Regulation and control scheme auxiliary analysis method based on disease specificity whole brain kinetic model

Abstract

The invention belongs to the technical field of brain dynamics modeling, particularly relates to a regulation and control scheme auxiliary analysis method based on a disease specificity whole brain dynamics model, and aims to realize accurate screening of nerve regulation and control schemes. Comprising the steps that S1, multi-mode magnetic resonance imaging data are obtained and preprocessed, structural connection, a brain region BOLD time sequence, static function connection and dynamic function connection are obtained, and prior constraint values of myelination and cortical atrophy are obtained through calculation based on the preprocessed data; s2, parameterizing the local return coupling strength of the model based on a prior constraint value, coupling the models of different brain regions through the structural connection obtained in the step S1, and constructing a whole brain dynamic model through parameter optimization; s3, training a variational auto-encoder by using different groups of static function connections and dynamic function connections; and S4, applying different regulation and control schemes on the whole brain dynamic model, obtaining a brain state evolution trajectory induced by regulation and control through the trained variational auto-encoder, and screening an effective regulation and control scheme.

Description

Technical Field

[0001] This invention relates to the field of brain dynamics modeling technology, and in particular to an auxiliary analysis method for regulation schemes based on disease-specific whole-brain dynamics models. Background Technology

[0002] The high prevalence, disability rate, and economic burden of neurodegenerative diseases have become a major global public health challenge. Currently, drug treatments have limited efficacy and often involve side effects. Neuromodulation, as a non-pharmacological intervention, modulates neural signal transmission through physical or chemical means, offering a new direction for disease treatment. However, ethical constraints limit large-scale clinical trials, and the "combinatorial explosion" problem in the search space for stimulus parameters, including location and dosage, restricts the optimization of modulation strategies. Whole-brain dynamics models offer a new opportunity to address these issues, potentially enabling the prediction and optimization of individualized neuromodulation protocols for diseases. However, traditional whole-brain dynamics models often employ the assumption of spatial homogeneity, neglecting the regional specificity of brain structure and function in disease states. This makes it difficult for the models to accurately reproduce brain activity under pathological conditions, significantly reducing the accuracy of virtual modulation. Summary of the Invention

[0003] The purpose of this invention is to provide an auxiliary analysis method for regulation schemes based on a disease-specific whole-brain dynamics model, aiming to achieve accurate screening and optimization of neural regulation schemes, and to provide reliable computer simulation technology support for preclinical neural regulation scheme decision-making for neurodegenerative diseases.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: The present invention provides an auxiliary analysis method for regulatory schemes based on a disease-specific whole-brain dynamics model, comprising: S1: acquiring multimodal magnetic resonance imaging data and preprocessing it to obtain structural connectivity, brain region BOLD time series, static functional connectivity, and dynamic functional connectivity; calculating prior constraint values ​​for myelination and cortical atrophy based on the preprocessed data; wherein, the multimodal magnetic resonance imaging data includes functional magnetic resonance imaging, diffusion tensor imaging, T1w, and T2w imaging; S2: applying the prior constraint values ​​obtained in step S1 to the model. The local recurrent coupling strength is parameterized, and the models of different brain regions are coupled through the structural connections obtained in step S1. The whole-brain dynamic model is constructed through parameter optimization; S3: The variational autoencoder is trained using static functional connections and dynamic functional connections of different groups; The variational autoencoder is used to capture the spatiotemporal characteristics of brain function and extract low-dimensional potential representations of brain states to distinguish the brain states of different groups; S4: Different regulatory schemes are applied to the whole-brain dynamic model constructed in step S2. The brain state evolution trajectory induced by regulation is obtained through the variational autoencoder trained in step S3, and effective regulatory schemes are screened.

[0005] Preprocessing includes: processing multimodal magnetic resonance imaging (MRI) data using a common automated processing pipeline based on brain atlas templates; functional magnetic resonance imaging (fMRI) data processing includes removing initial time points, correcting slice times and head movements, spatial smoothing, and filtering to extract brain region BOLD time series; diffusion tensor imaging (DTI) data processing to generate a structural connectivity matrix reflecting interregional connectivity; and processing T1w and T2w imaging data to extract myelination and cortical thickness information.

[0006] In step S1, according to The W score for cortical thickness was calculated for each subject, where β1 is the age-related regression coefficient for each brain region, and β2 is the intercept value. For residual plots, This is the original value of the cortical thickness. This represents the expected value of cortical thickness in the control group corresponding to the patient's age. The standard deviation of the residuals.

[0007] according to Myelination, represented by the T1w / T2w ratio, is introduced as a priori constraint reflecting the cortical hierarchy; where, This represents the T1w / T2w ratio in brain region i.

[0008] In step S2, the parameterization of the local recursive coupling strength is based on This is achieved through a linear combination of prior constraint values ​​for myelination and cortical atrophy.

[0009] The whole-brain dynamics model is a dynamic mean-field model used to describe the physiological firing patterns of neuronal populations. It characterizes the dynamic properties of brain regions through coupled nonlinear differential equations. Nonlinear differential equations include: in, Represents the average synaptic gating variable. This represents the total input current to each brain region. Indicates the discharge rate. It is a structural connection matrix. It scales the global coupling strength to the connection strength of the structural connections. It is the strength of cyclic connectivity in the activity of neuronal groups within a brain region. Represents external input, excitability-gated variable , , d .

[0010] In step S3, the variational autoencoder includes an encoder network, intermediate variational layers, and a decoder network; the encoder network dynamically maps the functional connections of the input to a Gaussian probability distribution in the latent space, and the decoder network is used to reconstruct the input data from samples of the Gaussian probability distribution.

[0011] In step S4, the modulation scheme includes applying external current stimulation to different brain regions of the model; the external current stimulation is a periodic pulse current, and different schemes are formed by adjusting the stimulation amplitude, frequency and pulse width.

[0012] After applying external electrical stimulation, brain functional activities simulated by a disease-specific whole-brain dynamics model are encoded, and the brain state evolution trajectory is plotted in a low-dimensional latent space using a variational autoencoder.

[0013] The criteria for selecting effective modulation schemes include: the brain state after modulation should be similar to that of the healthy group, and the nonlinear dynamic indicators and brain network properties after modulation should be closer to those of the healthy group than before modulation.

[0014] Compared with the prior art, the beneficial effects of this application are as follows: 1. This application provides an auxiliary analysis method for regulatory schemes based on a disease-specific whole-brain dynamics model. By integrating multimodal magnetic resonance imaging data, it extracts region-specific features of myelination and cortical atrophy, achieving precise modeling of the pathological features of neurodegenerative diseases. This innovation breaks through the spatial homogeneity assumption of traditional whole-brain models, enabling the model to more realistically reflect brain activity in disease states.

[0015] 2. A parameterization method for local recurrent coupling strength based on prior constraints of myelination and cortical atrophy is proposed, establishing the biological basis for structural-functional associations. This parameterization method not only improves the model's ability to fit resting-state functional connectivity but also effectively captures dynamic changes in brain functional activity.

[0016] 3. A variational autoencoder is used to achieve unsupervised learning and low-dimensional representation of brain states, overcoming the limitation of traditional methods that require a pre-defined number of states. By mapping the evolutionary trajectory of brain states in the latent space, the regulatory effect is intuitively quantified, providing a new dimension for optimizing neural regulation schemes. Simultaneously, by applying virtual stimuli to disease-specific models and combining the analysis of nonlinear dynamic indicators and brain network properties, the mechanism by which neural regulation affects brain activity is revealed, providing a theoretical basis for preclinical screening. Attached Figure Description

[0017] Figure 1This is a flowchart of an auxiliary analysis method for a regulation scheme based on a disease-specific whole-brain dynamics model provided in an embodiment of this application; Figure 2 This is a schematic diagram comparing the fitting effect of an embodiment of this application with that of a traditional model; Figure 3 This is a schematic diagram illustrating the results of assisting in the discovery of neuromodulation targets in Alzheimer's disease, provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0020] In embodiments of the invention, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, article, or apparatus that includes that element.

[0021] In embodiments of the present invention, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0022] This application provides an auxiliary analysis method for regulatory schemes based on a disease-specific whole-brain dynamics model, exemplified by, for example... Figure 1 As shown. The method includes: S1: Acquire multimodal magnetic resonance imaging data and preprocess it to obtain structural connectivity, brain region BOLD time series, static functional connectivity and dynamic functional connectivity. Based on the preprocessed data, calculate the prior constraint values ​​for myelination and cortical atrophy.

[0023] The multimodal magnetic resonance imaging (MRI) data included diffusion tensor imaging (DTI), T1w and T2w imaging, and functional magnetic resonance imaging (fMRI). For example, preprocessing included: DTI data was processed using the FSL, FreeSurfer, and MRtrix3 software toolkits. A structural connectivity matrix reflecting interregional connectivity was generated for each subject by correcting the ratio of fiber tracts to brain region surface area in paired brain regions. The group-averaged structural connectivity matrix was calculated by selecting all connections present in at least 50% of the subjects. T1w and T2w images were processed using FreeSurfer's recon-all command, creating cortical brain region segmentation, brain tissue masks, and white matter fiber tractography masks based on brain atlas templates. Steps included: head movement correction, intensity normalization, skull dissection, removal of non-brain tissue, brain tissue mask generation, cortical reconstruction, white matter segmentation, subcortical segmentation, and cortical surface generation. Functional magnetic resonance imaging (fMRI) data processing included removing the first 10 time points, correcting for slice time and head motion, and performing spatial smoothing and filtering to extract BOLD signals from each brain region. The functional connectivity matrix for each subject was calculated based on the Pearson correlation coefficient of the BOLD time series. The functional connectivity dynamic matrix reflects the temporal similarity of the functional connections (FCs). For each participant, a sliding time window method was used to calculate the functional connectivity matrix for each sliding window, followed by the correlation between the functional connectivity matrices of different time windows, ultimately forming a time-point × time-point size functional connectivity dynamic matrix. In step S2, prior constraint values ​​for myelination and cortical atrophy are calculated based on the preprocessed data, specifically including: Abnormal myelination and cortical atrophy are two important structural indicators reflecting the pathological changes in Alzheimer's disease. Cortical thickness represents macroscopic structural atrophy, while myelin reflects microstructural changes in white matter. The degeneration and loss of neurons in Alzheimer's disease lead to structural atrophy of the brain. Patterns of cortical atrophy have been shown to accurately track disease progression and hold great promise in differentiating subtypes.

[0024] A W score for cortical thickness was calculated for each participant to quantify cortical atrophy. The W score is a Z-score adjusted for age as a covariate, with the mean of the control group set at 0 and the standard deviation at 1. Therefore, a negative value indicates a degree of cortical atrophy higher than expected for cognitively normal older adults. The calculation formula is as follows: Where β1 is the age-related regression coefficient for each brain region, and β2 is the intercept value. For residual plots, This is the original value of the cortical thickness. This represents the expected value of cortical thickness in the control group corresponding to the patient's age. The standard deviation of the residuals.

[0025] White matter abnormalities, particularly damage to myelin and oligodendrocytes, accelerate cognitive impairment and pathological changes. Furthermore, cortical myelin damage occurs in the preclinical stage of Alzheimer's disease. Incorporating a T1w / T2w heterogeneity model yields more realistic spontaneous brain activity. Myelination, represented by the T1w / T2w ratio, is introduced as a priori constraint reflecting cortical hierarchical structure, calculated as follows: in, This represents the T1w / T2w ratio in brain region i.

[0026] In step S2, the local recurrent coupling strength of the model is parameterized based on the prior constraint values ​​obtained in step S1. The models of different brain regions are coupled through the structural connections obtained in step S1, and a whole-brain dynamic model is constructed through parameter optimization.

[0027] In disease-specific models, different brain regions exhibit varying degrees of recurrent connectivity. This is achieved through a linear combination of prior constraint values ​​for myelination and cortical atrophy.

[0028] The whole-brain dynamics model is a dynamic mean-field model, which is an effective simplification of the detailed spike model. It is used to describe the physiological firing patterns of neuronal groups and characterizes the dynamic properties of brain regions through coupled nonlinear differential equations. Nonlinear differential equations include: in, Represents the average synaptic gating variable. This represents the total input current to each brain region. Indicates the discharge rate. It is a structural connection matrix. It scales the global coupling strength to the connection strength of the structural connections. It is the strength of cyclic connectivity in the activity of neuronal groups within a brain region. Represents external input, excitability-gated variable , , d .

[0029] Furthermore, the simulated neural activity of each brain region was analyzed using a blood oxygen dynamics model. Converting it to a BOLD signal, the relevant differential equation is as follows: in, This triggers vasodilatory signals in the cortical region. Increase, accompanied by blood volume and deoxyhemoglobin content Changes, κ = 0.65, s⁻¹. γ = 0.41, s⁻¹. γ is the signal attenuation rate. This represents the blood oxygen kinetic time constant. , , , , , It is a set of parameter values ​​related to magnetic field strength.

[0030] The expectation-maximization algorithm is used for iterative optimization to maximize the goodness of fit between simulated and empirical static functional connections and dynamic functional connections. The Pearson correlation coefficient is used to quantify the goodness of fit of static functional connections, and the Kolmogorov-Smirnov distance is used to quantify the goodness of fit of dynamic functional connections.

[0031] For example, such as Figure 2 As shown. Figure 2 Figure A in the diagram illustrates how the model achieves optimal fit for both static and dynamic functional connections under optimal parameters. Figure 2 The B-figure in the diagram demonstrates through quantitative and qualitative analysis that the model can accurately reproduce static functional connections. Figure 2 The C diagrams in the document visualize the dynamic functional connections of empirical data, traditional homogeneous models, and the heterogeneous models provided in this application embodiment. The model provided in this application embodiment can reproduce the mode switching of dynamic functional activities of empirical data. Figure 2 The D-plot in the paper quantitatively compares the fit of traditional models, models with only myelin constraints, models with only atrophy constraints, and models provided in the embodiments of this application to static and dynamic functional connectivity.

[0032] S3: Train the variational autoencoder using static and dynamic functional connections with different groupings.

[0033] For example, different groups refer to constructing whole-brain dynamic models for healthy control groups and disease patients respectively, and optimizing the parameters to obtain the optimal model parameters. Then, each group of subjects undergoes 3000 simulations to obtain simulated static and dynamic functional connectivity, which is used to train the variational autoencoder so that the low-dimensional latent space extracted by the variational autoencoder can distinguish the brain states of different subject groups.

[0034] A variational autoencoder consists of a deep neural network and two fully connected layers. It comprises an encoder network, intermediate variational layers, and a decoder network. The encoder network dynamically maps the functional connections of the input to a Gaussian probability distribution in the latent space, while the decoder network reconstructs the input data from samples within the Gaussian probability distribution.

[0035] The loss function during model training consists of a standard reconstruction error term and a regularization term, as exemplified below: in, , These are the encoder and decoder model parameters, respectively. It is the difference between the reconstructed item and the corresponding input sample. is the regularization term, and is the Kullback-Leibler divergence between the latent space distribution and the standard Gaussian distribution, to ensure that meaningful outputs are generated after decoding and to maintain continuity between the output and the corresponding latent space value selection. Features of the same category can cluster together with high probability, which is used to find low-dimensional representations to encode the process of brain state changes.

[0036] S4: Apply different regulatory schemes to the whole-brain dynamics model constructed in step S2, obtain the brain state evolution trajectory induced by regulation through the variational autoencoder trained in step S3, and screen effective regulatory schemes.

[0037] In step S4, the modulation scheme includes applying external current stimulation to different brain regions of the model; the external current stimulation is a periodic pulse current, and different schemes are formed by adjusting the stimulation amplitude, frequency and pulse width.

[0038] For example, the applied external periodic pulsed current stimulation is as follows: Where A represents the amplitude of the stimulus, f represents the natural frequency of the node, and B represents the pulse width. The system is perturbed for 2 minutes, with the amplitude increasing from 0 to 1 in steps of 0.01 (each stimulus scheme is simulated 100 times independently).

[0039] After applying external electrical stimulation, brain functional activities simulated by a disease-specific whole-brain dynamics model are encoded, and the brain state evolution trajectory is plotted in a low-dimensional latent space using a variational autoencoder.

[0040] The study compares whether the brain state after stimulation can transition from a diseased state to a healthy state, and combines this with whether there is an improvement in nonlinear dynamic indicators and brain network properties between the stimulated and healthy control groups, ultimately selecting an effective regulatory strategy. The criteria for selecting an effective regulatory strategy include: the brain state after regulation transitions to the healthy brain state, and the nonlinear dynamic indicators and brain network properties after regulation are closer to those of the healthy group than before regulation.

[0041] This invention discloses an auxiliary analysis method for regulatory schemes based on a disease-specific whole-brain dynamics model, solving the problem of the lack of disease specificity in whole-brain dynamics models. By integrating multimodal magnetic resonance imaging data, regional specific features of myelination and cortical atrophy are extracted, enabling precise modeling of the pathological features of neurodegenerative diseases.

[0042] A parameterization method for local recurrent coupling strength based on prior constraints of myelination and cortical atrophy is proposed, which not only improves the model's ability to fit resting-state functional connectivity but also effectively captures dynamic pattern switching of brain functional activity. Deep learning methods were successfully used to capture brain states of both static and dynamic functional activity. Variational autoencoders were used to achieve unsupervised learning and low-dimensional representation of brain states, overcoming the limitation of traditional methods requiring a pre-defined number of states. By mapping the evolutionary trajectory of brain states in the latent space, the regulatory effect was intuitively quantified, providing a new dimension for optimizing neural modulation schemes.

[0043] Simultaneously, by applying virtual stimulation to disease-specific models and combining the analysis of nonlinear dynamic indicators and brain network properties, the study revealed the impact of neural modulation on brain functional activity, providing a theoretical basis for preclinical screening. For example, ... Figure 3 As shown, Figure 3 The left side of the middle section shows the brain state trajectories under different regulatory schemes. Figure 3 The right side of the image shows the corresponding disease-regulating targets and their locations obtained according to the method provided in the embodiments of this application. The method proposed in this invention can be applied to Alzheimer's disease, and through virtual intervention, it can discover key targets for disease regulation, which has important clinical significance for the treatment of Alzheimer's disease.

[0044] In the description of this specification, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0045] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for auxiliary analysis of regulatory schemes based on disease-specific whole-brain dynamics models, characterized in that, include: S1: Acquire multimodal magnetic resonance imaging (MRI) data and preprocess it to obtain structural connectivity, brain region BOLD time series, static functional connectivity, and dynamic functional connectivity. Based on the preprocessed data, calculate the prior constraint values ​​for myelination and cortical atrophy. The multimodal MRI data includes functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI), and T1w and T2w imaging. S2: Based on the prior constraint values ​​obtained in step S1, the local recurrent coupling strength of the model is parameterized. The models of different brain regions are coupled through the structural connections obtained in step S1, and a whole-brain dynamic model is constructed through parameter optimization. S3: Training variational autoencoders using static and dynamic functional connections with different groups; the variational autoencoders are used to capture the spatiotemporal characteristics of brain function and extract low-dimensional latent representations of brain states to distinguish different groups of brain states. S4: Apply different regulatory schemes to the whole-brain dynamics model constructed in step S2, obtain the brain state evolution trajectory induced by regulation through the variational autoencoder trained in step S3, and screen effective regulatory schemes.

2. The method for auxiliary analysis of regulatory schemes based on disease-specific whole-brain dynamics models according to claim 1, characterized in that, The preprocessing includes: processing multimodal magnetic resonance imaging data using a general automated processing pipeline based on a brain atlas template; wherein, functional magnetic resonance imaging data processing includes removing initial time points, correcting slice times and head movements, spatial smoothing and filtering to extract brain region BOLD time series; generating a structural connectivity matrix reflecting interregional connectivity through diffusion tensor imaging data processing; and extracting myelination and cortical thickness information through T1w and T2w imaging data processing.

3. The method for auxiliary analysis of regulatory schemes based on disease-specific whole-brain dynamics models according to claim 1, characterized in that, In step S1, according to The W score for cortical thickness was calculated for each subject, where β1 is the age-related regression coefficient for each brain region, and β2 is the intercept value. For residual plots, This is the original value of the cortical thickness. This represents the expected value of cortical thickness in the control group corresponding to the patient's age. The standard deviation of the residuals.

4. The method for auxiliary analysis of regulatory schemes based on disease-specific whole-brain dynamics models according to claim 1, characterized in that, In step S1, according to Myelination, represented by the T1w / T2w ratio, is introduced as a priori constraint reflecting the cortical hierarchy; where, This represents the T1w / T2w ratio in brain region i.

5. The method for auxiliary analysis of regulatory schemes based on disease-specific whole-brain dynamics models according to claim 1, characterized in that, In step S2, the parameterization of the local recursive coupling strength is based on This is achieved through a linear combination of prior constraint values ​​for myelination and cortical atrophy.

6. The method for auxiliary analysis of regulatory schemes based on disease-specific whole-brain dynamics models according to claim 1, characterized in that, In step S2, the whole-brain dynamics model is a dynamic mean-field model used to describe the physiological firing patterns of neuronal groups and characterizes the dynamic properties of brain regions through coupled nonlinear differential equations. The nonlinear differential equations include: in, Represents the average synaptic gating variable. This represents the total input current to each brain region. Indicates the discharge rate. It is a structural connection matrix. It scales the global coupling strength to the connection strength of the structural connections. It refers to the local recurrent coupling strength of neuronal group activity within a brain region. Represents external input, excitability-gated variable , , d .

7. The method for auxiliary analysis of regulatory schemes based on disease-specific whole-brain dynamics models according to claim 1, characterized in that, In step S3, the variational autoencoder includes an encoder network, an intermediate variational layer, and a decoder network; the encoder network dynamically maps the input functional connections and functional connections to a Gaussian probability distribution in the latent space, and the decoder network is used to reconstruct the input data from samples of the Gaussian probability distribution.

8. The method for auxiliary analysis of regulatory schemes based on disease-specific whole-brain dynamics models according to claim 1, characterized in that, In step S4, the modulation scheme includes applying external current stimulation to different brain regions of the model; the external current stimulation is a periodic pulse current, and different schemes are formed by adjusting the stimulation amplitude, frequency and pulse width.

9. The method for auxiliary analysis of regulatory schemes based on disease-specific whole-brain dynamics models according to claim 7, characterized in that, After applying external electrical stimulation, brain functional activities simulated by a disease-specific whole-brain dynamics model are encoded, and the brain state evolution trajectory is plotted in a low-dimensional latent space using a variational autoencoder.

10. The method for auxiliary analysis of regulatory schemes based on disease-specific whole-brain dynamics models according to claim 1, characterized in that, The criteria for selecting effective regulation schemes include: the brain state after regulation changes to the brain state of the healthy group, and the nonlinear dynamic indicators and brain network properties after regulation are closer to those of the healthy group than before regulation.

Images