following convex optimization problem is considered to be solved:

[0057]

[0058] Where: is the nuclear norm of matrix L n , that is, the sum of all its singular values, ‖S n ‖1 = Σ ij |s nij | is the l1 norm of S n , and λ1 > 0 is a regularization parameter that controls the trade-off between the two matrices L n and S n . ‖Bvec(L n )‖1 is regularized by adjusting the parameter λ2 > 0. By penalizing the differences in the connectivity profiles, it encourages the individual-induced group homogeneity. B is a matrix defined by , and D is the first-order difference matrix is an identity matrix of M 2 × M 2 .

[0059] Step 3.2: According to the spatio-temporal weight matrix, simulate the brain activity signal by calculating the relationship between the spontaneous local activity of different brain regions (i.e., neuron groups) and their outputs to other brain regions. Specifically, the process of simulating brain nerve activity is as follows:

[0060] Based on the theory of neural dynamics, taking the spatio-temporal weight matrix constructed in Step 3.2 above as the input of the co-spatio-temporal brain computational model, the evolution process between each brain region is described by a set of differential equations:

[0061]

[0062] Where: η t is the noise received by the brain region, describes the spatio-temporal reconstruction information captured from the brain activity, is the relationship between the spontaneous local activity of a brain region (i.e., neuron population) and its output to other brain regions.

[0063] Convert the simulated neural activity into the simulated BOLD signal through the hemodynamic response Balloon-Windkessel model. Simply put, the synaptic impulse activity S in each brain region i results in an increase in the vasodilation signal Z i which is affected by autoregulatory feedback, and the blood flow f i responds proportionally to the signal Z i accompanied by changes in blood volume v i and deoxyhemoglobin content q i The specific equations are as follows:

[0064]

[0065] where: S i represents the neural activity signal, ρ = 0.34 is the resting oxygen extraction fraction, k = 0.65 (s -1 ) is the signal decay rate, Z i is the vasodilation signal, γ = 0.41 (s -1 ) is the blood flow-dependent elimination rate, f i is the blood inflow, v i is the blood volume, τ = 0.98 (s) represents the hemodynamic time constant, α = 0.32 is the Grubb constant, V0 = 0.02 is the resting blood volume ratio, and a set of parameter values related to the 3T magnetic field strength are k1 = 2.38, k2 = 2, and k3 = 0.48.

[0066] Step 4: Calculate the corresponding phase space similarity loss function based on the empirically collected data (fMRI data) and the brain activities simulated by the brain computational model under different parameters. Specifically, the process of constructing the phase space similarity loss function is as follows:

[0067] Convert the obtained brain activities into the corresponding phase spaces of each brain region respectively, and then use the following formula to measure the spatial similarity between the phase spaces of brain regions to construct a phase space similarity matrix:

[0068]

[0069] where: PSA ij is the phase space similarity between the i-th and j-th regions of the brain, PS i is the phase space map of the i-th region of the brain, PS j is the phase space map of the j-th region of the brain, is the average value of the corresponding terms, σ PS is the standard deviation of the corresponding terms, c1 = (k 1 L) 2 , c2 = (k 2 L)2 are all constants used to maintain stability, where L is the dynamic range of the input image, and k 1 = 0.01, k 2 = 0.03;

[0070] Calculate the phase space similarity matrix of the brain activity collected empirically and the brain activity simulated by the brain computational model respectively, and then compare the similarity between the two matrices through Pearson correlation to construct a phase space similarity loss function.

[0071] Step 5: Determine the optimal parameters of the brain computational model through the phase space similarity loss function to construct a personalized spatiotemporal brain computational model. Specifically, the process of constructing the spatiotemporal brain computational model is as follows:

[0072] Solve the phase space similarity loss function values of the brain computational model under different parameters through the genetic algorithm, and find the parameters with the most similar empirical signal and simulated signal to construct a personalized spatiotemporal brain computational model.

[0073] Corresponding to Figure 1 the construction method of the spatiotemporal brain computational model shown, the present invention also provides a system for constructing a spatiotemporal brain computational model, as Figure 3 shown, the system includes the following structures:

[0074] An image data acquisition module 201, configured to acquire brain image data collected by a nuclear magnetic resonance machine;

[0075] A preprocessing module 202, configured to preprocess the brain image data to obtain the BOLD signal of each brain region;

[0076] A spatiotemporal weight matrix construction module 203, configured to extract the information inherent in the brain;

[0077] A neural activity simulation module 204, configured to describe the evolution process between each brain region;

[0078] A loss function construction module 205, configured to determine the optimal parameters of the brain computational model and evaluate the accuracy of the simulated signal;

[0079] A spatiotemporal brain computational model construction module 206, configured to simulate accurate and personalized brain activities according to the determined optimal parameters.

[0080] Figure 4 is a comparison chart of the simulated signals and empirical data obtained from 30 healthy subjects using the method and system provided by the present invention on evaluation indicators (edge FC, node FC, DFC).

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a co-spatiotemporal brain computing model, characterized in that: Including the following steps: Step 1: Obtain functional magnetic resonance imaging (fMRI) data of the brain by continuously scanning the brain; Step 2: Preprocess the obtained fMRI data to obtain the BOLD signals of each brain region at each moment; Step 3: Construct a brain computational model, specifically including: Step 3.1: Divide the BOLD signals into multiple transmission intensity matrices, and extract the spatio-temporal weight matrix containing the intrinsic common information of the brain from these transmission intensity matrices; Step 3.2: According to the spatio-temporal weight matrix, simulate the brain activity signals by calculating the relationship between the spontaneous local activity of different brain regions and their outputs to other brain regions. The process of simulating brain nerve activity is as follows: Based on the neurodynamics theory, the constructed spatio-temporal weight matrix above is used as the input of the brain computational model, and the evolution process between each brain region is described by the following differential equation where: η t is the noise received by the brain region, and ζ(W) describes the spatio-temporal reconstruction information captured from brain activities, is the relationship between the spontaneous local activity of a brain region and its output to other brain regions; Convert the simulated nerve activity into the simulated BOLD signal through the hemodynamic response Balloon-Windkessel model; Step 4: According to experience, collect the data and the brain activities obtained by simulating the brain computing model under different parameters, and calculate the corresponding phase space similarity loss function; Step 5: Determine the optimal brain computational model parameters through the phase space similarity loss function to construct a personalized co-spatiotemporal brain computational model.

2. The construction method of a co-spatiotemporal brain computing model according to claim 1, wherein: The fMRI data is collected by a nuclear magnetic resonance machine.

3. The construction method of a co-spatiotemporal brain computing model according to claim 1, wherein: The preprocessing of the obtained fMRI data includes: data deletion, image time slice correction, image head motion correction, spatial smoothing, filtering, and image covariate regression processing.

4. The construction method of a co-spatiotemporal brain computing model according to claim 3, characterized in that: Use a brain atlas segmentation template to partition the preprocessed fMRI data and extract the BOLD signals of each brain region.

5. The construction method of a co-spatiotemporal brain computing model according to claim 4, characterized in that: The process of constructing the spatio-temporal weight matrix is as follows: Divide the BOLD signals into multiple time blocks by dividing the sliding time window, and calculate the Pearson correlation between brain regions within each time block respectively to construct the corresponding transmission intensity matrix; then extract the common information of these transmission intensity matrices through the fusion algorithm of augmented Lagrangian multiplier and principal component pursuit to construct the spatio-temporal weight matrix containing the intrinsic common information of the brain.

6. The construction method of a co-spatiotemporal brain computing model according to claim 5, wherein: The process of constructing the phase space similarity loss function is as follows: Convert the brain activity simulated in Step 3.2 into the corresponding phase space of each brain region respectively, and then use the following formula to measure the spatial similarity between the phase spaces of brain regions to construct a phase space similarity matrix: where: PSA ij is the phase space similarity between the i-th and j-th regions of the brain, PS i is the phase space diagram of the i-th region of the brain, PS j is the phase space diagram of the j-th region of the brain, is the average value of the corresponding term, σ PS is the standard deviation of the corresponding term, c1 = (k 1 L) 2 、c2 = (k 2 L) 2 are both constants used to maintain stability, where L is the dynamic range of the input image, k 1 = 0.01, k 2 = 0.03; Calculate the phase space similarity matrices of the brain activity collected according to experience and the brain activity simulated by the Balloon-Windkessel model respectively, and then compare the similarity between the two matrices through Pearson correlation to construct the phase space similarity loss function.

7. The construction method of a co-spatiotemporal brain computing model according to claim 6, wherein: The process of constructing the spatiotemporal brain computational model is as follows: Solve the value of the phase space similarity loss function of the brain computational model under different parameters through the genetic algorithm, find the parameters with the most similar empirical BOLD signal and simulated BOLD signal, and construct a personalized spatiotemporal brain computational model.

8. The construction method of a co-spatiotemporal brain computing model according to claim 7, wherein: The empirical acquisition data in Step 4 is the fMRI data collected by a nuclear magnetic resonance machine.

9. A system constructed by using the construction method of the co-spatiotemporal brain computing model according to any one of claims 1-8, characterized in that: Including: An image data acquisition module for acquiring brain image data collected by a nuclear magnetic resonance machine; A preprocessing module for preprocessing the brain image data to obtain the BOLD signals of each brain region; A spatio-temporal weight matrix construction module for extracting the intrinsic common information of the brain; A nerve activity simulation module for describing the evolution process between each brain region; A loss function construction module for determining the optimal parameters of the brain computational model and evaluating the accuracy of the simulated BOLD signal; a spatiotemporal brain computational model construction module for constructing a spatiotemporal brain computational model according to the determined optimal parameters to simulate accurate and personalized brain activities.

Images