Individualized neuro-modulation parameter optimization method based on tDCS stimulation response

Abstract

The application discloses an individualized neural regulation parameter optimization method based on a tDCS stimulation response model and belongs to the field of neural regulation scheme parameter optimization. The method solves the problems that a general tDCS stimulation scheme has high treatment effect heterogeneity on Alzheimer's disease patients, cannot be directly verified on the patients due to ethical constraints, and the response of whole brain neural activity to the tDCS stimulation is unclear. The method mainly reflects the influence of stimulation on whole brain neural activity through a large-scale dynamic model, simulates different neural regulation schemes by using a tDCS finite element modeling, solves the problem that the verification cannot be performed on the patients due to ethical constraints, and constructs a tDCS stimulation response model based on a subject level to effectively solve the problem of high treatment effect heterogeneity. Finally, the current dose parameters of the tDCS individualized neural regulation of the AD patients are optimized, the heterogeneity of the tDCS neural regulation among the subjects is reduced as much as possible, and the patient's condition is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of neural regulation parameter optimization, and particularly relates to an Alzheimer's disease individualized neural regulation parameter optimization method based on tDCS stimulation response. BACKGROUND

[0002] Alzheimer's disease (AD) is also known as senile dementia, and is mainly characterized by progressive loss of neurons, and its clinical features are progressive cognitive and behavioral dysfunction. AD has a high incidence, and there is no cure on the clinic, which causes serious harm to patients and society.

[0003] In recent years, transcranial direct current stimulation (tDCS) as a neural regulation method has been widely concerned in the field of neural rehabilitation at home and abroad, providing a new direction for the early intervention and treatment of AD. As a potential clinical non-invasive intervention method, tDCS can transmit a continuous low-intensity direct current to a specific area (stimulation target) of the brain by placing an electrode on the scalp, change the excitability of the cerebral cortical neurons by regulating the depolarization or hyperpolarization of the cell transmembrane potential, and produce a long-term potentiation / inhibition effect after repeated tDCS intervention, thereby achieving neural remodeling and improving the cognitive level of patients.

[0004] Although preliminary results have been achieved in the tDCS clinical trials of AD patients, the treatment effect is highly heterogeneous, and there are still many limitations. (1) The universal regulation scheme of tDCS has variability between subjects, and the regulation parameters such as electrode position, current intensity, number of stimulation sessions, and stimulation duration cannot be set according to experience, and an effective individual tDCS regulation scheme needs to be developed according to the patient's condition; (2) When verifying the parameters of the individualized tDCS regulation scheme, direct experiments on the human body cannot be conducted due to ethical restrictions; (3) The use of tDCS finite element modeling can be used to simulate the electric field intensity distribution of the whole brain after stimulation, and a communication bridge between the regulation scheme and brain neural activity is established. However, the finite element model is not enough to predict the immediate stimulation results, because in addition to the electric field distribution, the stimulation response effect also depends on cell specificity, local cortical circuit, brain connectivity and other physiological factors, and the response of whole brain neural activity to stimulation cannot be reflected.

[0005] Therefore, there is currently no suitable AD individualized tDCS neural regulation parameter, and it is impossible to verify whether the parameters of the tDCS regulation scheme are optimal directly on the human body. SUMMARY

[0006] The application provides an individualized neural regulation parameter optimization method based on tDCS stimulation response, which aims to solve the above problems.

[0007] The application is realized based on the individualized neural regulation parameter optimization method of tDCS stimulation response, which comprises the following steps:

[0008] Step S1, acquiring T1w, T2w, DTI and rs-fMRI multi-modal data of AD patients, and pre-processing the multi-modal data to obtain pre-processed multi-modal data;

[0009] Step S2, based on the pre-processed multi-modal data: finite element modeling is carried out to simulate the electric field distribution of tDCS stimulation in the brain of AD patients; at the same time, a large-scale dynamics model integrating synaptic plasticity mechanism is constructed;

[0010] Step S3: combining finite element modeling and large-scale dynamics model, constructing individualized tDCS stimulation response model;

[0011] Step S4: through the individualized tDCS stimulation response model, inversely calculating the individualized stimulation current intensity I best , completing the parameter optimization of current dose in the individualized tDCS regulation scheme of AD patients.

[0012] Further, in step S2, the large-scale dynamics model integrating synaptic plasticity mechanism is constructed, specifically including:

[0013] Step S21, dividing brain regions according to Desikan-Killiany atlas as nodes of network;

[0014] Step S22, constructing reconstructed empirical structural functional connection body FC(i,j) by using Pearson correlation method, the calculation formula is as follows:

[0015]

[0016] Wherein, x t and y t represent the signal values of two brain regions i and j time series at the moment; And represent the average values of brain region i and j time series respectively;

[0017] Step S23, modeling neuron population activity by using dynamic mean field model integrating synaptic plasticity, describing the neural field model of average neural activity of each brain region through mutually connected excitatory and inhibitory spike neural populations, and the dynamics of mutually connected brain regions is described by the following set of coupled nonlinear stochastic differential equations:

[0018]

[0019]

[0020]

[0021]

[0022] where x i , H(x i ), S i represent the total input current, average firing rate, average membrane potential of each population in cortical region i, respectively; Δw ij (t) represents the synaptic strength, where w ij is the synaptic weight representing synaptic efficacy; a i (t), a j (t) are the pre- and post-synaptic neuronal activity measures, c<<1 is a constant representing the learning rate; g(x, x0, v) represents the remodeling of synaptic structure between neurons, i.e. structural plasticity, which dynamically changes the network structure according to synaptic weight and homeostatic mechanisms, and is described by the following set of equations:

[0023]

[0024]

[0025]

[0026]

[0027]

[0028] Step S24, using the Balloon-Windkessel model to convert the simulated neural spike signals S 转 of each cortical region into BOLD signals:

[0029]

[0030] Step S25, using the expectation maximization algorithm to iteratively optimize, maximizing the degree of fitting between the simulated and empirical FCs, using the Pearson correlation coefficient R emp between the vector f emp of the empirical FC sim and the vector f sim of the simulated FC fitting to quantify the degree of fitting, the calculation formula is as follows:

[0031]

[0032] Further, in step S3, the individualized tDCS stimulation response model is constructed, specifically including:

[0033] Step S31, calculate the electric field intensity of each node parallel to the dendrite-axonal direction through finite element modeling Then calculate the bias voltage f(E tDCS ) through the lambda E model, wherein lambda belongs to the scaling factor of the voltage:

[0034]

[0035] Step S32, take the bias voltage f(E tDCS ) as an external stimulus in the large-scale dynamics model of the synaptic plasticity mechanism to induce changes in membrane potential:

[0036]

[0037] Step S33, fit the optimal parameters to calculate the computational synchrony, metastable state, BOLD signal and FC of tDCS.

[0038] Further, in step S4, specifically comprising:

[0039] Step S41, respectively construct the large-scale dynamics model of the synaptic plasticity of healthy subjects and AD patients;

[0040] Step S42, compare the nonlinear dynamics indexes and brain network attributes of healthy subjects and AD patients to mine the specificity of AD patients;

[0041] Step S43, simulate the electric field intensity of AD patients under different current dose stimulations on the F3 and Fp2 electrodes, simulate the current stimulation in the range of 0-6mA, and the step is 0.5;

[0042] Step S44, simulate the changes of the whole brain activity of AD patients under different current stimulation intensities through the tDCS stimulation response model;

[0043] Step S45, calculate the changes of AD patients in the specific indexes before tDCS stimulation and after different current dose stimulations, and reversely calculate the individualized optimal stimulation current intensity I best , complete the optimization of the current dose parameters of tDCS neuroregulation.

[0044] Compared with the prior art, the beneficial effects of the present application are: the present application provides a large-scale dynamics model of the synaptic plasticity mechanism reflecting the neuron population activity, constructs a tDCS stimulation response model by fusing tDCS finite element modeling and large-scale dynamics, simulates the neural activity of individuals in the whole brain under different neuroregulation schemes, effectively solves the heterogeneity of the universal neuroregulation scheme in the treatment effect, and completes the optimization of the individualized neuroregulation current dose parameters. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 Flow chart for individualized neuro-modulation parameter optimization of tDCS stimulation response modeling of the present application;

[0046] Figure 2 Technological diagram for tDCS stimulation response modeling of the present application;

[0047] Figure 3 Result diagram for global synchrony and metastable group difference and cognitive scale correlation of the present application;

[0048] Figure 4 Result diagram for individualized modulation of three AD patients on aLp under different stimulation intensities of the present application;

[0049] Figure 5 Result diagram for individualized modulation of three AD patients on aNe under different stimulation intensities of the present application. DETAILED DESCRIPTION

[0050] In order to make the objectives, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application in combination with the accompanying drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application, and are not used to limit the present application.

[0051] Please refer to Figures 1-2 The present application provides a technical solution: parameter optimization method of individualized neuro-modulation scheme for Alzheimer's disease based on tDCS stimulation response model, including the following steps:

[0052] Step S1: collecting T1w, T2w, DTI and rs-fMRI data of Alzheimer's disease patients and pre-processing the multi-modal data;

[0053] Step S2: constructing an individualized dynamic large-scale dynamics model based on the pre-processed data of S1, and innovatively integrating synaptic plasticity mechanism to predict the long-term neural effects of neuro-modulation.

[0054] Step S3: performing finite element modeling simulation based on the pre-processed data of S1 to calculate the electric field distribution under each network node

[0055] Step S4: combining the fusion synaptic plasticity large-scale dynamics model of S2 and the tDCS finite element modeling in S3 to construct an individualized tDCS stimulation response model;

[0056] Step S5: reverse calculating the individualized stimulation current intensity I through the tDCS stimulation response modeling technology in S4 best , completing the parameter optimization of current dosage in the individualized tDCS modulation scheme for AD patients.

[0057] The large-scale dynamic modeling process of the fusosynaptic plasticity mechanism in step S2 is as follows:

[0058] (1) According to the Desikan-Killiany atlas, the brain regions are divided as nodes of the network;

[0059] (2) The Pearson correlation method is used to construct the reconstructed empirical structure functional connectivity (FC) matrix FC(i, j);

[0060] (3) The dynamic mean field model (DMF) of fusosynaptic plasticity is used to model the population activity of neurons;

[0061] (4) The Balloon-Windkessel model is used to convert the simulated neural pulse signal S i of each cortical region into a BOLD signal;

[0062] (5) The expectation maximization (EM) algorithm is used to iteratively optimize and maximize the fitting degree R fitting between the simulation and the empirical FC.

[0063] In step S4, the fusosynaptic plasticity large-scale dynamic model of step S2 and the tDCS finite element modeling based on step S3 are combined to construct a tDCS stimulation response model, and the specific steps are as follows:

[0064] (1) Based on step S3, the electric field strength parallel to the dendrite-axonal direction of each node is calculated The bias voltage f(E tDCS ) is calculated through the λE model;

[0065] (2) The bias voltage f(E tDCS ) calculated in (1) is used as the change of the external stimulus-induced membrane potential in the dynamic mean field model in step S2;

[0066] (3) The synchrony, metastability, BOLD signal and FC after tDCS stimulation are calculated according to the optimal parameters fitted in step S2.

[0067] In step S5, based on the tDCS stimulation response modeling technology in step S4, the individualized stimulation current intensity I best is calculated, and the specific method for optimizing the individualized tDCS regulation parameter current dose for AD patients is as follows:

[0068] The large-scale dynamics model of S2-based synaptic plasticity mechanism calculates the synchrony, metastability, BOLD signal and FC of healthy subjects and AD patients.

[0069] The S4-based tDCS stimulation response model calculates the nonlinear dynamics indicators of AD patients under different current dose stimulation.

[0070] The change relationship of the nonlinear dynamics indicators of AD patients under different tDCS stimulation intensities is calculated, so as to inversely calculate the individualized optimal current stimulation intensity I best The individualized tDCS regulation current dose parameters of the AD patient are optimized.

[0071] The present application verifies the reliability of the tDCS individualized neural regulation current dose parameter optimization on the ADNI dataset. First, the nonlinear dynamics indicators and brain network properties of the AD patients and healthy subjects in the ADNI dataset are compared; then, the large-scale dynamics model based on the ADNI dataset is constructed; then, the tDCS finite element modeling electric field strength is used to construct the tDCS stimulation response model; finally, the relationship between the stimulation intensity and the indicators is calculated, and the current dose parameter optimization is completed.

[0072] Specifically, step S1, acquiring T1w, T2w, DTI and rs-fMRI multi-modal data of AD patients, and pre-processing the multi-modal data to obtain pre-processed multi-modal data;

[0073] Step S2, based on the pre-processed multi-modal data: finite element modeling is carried out to simulate the electric field distribution of tDCS stimulation in the brain of AD patients; at the same time, a large-scale dynamics model of synaptic plasticity mechanism is constructed;

[0074] Step S3: combining finite element modeling and large-scale dynamics model, constructing individualized tDCS stimulation response model;

[0075] Step S4: through the individualized tDCS stimulation response model, inversely calculating the individualized stimulation current intensity I best The current dose parameter optimization of the individualized tDCS regulation scheme of the AD patient is completed.

[0076] In step S2, the large-scale dynamics model of the synaptic plasticity mechanism is constructed, which specifically includes:

[0077] Step S21, according to the Desikan-Killiany atlas, 68 brain regions are divided as nodes of the network;

[0078] Step S22, the Pearson correlation method is used to construct the reconstructed empirical structural functional connection body FC(i,j), and the calculation formula is as follows:

[0079]

[0080] where x t and y t denote the signal values of the time series of two brain regions i and j at time t; and denote the mean values of the time series of brain regions i and j, respectively;

[0081] Step S23, constructing a dynamic mean field model of the fused synaptic plasticity mechanism, modeling the activity of the neuronal population by a mean field model of neural activity that represents each brain region by a population of mutually connected excitatory and inhibitory spiking neurons, the dynamics of the interconnected brain regions are described by the following set of coupled nonlinear stochastic differential equations:

[0082]

[0083]

[0084]

[0085] Δw ij (t) = a i (t) [ca j (t) - w ij (t)]

[0086] where x i , H(x i ), S i denote the total input current, the average firing rate, and the average membrane potential of each population in cortical region i, respectively; Δw ij (t) denotes the synaptic strength, where w ij is the synaptic connection weight representing the synaptic efficacy; a i (t), a j (t) are measures of presynaptic and postsynaptic neuronal activity, such as firing rate, calcium concentration, etc.; c<<1 is a constant representing the learning rate; g(x, x0, v) represents the remodeling of the synaptic structure between neurons, i.e., structural plasticity, which dynamically changes the network structure according to synaptic weights and homeostatic mechanisms, and is described by the following set of equations:

[0087]

[0088]

[0089]

[0090]

[0091]

[0092] Step S24, using the Balloon-Windkessel model to simulate the neural pulse signal S i of each cortical region into BOLD signal:

[0093]

[0094] Step S25, using the expectation maximization algorithm to iteratively optimize, maximize the fitting degree between the simulated and empirical FCs, using the empirical FC emp vector f emp and the simulated FC sim vector f sim to quantify the fitting degree R fitting between the Pearson correlation coefficient, the calculation formula is as follows:

[0095]

[0096] In step S3, based on the large-scale dynamics model of the synaptic plasticity mechanism and the tDCS finite element modeling, an individual tDCS stimulation response model is constructed, which specifically includes:

[0097] Step S31, through finite element modeling, the electric field intensity of each node parallel to the dendrite-axonal direction is calculated Then the bias voltage f(E tDCS ) is calculated through the λE model, where λ is the scaling factor of the voltage:

[0098]

[0099] Step S32, the bias voltage f(E tDCS ) is used as an external stimulus in the large-scale dynamics model of the synaptic plasticity mechanism to induce changes in membrane potential:

[0100]

[0101] Step S33, the optimal parameters are fitted to calculate the computational synchrony, metastable state, BOLD signal and FC of tDCS.

[0102] In step S4, based on the data of AD patients and healthy subjects, the individualized stimulation current intensity I best is calculated by using the tDCS stimulation response modeling technology, and the current dose parameters of the individualized tDCS neuromodulation of AD are optimized, which specifically includes:

[0103] Step S41, the large-scale dynamics model of the synaptic plasticity of healthy subjects and AD patients is constructed respectively;

[0104] Step S42, comparing the nonlinear dynamics index and the brain network attribute of the healthy subject and the AD patient, and mining the specificity of the AD patient;

[0105] Step S43, simulating the electric field intensity of the AD patient under the stimulation of different current doses at the F3 and Fp2 electrodes, simulating the current stimulation in the range of 0-6 mA, and the step is 0.5;

[0106] Step S44, simulating the change of the whole brain activity of the AD patient under different current stimulation intensities through the tDCS stimulation response model;

[0107] Step S45, calculating the change of the specificity index of the AD patient before tDCS stimulation and after stimulation under different current doses, and reversely calculating the individualized optimal stimulation current intensity I best , and completing the optimization of the current dose parameter of the tDCS neural regulation.

[0108] Experimental results: The reliability of the tDCS stimulation response model is verified by simulating the FC fitting degree of the AD patient and the healthy subject in the ADNI data set on the large-scale dynamics model, the metastable state and the synchronism, and the results are shown in the following table.

[0109] As shown in the following table, there are significant differences in the synchronism and the metastable state of the AD, MCI and healthy subjects based on the large-scale dynamics model. Figure 3

[0110] As shown in the following table, the results show that the shortest path length after the tDCS stimulation based on the tDCS stimulation response model is shortened, and the network attribute results and the stimulation intensity are not in a linear relationship. Figure 4 As shown in the following table, the results show that the network efficiency after the tDCS stimulation is improved.

[0111] Figure 5 As shown in the following table, the results show that the network efficiency after the tDCS stimulation is improved.

[0112] It is proved that the optimization of the current dose parameter of the tDCS neural regulation based on the tDCS stimulation response model is reliable.

[0113] The above is only a preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.​​

Claims

1. A method for individualized neuro-modulation parameter optimization based on tDCS stimulation response, characterized in that, The method comprises the following steps: Step S1, acquiring T1w, T2w, DTI and rs-fMRI multi-modal data of an AD patient, and pre-processing the multi-modal data to obtain pre-processed multi-modal data; Step S2, based on the pre-processed multi-modal data: performing finite element modeling to simulate the electric field distribution of tDCS stimulation in the brain of the AD patient; at the same time, constructing a large-scale dynamic model integrating synaptic plasticity mechanism; Step S3: combining the finite element modeling and the large-scale dynamic model, constructing an individualized tDCS stimulation response model; Step S4: inversely calculating the individualized stimulation current intensity I through the individualized tDCS stimulation response model best , completing the parameter optimization of the current dose in the individualized tDCS regulation scheme for AD patients; In step S2, the large-scale dynamic model integrating synaptic plasticity mechanism is constructed, specifically including: Step S21, dividing brain regions according to the Desikan-Killiany atlas as nodes of the network; Step S22, constructing a reconstructed empirical structural functional connection body FC(i,j) by using the Pearson correlation method, and the calculation formula is as follows: where x t and y t denote the signal values of the time series of the two brain regions i and j at time t; and denote the mean values of the time series of the brain regions i and j, respectively; Step S23, modeling the neuron population activity by using the dynamic mean field model integrating synaptic plasticity, describing the neural field model of the average neural activity of each brain region through the excitatory and inhibitory spiking neural populations connected with each other, and the dynamics of the brain regions connected with each other are described by the following set of coupled nonlinear stochastic differential equations: Δw ij (t) = a i (t) [ca j (t) - w ij (t)] where x i , H(x i ), S i represent the input total current, average firing rate, average membrane potential of each population in cortical region i, respectively; Δw ij (t) represents the synaptic strength, where w ij is the synaptic interconnection weight representing synaptic efficacy; a i (t), a j (t) are the pre- and post-synaptic neuronal activity measures, c « 1 is a constant representing the learning rate; g(x, x0, v) represents the remodeling of the synaptic structure between neurons, i.e. structural plasticity, which dynamically changes the network structure according to synaptic weights and homeostatic mechanisms, and is described by the following set of equations: Step S24, using the Balloon-Windkessel model to convert the simulated neural spike signals S i into BOLD signals: Step S25, iterative optimization using expectation-maximization algorithm, maximizing the degree of fitting between simulated and empirical FCs, using the empirical FCs emp vector f emp and the simulated FCs sim vector f sim between the Pearson correlation coefficient R fitting quantifying the degree of fitting, is calculated as follows:

2. The individualized neuromodulation parameter optimization method based on tDCS stimulation response according to claim 1, characterized in that, In step S3, the individualized tDCS stimulation response model is constructed, specifically including: Step S31, calculate the electric field intensity of each node parallel to the dendrite-axonal direction through finite element modeling Then calculate the bias voltage f(E tDCS ) through the λE model, where λ belongs to the scaling factor of the voltage: Step S32, the bias voltage f(E tDCS ) as the external stimulus in the large-scale dynamics model of the fusion synaptic plasticity mechanism, evokes changes in membrane potential: Step S33, fitting the optimal parameters to calculate the computational synchrony, metastable state, BOLD signal and FC of tDCS.

3. The tDCS stimulation response based individualized neuromodulation parameter optimization method of claim 1, wherein, In step S4, specifically including: Step S41, constructing a large-scale dynamic model integrating synaptic plasticity for healthy subjects and AD patients respectively; Step S42, comparing the nonlinear dynamics indexes and brain network attributes of healthy subjects and AD patients to mine the specificity of AD patients; Step S43, simulating the electric field intensity of the AD patient under different current dose stimulations on the F3 and Fp2 electrodes, simulating the current stimulation in the range of 0-6mA with a step of 0.5; Step S44, simulating the changes of the whole brain activity of the AD patient under different current stimulation intensities by the tDCS stimulation response model; Step S45, the changes in the specific indicators of the AD patient before tDCS stimulation and after stimulation at different current doses are calculated, and the individualized optimal stimulation current intensity I is calculated in reverse best , the optimization of the current dose parameters of tDCS neuromodulation is completed.

Images