Method and device for determining individualized transcranial stimulation parameters based on target regulatory effects

Abstract

The application provides a method and device for determining individual transcranial stimulation parameters based on target regulation effects, and relates to the technical field of intelligent medical treatment. The method comprises the following steps: determining brain function target regulation data based on the target regulation effect; inputting the brain function target regulation data and corresponding target key physiological parameters into a field characteristic determination model to obtain field characteristic parameters output by the field characteristic determination model, wherein the field characteristic parameters are used to realize the brain function target regulation data; the field characteristic determination model is trained based on sample brain function target regulation data, sample subject key physiological parameters and sample field characteristic parameters; and determining corresponding transcranial stimulation parameters based on the field characteristic parameters, which can accurately determine individual stimulation parameters to realize the expected brain function target regulation data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart medical treatment, and in particular to a method and device for determining individual transcranial stimulation parameters based on target regulation effect. BACKGROUND

[0002] Brain function regulation techniques, such as transcranial magnetic stimulation (TMS) and transcranial electrical stimulation (tES), play an important role in neuroscience research and clinical treatment, and are widely used in the intervention of diseases such as depression, Parkinson's disease, and stroke rehabilitation. Previous studies have shown that the effects of brain function regulation techniques vary significantly among different individuals, which is a major reason limiting their widespread application. Currently, optimization research on brain function regulation techniques mainly focuses on improving the intensity or focusing degree of the electric field distribution in the target region. For example, a convex optimization algorithm is used to calculate the optimal position of the electrode and the stimulation intensity corresponding to the electrode in multi-channel transcranial electrical stimulation, with the area of the target stimulation target point exceeding a certain electric field intensity threshold as the objective function. However, the intensity or focusing degree of the electric field distribution in the target brain region is not necessarily directly positively correlated with the final brain function target regulation data. The target regulation data of brain regulation techniques on specific individuals should also be related to specific tasks and individual physiological parameter differences.

[0003] Therefore, for specific application scenarios, how to accurately determine individualized stimulation parameters to achieve the expected brain function target regulation data is an important issue that needs to be addressed in the industry. SUMMARY

[0004] To solve the problems in the prior art, the present application provides a method and device for determining individual transcranial stimulation parameters based on target regulation effect.

[0005] The present application provides a method for determining individual transcranial stimulation parameters based on target regulation effect, comprising:

[0006] determining brain function target regulation data based on the target regulation effect;

[0007] inputting the brain function target regulation data and the corresponding target key physiological parameters into a field feature determination model to obtain field feature parameters output by the field feature determination model, the field feature parameters being used to achieve the brain function target regulation data; the field feature determination model being trained based on sample brain function target regulation data, sample subject key physiological parameters, and sample field feature parameters;

[0008] determining corresponding transcranial stimulation parameters based on the field feature parameters.

[0009] According to the method for determining individual transcranial stimulation parameters based on target regulation effect provided by the application, the field feature determination model is trained based on sample brain function target regulation data, sample subject key physiological parameters and sample field feature parameters, and the method comprises the following steps:

[0010] The sample field feature parameters are input into a forward module of the original field feature determination model to obtain Gaussian noise data output by the forward module; the Gaussian noise data is obtained by gradually adding noise to the sample field feature parameters and performing forward diffusion;

[0011] The Gaussian noise data is input into a reverse module of the original field feature determination model to obtain predicted field feature parameters recovered by the reverse module; the predicted field feature parameters are obtained by performing a reverse process on the Gaussian noise data;

[0012] A loss function of the predicted field feature parameters and the sample field feature parameters is determined, and parameters of the original field feature determination model are adjusted by minimizing the loss function;

[0013] The step of inputting the sample field feature parameters into the forward module of the original field feature determination model is repeated until the original field feature determination model converges, and the field feature determination model is obtained;

[0014] The sample brain function target regulation data and the corresponding subject key physiological parameters are control conditions of the reverse process.

[0015] According to the method for determining individual transcranial stimulation parameters based on target regulation effect provided by the application, the field feature determination model is trained based on sample brain function target regulation data, sample subject key physiological parameters and sample field feature parameters, and the method comprises the following steps:

[0016] The sample brain function target regulation data and the sample field feature parameters are input into a pre-training matching model of the original field feature determination model to obtain an encoding vector of the sample field feature parameters output by the pre-training matching model;

[0017] The encoding vector is input into a forward module of the original field feature determination model to obtain Gaussian noise data output by the forward module; the Gaussian noise data is obtained by gradually adding noise to the encoding vector and performing forward diffusion;

[0018] The Gaussian noise data is input into a reverse module of the original field feature determination model to obtain a predicted field feature parameter encoding vector recovered by the reverse module; the predicted field feature parameter encoding vector is obtained by performing a reverse process on the Gaussian noise data;

[0019] determining a loss function of the predicted field characteristic parameter encoding vector and the encoding vector of the sample field characteristic parameter, and adjusting parameters of the original field characteristic determination model by minimizing the loss function;

[0020] repeating the step of inputting the sample brain function target regulation data and the sample field characteristic parameter into the pre-training matching model of the original field characteristic determination model until the original field characteristic determination model converges, to obtain the field characteristic determination model;

[0021] The sample brain function target regulation data and the corresponding subject key physiological parameter are control conditions of the reverse process, and the field characteristic determination model comprises a decoding module configured to decode the predicted field characteristic parameter encoding vector output by the reverse module to obtain the predicted field characteristic parameter.

[0022] According to the method for determining individual transcranial stimulation parameters based on target regulation effects provided by the application, before the field characteristic determination model is trained, the method further comprises:

[0023] determining structural image data, and constructing a corresponding individual head model based on the structural image data;

[0024] determining a plurality of groups of transcranial stimulation parameters based on the individual head model, wherein each group of the transcranial stimulation parameters is used to realize one kind of transcranial stimulation;

[0025] performing simulation calculation to obtain sample field characteristic parameters in the cerebral cortex according to the transcranial stimulation parameters and the individual head model;

[0026] inputting the sample field characteristic parameters into a training data acquisition model to obtain sample brain function target regulation data output by the training data acquisition model, wherein the training data acquisition model is trained based on experimental sample field characteristic parameters and experimental sample brain function target regulation data.

[0027] According to the method for determining individual transcranial stimulation parameters based on target regulation effects provided by the application, the training data acquisition model is trained based on experimental sample field characteristic parameters and experimental sample brain function target regulation data, and comprises:

[0028] inputting the experimental sample field characteristic parameters into an original training data acquisition model to obtain predicted brain function target regulation data output by the original training data acquisition model;

[0029] calculating a loss function based on the predicted brain function target regulation data and the experimental sample brain function target regulation data, and adjusting parameters of the original training data acquisition model according to the loss function;

[0030] The step of inputting the experimental sample field characteristic parameter into the original training data acquisition model is repeated until the original training data acquisition model converges, and the training data acquisition model is obtained.

[0031] According to the method for determining individual transcranial stimulation parameters based on target regulation effects provided by the application, before the experimental sample field characteristic parameter is input into the original training data acquisition model, the method further comprises:

[0032] The sample field characteristic parameter is subjected to mask processing to obtain a mask field characteristic parameter.

[0033] The mask field characteristic parameter is input into an original feature acquisition model to obtain a predicted field characteristic parameter output by the original feature acquisition model.

[0034] A second loss function between the sample field characteristic parameter and the predicted field characteristic parameter is determined, and the parameters of the original feature acquisition model are adjusted by minimizing the second loss function.

[0035] The step of subjecting the sample field characteristic parameter to mask processing to obtain a mask field characteristic parameter is repeated until the original feature acquisition model converges, and the original feature acquisition model is obtained.

[0036] A linear layer is added to the original feature acquisition model to obtain the original training data acquisition model.

[0037] The application further provides a device for determining individual transcranial stimulation parameters based on target regulation effects, comprising:

[0038] A target regulation data determination module is configured to determine brain function target regulation data based on the target regulation effect.

[0039] A field characteristic parameter determination module is configured to input the brain function target regulation data and corresponding target key physiological parameters into a field characteristic determination model to obtain field characteristic parameters output by the field characteristic determination model, wherein the field characteristic parameters are used to realize the brain function target regulation data; the field characteristic determination model is trained based on sample brain function target regulation data, sample subject key physiological parameters and sample field characteristic parameters.

[0040] A stimulation parameter determination module is configured to input the field characteristic parameters as targets into an optimization module to obtain transcranial stimulation parameters, wherein the transcranial stimulation parameters are used to realize the field characteristic parameters; the optimization module can refer to an existing multi-channel linear convex optimization algorithm.

[0041] The application further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for determining individualized transcranial stimulation parameters based on target regulation effect according to any one of the above.

[0042] The application further provides a non-transitory computer readable storage medium, which stores a computer program, wherein the computer program is executed by a processor to implement the method for determining individualized transcranial stimulation parameters based on target regulation effect according to any one of the above.

[0043] The application further provides a computer program product, comprising a computer program, wherein the computer program is executed by a processor to implement the method for determining individualized transcranial stimulation parameters based on target regulation effect according to any one of the above.

[0044] The application provides the method and device for determining individualized transcranial stimulation parameters based on target regulation effect, which captures the relationship between brain function target regulation data, target key physiological parameters and field characteristic parameters by a field characteristic determination model, and outputs field characteristic parameters, so that the corresponding transcranial stimulation parameters are determined by combining the field characteristic parameters obtained by the target key physiological parameters, the individualized stimulation parameters are accurately determined, and the expected brain function target regulation data is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0046] Figure 1 is a flowchart of the method for determining individualized transcranial stimulation parameters based on target regulation effect provided by the application.

[0047] Figure 2 is a structural schematic diagram of the device for determining individualized transcranial stimulation parameters based on target regulation effect provided by the application.

[0048] Figure 3 is a structural schematic diagram of the electronic device provided by the application. DETAILED DESCRIPTION

[0049] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below in conjunction with the drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0050] The present application will be described below in conjunction with Figures 1-3 The method and device for determining individual transcranial stimulation parameters based on target regulation effect of the present application can be used in the field of brain-computer interface and neural regulation.

[0051] Figure 1 The present application provides a flowchart of the method for determining individual transcranial stimulation parameters based on target regulation effect, as shown in Figure 1 The method comprises the following steps:

[0052] Step 101, determining brain function target regulation data based on the target regulation effect.

[0053] The brain function target regulation data refers to the target regulation effect of the expected change or expected enhancement of brain function produced by non-invasive neural regulation means such as transcranial stimulation. For example, the brain function target regulation data can be to realize the improvement of cognitive ability, or to realize the relief of disease symptoms, etc.

[0054] Step 102, inputting the brain function target regulation data and the corresponding target key physiological parameter into a field feature determination model to obtain the field feature parameter output by the field feature determination model, the field feature parameter being used to realize the brain function target regulation data; the field feature determination model being trained based on sample brain function target regulation data, sample subject key physiological parameter and sample field feature parameter.

[0055] The target key physiological parameter refers to the key physiological parameter of the individual who needs to be regulated in brain function. The key physiological parameter refers to the preset physiological characteristic that can affect the brain function target regulation data, such as the head structure parameter, etc.

[0056] The field feature parameter can be a specific field intensity distribution, such as electric field distribution, magnetic field distribution or acoustic field distribution, etc., or a distribution feature obtained based on the field intensity distribution, such as focusing degree, extreme value, etc.

[0057] The sample brain function target regulation data, the sample subject key physiological parameter, and the sample field characteristic parameter can be a corresponding set of parameters obtained based on the experiment in which the subject participates. Specifically, the sample brain function target regulation data can be an expected change or an expected enhancement effect on the brain function of the subject participating in the experiment, the sample subject key physiological parameter can be a preset physiological characteristic of the subject participating in the experiment that can affect the brain function target regulation data, and the sample field characteristic parameter can be a field characteristic parameter that can generate the sample brain function target regulation data.

[0058] For example, in the process of training the field characteristic determination model, the brain function target regulation data can be determined based on actual research conditions, for example, the brain function target regulation data can be enhancement of visual cognitive function, the sample subject key physiological parameter can be determined according to the determined enhancement of visual cognitive function, for example, the head structure parameter of the subject; and the field characteristic parameter for enhancing the visual cognitive function of the subject participating in the experiment can be determined through a specific experiment, so as to obtain the sample field characteristic parameter.

[0059] Step 103, determining corresponding transcranial stimulation parameters based on the field characteristic parameter.

[0060] The transcranial stimulation parameter refers to a specific configuration parameter of the transcranial stimulation required for realizing the field intensity distribution corresponding to the brain function target regulation data.

[0061] For example, based on the selected specific non-invasive neural regulation means, the field intensity distribution can be an electric field intensity distribution, a magnetic field intensity distribution, or an acoustic field intensity distribution, and the corresponding transcranial stimulation parameter can be a transcranial electric stimulation parameter, a transcranial magnetic stimulation parameter, or a transcranial ultrasonic stimulation parameter.

[0062] For example, the sample field intensity distribution and the sample transcranial stimulation parameter can be pre-acquired, a lead field matrix and a target optimization function can be constructed based on the sample field intensity distribution and the sample transcranial stimulation parameter, and a corresponding transcranial stimulation parameter can be obtained by solving the target optimization function by using a numerical optimization method such as convex optimization; or an optimization module can be constructed based on a multi-channel linear convex optimization algorithm, the field characteristic parameter can be input into the optimization module as a target, and a transcranial stimulation parameter can be obtained, wherein the transcranial stimulation parameter is used to realize the field characteristic parameter.

[0063] The method for determining individualized transcranial stimulation parameters based on target regulation effects provided by the embodiment of the application can capture the relationship among the brain function target regulation data, the target key physiological parameter, and the field characteristic parameter by using the field characteristic determination model to output the field characteristic parameter. In this way, the corresponding transcranial stimulation parameter can be determined by combining the field characteristic parameter obtained by the target key physiological parameter, the individualized stimulation parameter can be accurately determined, and the expected brain function target regulation data can be realized.

[0064] Based on the above embodiments, the field feature determination model is trained based on sample brain function target regulation data, key physiological parameters of sample subjects, and sample field feature parameters, including:

[0065] The sample field feature parameters are input into the forward module of the original field feature determination model to obtain Gaussian noise data output by the forward module; the Gaussian noise data is obtained by forward diffusion of noise gradually added to the sample field feature parameters.

[0066] The Gaussian noise data is input into the inverse module of the original field feature determination model to obtain the predicted field feature parameters recovered by the inverse module; the predicted field feature parameters are obtained by inverting the Gaussian noise data.

[0067] Determine the loss function for the predicted field feature parameters and the sample field feature parameters, and adjust the original field features by minimizing the loss function to determine the parameters of the model;

[0068] Repeat the step of inputting the sample field feature parameters into the forward module of the original field feature determination model until the original field feature determination model converges to obtain the field feature determination model;

[0069] Among them, the target regulation data of brain function of the sample and the corresponding key physiological parameters of the subjects are the control conditions of the reverse process.

[0070] Specifically, the forward module, also known as the forward module or the precursor module, is used to implement the forward process in the field feature determination model. For example, based on parameterized Markov chain rules, the input sample brain function target regulation data can be progressively transformed into Gaussian noise data. This process can be represented by the following formula:

[0071]

[0072] in, Forward process, These are the input sample field feature parameters. The data contains Gaussian noise. For the first t Noise data of the step, It follows a Gaussian distribution. The noise scheduling parameters added at each step of the forward diffusion process determine the amount of noise added at each step. I It is the identity matrix. This represents the total number of steps in the forward process.

[0073] The variance is used to control the intensity of the noise.

[0074] In this way, the sample brain function target regulation data is gradually added with noise in the iteration of the number of steps of forward diffusion, gradually loses its original distinctive features, and finally when T tends to infinity, is equivalent to a noise conforming to an isotropic Gaussian distribution.

[0075] In an embodiment, the different time steps are predefined, may be linear decay, exponential decay, etc., satisfying .

[0076] The reverse module, also known as the backward module, the after-trend module, etc., is used to realize the inverse process in the field feature determination model. Illustratively, based on the parameterized Markov chain rule and the control condition of the inverse process, the Gaussian noise data can be gradually reduced in noise and gradually approach the sample field feature parameters. Specifically, a neural network can be used to fit the inverse process.

[0077]

[0078] wherein, c is a control condition, i.e., brain function target regulation data and corresponding key physiological parameters of the subject, used to guide the generation of specific features, is the number of steps of reverse diffusion, is the true reverse mean, which depends on and , is the variance coefficient of the true reverse process, which is related to the noise scheduling parameter , is the neural network fitting mean, is the noise scheduling parameter in the diffusion process, is the cumulative product of , is the noise item predicted by the neural network.

[0079] Illustratively, controls the proportion of original information retained by the current step data, and its value is between 0 and 1; represents the overall decay degree of the original signal from the initial data to the step, used to adjust the degree of dependence on noise when generating data; is an estimate of the noise data at the current step under the control condition c and the number of reverse diffusion steps , used to approximate the addition of the original sample field feature parameters actual noise so as to gradually denoise in the reverse process, and generate target data meeting the control condition c.

[0080] Exemplarily, a mean square error formula can be designed to define a loss function between the predicted field characteristic parameter and the sample field characteristic parameter. In this way, the parameters of the field characteristic determination model can be adjusted by minimizing the mean square error formula using the gradient descent method. When the loss function converges to a preset threshold, for example, the value of the loss function is less than 0.001, or the maximum number of iterations is repeated, for example, 1000 times, it is determined that the original field characteristic determination model converges, and the field characteristic determination model is obtained.

[0081] The brain function target regulation data that can be achieved by the field characteristic parameter generated by the transcranial stimulation parameter has individual differences. In the embodiment, by taking the sample brain function target regulation data and the corresponding subject key physiological parameters as control conditions, the original field characteristic determination model of the model can capture the influence of individual differences on the field characteristic parameter, and the accuracy of the determined field characteristic parameter facing individualized brain function regulation can be improved.

[0082] Specifically, in the reverse process, the sample brain function target regulation data and the corresponding subject key physiological parameters are taken as control conditions, so that the predicted field characteristic parameter obtained by denoising is highly correlated with the sample brain function target regulation data; the sample brain function target regulation data and the corresponding subject key physiological parameters run through the reverse process, so that the predicted field characteristic parameter obtained is coupled with individualized characteristics, and the accuracy of the determined field characteristic parameter facing individualized brain function regulation can be improved.

[0083] Based on any of the above embodiments, the field characteristic determination model is trained based on the sample brain function target regulation data, the sample subject key physiological parameters and the sample field characteristic parameter, and includes:

[0084] The sample brain function target regulation data and the sample field characteristic parameter are input into a pre-training matching model of the original field characteristic determination model, to obtain an encoding vector of the sample field characteristic parameter output by the pre-training matching model;

[0085] The encoding vector is input into a forward module of the original field characteristic determination model, to obtain Gaussian noise data output by the forward module; the Gaussian noise data is obtained by adding noise to the encoding vector and performing forward diffusion;

[0086] The Gaussian noise data is input into a reverse module of the original field characteristic determination model, to obtain a predicted field characteristic parameter encoding vector recovered by the reverse module; the predicted field characteristic parameter encoding vector is obtained by performing a reverse process on the Gaussian noise data;

[0087] determining a loss function of the encoding vector of the predicted field characteristic parameter and the encoding vector of the sample field characteristic parameter, and adjusting parameters of the original field characteristic determination model by minimizing the loss function;

[0088] repeating the step of inputting the sample brain function target regulation data and the sample field characteristic parameter into the pre-training matching model of the original field characteristic determination model until the original field characteristic determination model converges, to obtain the field characteristic determination model;

[0089] The sample brain function target regulation data and the corresponding subject key physiological parameters are control conditions of the reverse process. The field characteristic determination model comprises a decoding module configured to decode the predicted field characteristic parameter encoding vector output by the reverse module to obtain a predicted field characteristic parameter.

[0090] The pre-training matching model can use contrastive learning to encode the sample brain function target regulation data and the sample field characteristic parameter to obtain an encoding vector of the sample brain function target regulation data and an encoding vector of the sample field characteristic parameter, so as to learn the mapping relationship thereof in a common embedding space and obtain key correlation features.

[0091] For example, the pre-training matching model can be a CLIP (Contrastive Language-Image Pre-Training) contrastive learning model, and the decoding module can be a contrastive learning decoder.

[0092] The working principle and technical effects of training the field characteristic determination model in this embodiment are basically the same as those in the foregoing embodiments, and will not be repeated here; the difference is that in this embodiment, the sample brain function target regulation data and the sample field characteristic parameter are input into the pre-training matching model of the original field characteristic determination model before being input into the forward module, to obtain an encoding vector of the sample field characteristic parameter generated by using contrastive learning to encode the sample brain function target regulation data and the sample field characteristic parameter, which can better capture the key correlation features between the two types of data.

[0093] Compared with directly training the original field characteristic determination model using the sample brain function target regulation data and the sample field characteristic parameter, using the encoding vector can provide more consistent and standardized input, and learn the correlation features of the two in the encoding space, thereby increasing the convergence speed of the original field characteristic determination model in the early training stage.

[0094] Moreover, the pre-training matching model can convert the high-dimensional sample field strength distribution into a low-dimensional encoding vector, which can reduce the computational overhead required for training the field characteristic determination model by dimension reduction, and speed up the training process of the field characteristic determination model.

[0095] Based on any of the above embodiments, before the field feature determination model is trained, the method further comprises:

[0096] determining structural image data, and constructing a corresponding individualized head model based on the structural image data;

[0097] determining a plurality of transcranial stimulation parameters based on the individualized head model; each set of the transcranial stimulation parameters is used to implement a transcranial stimulation;

[0098] performing simulation calculation to obtain sample field feature parameters in the cerebral cortex according to the transcranial stimulation parameters and the individualized head model;

[0099] inputting the sample field feature parameters into a training data acquisition model to obtain sample brain function target regulation data output by the training data acquisition model; the training data acquisition model is trained based on experimental sample field feature parameters and experimental sample brain function target regulation data.

[0100] Specifically, the structural image data can be structural image data collected in a previous experiment with the consent of the subject, or an open-source structural image data set, etc. An exemplary preprocessing can be performed on the structural image data, such as denoising, image segmentation, and edge detection, etc., to extract corresponding key geometric features; a three-dimensional reconstruction algorithm such as volume rendering or surface rendering is used to construct a corresponding individualized head model based on the extracted structural image data.

[0101] In an embodiment, the target key physiological parameter is the structural image data of the individual to be regulated for brain function. The field feature determination model can output field feature parameters based on the input brain function target regulation data and the individualized head model corresponding to the structural image data of the individual.

[0102] Taking a non-invasive neuroregulation means for achieving brain function target regulation data, i.e., transcranial electrical stimulation, as an example, different transcranial direct current stimulation (tDCS) stimulation parameters can be determined for each individualized head model. The tDCS stimulation parameters can include bipolar tDCS stimulation parameters, high-resolution tDCS stimulation parameters, and multi-channel tDCS stimulation parameters under different optimization methods, etc. Each set of tDCS stimulation parameters can specifically include parameters such as the position of the positioning electrode on the individualized head model, the stimulation intensity, and the waveform.

[0103] Exemplarily, the individualized head model and the corresponding set of transcranial stimulation parameters can be input into a simulation calculation system to obtain sample field feature parameters in the individualized head model calculated by the simulation calculation system using a finite element analysis method; a plurality of individualized head models and the corresponding set of transcranial stimulation parameters can be repeatedly input into the simulation calculation system to obtain corresponding sample field feature parameters.

[0104] In this embodiment, the structural image data is obtained by an open-source structural image dataset, and therefore the training data acquisition model is based on the structural image data to construct an individualized head model simulation to obtain corresponding sample field characteristic parameter output sample brain function target regulation data, which can obtain a large amount of sample field characteristic parameters and sample brain function target regulation data for training the field characteristic determination model.

[0105] Based on any of the above embodiments, the training data acquisition model is trained based on the experimental sample field characteristic parameters and the experimental sample brain function target regulation data, and includes:

[0106] The experimental sample field characteristic parameters are input into the original training data acquisition model to obtain the predicted brain function target regulation data output by the original training data acquisition model.

[0107] A loss function is calculated based on the predicted brain function target regulation data and the experimental sample brain function target regulation data, and the parameters of the original training data acquisition model are adjusted according to the loss function.

[0108] The step of inputting the experimental sample field characteristic parameters into the original training data acquisition model is repeated until the original training data acquisition model converges, and the training data acquisition model is obtained.

[0109] Specifically, the experimental sample field characteristic parameters refer to sample field characteristic parameters obtained by simulating and calculating the transcranial stimulation parameters and the subject's key physiological parameters for achieving brain function target regulation data. The experiment can be a brain function effect regulation experiment under different transcranial stimulation parameters completed by a small number of healthy adults as subjects and divided into several groups.

[0110] The working principle and technical effect of determining the loss function and adjusting the parameters of the original training data acquisition model based on the loss function are basically the same as those of determining the loss function and adjusting the parameters of the original field characteristic determination model based on the loss function. The working principle and technical effect of judging the convergence of the original training data acquisition model are basically the same as those of judging the convergence of the original field characteristic determination model. In this embodiment, they will not be described again.

[0111] In this embodiment, the original training data acquisition model is based on the experimental sample field characteristic parameters and the experimental sample brain function target regulation data, which extracts the correlation between the field characteristic parameters and the brain function target regulation data from accurate experimental data. Therefore, the training data acquisition model trained based on the brain function target regulation data can provide training data for the field characteristic determination model suitable for actual application.

[0112] Based on any of the above embodiments, before inputting the experimental sample field feature parameter into the original training data acquisition model, the method further comprises:

[0113] Mask processing the sample field feature parameter to obtain a masked field feature parameter;

[0114] Inputting the masked field feature parameter into the original feature acquisition model to obtain a predicted field feature parameter output by the original feature acquisition model;

[0115] Determining a second loss function between the sample field feature parameter and the predicted field feature parameter, and adjusting the parameters of the original feature acquisition model by minimizing the second loss function;

[0116] Repeating the step of masking the sample field feature parameter to obtain a masked field feature parameter until the original feature acquisition model converges, and obtaining the original feature acquisition model;

[0117] Adding a linear layer to the original feature acquisition model to obtain the original training data acquisition model.

[0118] Specifically, the mask processing refers to masking, randomizing, or zeroing part of the sample field feature parameter to obtain a modified incomplete masked field feature parameter. The masked field feature parameter is used to train the original feature acquisition model to infer the complete field feature parameter from the incomplete field feature parameter.

[0119] The working principle and technical effect of determining the second loss function and adjusting the parameters of the original feature acquisition model based on the second loss function are basically the same as those of determining the loss function and adjusting the parameters of the original field feature determination model based on the loss function. The working principle and technical effect of judging the convergence of the original feature acquisition model are basically the same as those of judging the convergence of the original field feature determination model. In this embodiment, they will not be described again.

[0120] The linear layer refers to a basic layer for linear processing of the predicted field feature parameter output by the original feature acquisition model.

[0121] In this embodiment, by masking and training the sample field feature parameter to obtain the original feature acquisition model, the robustness of the original training data acquisition model can be enhanced when the field feature parameter is incomplete or has noise in application, and the robustness of the training data acquisition model trained based on the original training data acquisition model can be enhanced.

[0122] And, in the embodiment, the original feature acquisition model is obtained by masking and training the sample field characteristic parameters, the ability of the original feature acquisition model to understand the field characteristic parameter rules is increased, a basis is provided for subsequent inference of brain function target regulation data from the experimental sample field characteristic parameters, the prediction accuracy of the training data acquisition model trained based on the original feature acquisition model is improved, and the number of iteration steps required for convergence of the original training data acquisition model in the training process is reduced.

[0123] The sample field characteristic parameters are obtained by simulating calculation on structural image data obtained from an open-source structural image data set, and therefore the training data acquisition model in the embodiment is obtained by further fine-tuning a small amount of accurate experimental sample field characteristic parameters and experimental sample brain function target regulation data based on the original training data acquisition model obtained by masking and training a large amount of sample field characteristic parameters, so as to improve the accuracy of the predicted brain function target regulation data output by the training data acquisition model.

[0124] In order to specifically describe the function of the method for determining individual transcranial stimulation parameters based on target regulation effect provided in the embodiment, a specific example is provided below.

[0125] Determine structural image data, construct a corresponding individual head model based on the structural image data, determine a plurality of groups of transcranial stimulation parameters based on the individual head model, each group of transcranial stimulation parameters is used to realize one kind of transcranial stimulation, and simulate calculation is performed according to the transcranial stimulation parameters and the individual head model to obtain sample field characteristic parameters in the cerebral cortex;

[0126] Mask the sample field characteristic parameters to obtain masked field characteristic parameters, input the masked field characteristic parameters into the original feature acquisition model to obtain predicted field characteristic parameters output by the original feature acquisition model, determine a second loss function between the sample field characteristic parameters and the predicted field characteristic parameters, adjust the parameters of the original feature acquisition model by minimizing the second loss function, repeat the step of masking the sample field characteristic parameters to obtain the masked field characteristic parameters until the original feature acquisition model converges, and obtain the original feature acquisition model; add a linear layer to the original feature acquisition model to obtain an original training data acquisition model;

[0127] Input the experimental sample field characteristic parameters into the original training data acquisition model to obtain predicted brain function target regulation data output by the original training data acquisition model, calculate a loss function based on the predicted brain function target regulation data and the experimental sample brain function target regulation data, adjust the parameters of the original training data acquisition model according to the loss function, and repeat the step of inputting the experimental sample field characteristic parameters into the original training data acquisition model until the original training data acquisition model converges, and obtain a training data acquisition model;

[0128] input the sample brain function target regulation data and the sample field characteristic parameter into a pre-training matching model of the original field characteristic determination model to obtain an encoding vector of the sample field characteristic parameter output by the pre-training matching model; input the encoding vector into a forward module of the original field characteristic determination model to obtain Gaussian noise data output by the forward module; the Gaussian noise data is obtained by adding noise to the encoding vector and performing forward diffusion; input the Gaussian noise data into a reverse module of the original field characteristic determination model to obtain a predicted field characteristic parameter encoding vector recovered by the reverse module; the predicted field characteristic parameter encoding vector is obtained by performing an inverse process on the Gaussian noise data; determine a loss function of the predicted field characteristic parameter encoding vector and the encoding vector of the sample field characteristic parameter, and adjust parameters of the original field characteristic determination model by minimizing the loss function; repeat the step of inputting the sample brain function target regulation data and the sample field characteristic parameter into the pre-training matching model of the original field characteristic determination model until the original field characteristic determination model converges, and obtain a field characteristic determination model; wherein the sample brain function target regulation data and the corresponding subject key physiological parameter are control conditions of the inverse process; the field characteristic determination model comprises a decoding module, and the decoding module is configured to decode the predicted field characteristic parameter encoding vector output by the reverse module to obtain a predicted field characteristic parameter.

[0129] determine brain function target regulation data; input the brain function target regulation data and the corresponding target key physiological parameter into the field characteristic determination model to obtain field characteristic parameters output by the field characteristic determination model, and the field characteristic parameters are used to realize the brain function target regulation data; determine corresponding transcranial stimulation parameters based on the field characteristic parameters.

[0130] The device for determining individualized transcranial stimulation parameters based on target regulation effects provided by the application is described below, and the device for determining individualized transcranial stimulation parameters based on target regulation effects described below can be correspondingly referred to the method for determining individualized transcranial stimulation parameters based on target regulation effects described above.

[0131] Figure 1 is a structural schematic diagram of the device for determining individualized transcranial stimulation parameters based on target regulation effects provided by the application, as Figure 1 shown, the device comprises:

[0132] a target regulation data determination module 210 configured to determine brain function target regulation data based on the target regulation effect;

[0133] The field characteristic parameter determination module 220 is configured to input the brain function target regulation data and the corresponding target key physiological parameter into a field characteristic determination model to obtain a field characteristic parameter output by the field characteristic determination model, and the field characteristic parameter is used to realize the brain function target regulation data; the field characteristic determination model is trained based on sample brain function target regulation data, sample target key physiological parameters and sample field characteristic parameters;

[0134] The stimulation parameter determination module 230 is configured to determine corresponding transcranial stimulation parameters based on the field characteristic parameters.

[0135] According to any of the above embodiments, the device for determining individualized transcranial stimulation parameters based on target regulation effects further comprises a field characteristic determination model training module configured to:

[0136] The sample field characteristic parameters are input into a forward module of the original field characteristic determination model to obtain Gaussian noise data output by the forward module; the Gaussian noise data is obtained by gradually adding noise to the sample field characteristic parameters and performing forward diffusion;

[0137] The Gaussian noise data is input into a reverse module of the original field characteristic determination model to obtain predicted field characteristic parameters recovered by the reverse module; the predicted field characteristic parameters are obtained by performing an inverse process on the Gaussian noise data;

[0138] A loss function of the predicted field characteristic parameters and the sample field characteristic parameters is determined, and parameters of the original field characteristic determination model are adjusted by minimizing the loss function;

[0139] The step of inputting the sample field characteristic parameters into the forward module of the original field characteristic determination model is repeated until the original field characteristic determination model converges, and the field characteristic determination model is obtained;

[0140] The sample brain function target regulation data and the corresponding target key physiological parameters are control conditions of the inverse process.

[0141] According to any of the above embodiments, the device for determining individualized transcranial stimulation parameters based on target regulation effects further comprises a field characteristic determination model training module configured to:

[0142] The sample brain function target regulation data and the sample field characteristic parameters are input into a pre-training matching model of the original field characteristic determination model to obtain an encoding vector of the sample field characteristic parameters output by the pre-training matching model;

[0143] The encoding vector is input into a forward module of the original field characteristic determination model to obtain Gaussian noise data output by the forward module; the Gaussian noise data is obtained by gradually adding noise to the encoding vector and performing forward diffusion.

[0144] inputting the Gaussian noise data into a reverse module of the original field feature determination model to obtain a predicted field feature parameter code vector recovered by the reverse module; the predicted field feature parameter code vector is obtained by performing an inverse process on the Gaussian noise data;

[0145] determining a loss function of the predicted field feature parameter code vector and a code vector of the sample field feature parameter, and adjusting parameters of the original field feature determination model by minimizing the loss function;

[0146] repeating the step of inputting the sample brain function target regulation data and the sample field feature parameter into the pre-training matching model of the original field feature determination model until the original field feature determination model converges, to obtain the field feature determination model;

[0147] wherein the sample brain function target regulation data and the corresponding target key physiological parameter are control conditions of an inverse process; the field feature determination model comprises a decoding module, and the decoding module is configured to decode the predicted field feature parameter code vector output by the reverse module to obtain a predicted field feature parameter.

[0148] Based on any of the above embodiments, the device for determining individual transcranial stimulation parameters based on target regulation effects further comprises a training data acquisition module configured to:

[0149] determine structural image data, and construct a corresponding individual head model based on the structural image data;

[0150] determine a plurality of transcranial stimulation parameters based on the individual head model; each set of the transcranial stimulation parameters is used to implement one transcranial stimulation;

[0151] perform simulation calculation to obtain sample field feature parameters in the cerebral cortex according to the transcranial stimulation parameters and the individual head model;

[0152] input the sample field feature parameters into a training data acquisition model to obtain sample brain function target regulation data output by the training data acquisition model; the training data acquisition model is trained based on experimental sample field feature parameters and experimental sample brain function target regulation data.

[0153] Based on any of the above embodiments, the device for determining individual transcranial stimulation parameters based on target regulation effects further comprises a training data acquisition model training module configured to:

[0154] input the experimental sample field feature parameters into an original training data acquisition model to obtain predicted brain function target regulation data output by the original training data acquisition model;

[0155] calculate a loss function based on the predicted brain function target regulation data and the experimental sample brain function target regulation data, and adjust parameters of the original training data acquisition model according to the loss function;

[0156] repeat the step of inputting the experimental sample field characteristic parameter into the original training data acquisition model until the original training data acquisition model converges, to obtain the training data acquisition model.

[0157] Based on any of the above embodiments, the device for determining individual transcranial stimulation parameters based on target regulation effects further comprises an original training data acquisition model training module for:

[0158] mask processing the sample field characteristic parameter to obtain a masked field characteristic parameter;

[0159] input the masked field characteristic parameter into the original feature acquisition model to obtain predicted field characteristic parameters output by the original feature acquisition model;

[0160] determine a second loss function between the sample field characteristic parameter and the predicted field characteristic parameter, and adjust parameters of the original feature acquisition model by minimizing the second loss function;

[0161] repeat the step of mask processing the sample field characteristic parameter to obtain a masked field characteristic parameter until the original feature acquisition model converges, to obtain the original feature acquisition model;

[0162] add a linear layer to the original feature acquisition model to obtain the original training data acquisition model.

[0163] Figure 3 An example of a structural diagram of an electronic device is shown in Figure 3 As shown, the electronic device can include a processor 310, a communications interface 320, a memory 330, and a communications bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other through the communications bus 340. The processor 310 can invoke logical instructions in the memory 330 to execute a method for determining individual transcranial stimulation parameters based on target regulation effects, which includes: determining brain function target regulation data; inputting the brain function target regulation data and corresponding target key physiological parameters into a field characteristic determination model to obtain field characteristic parameters output by the field characteristic determination model, the field characteristic parameters being used to achieve the brain function target regulation data; the field characteristic determination model being trained based on sample brain function target regulation data, sample target key physiological parameters, and sample field characteristic parameters; and determining corresponding transcranial stimulation parameters based on the field characteristic parameters.

[0164] Moreover, the logic instructions in the memory 330 described above can be implemented in the form of software functional units and sold or used as independent products, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0165] In another aspect, the present application also provides a computer program product, which comprises a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program can be executed by a processor to enable a computer to execute the method for determining individual transcranial stimulation parameters based on target regulation effect provided by the above-mentioned methods, the method comprising: determining brain function target regulation data; inputting the brain function target regulation data and corresponding target key physiological parameters into a field characteristic determination model to obtain field characteristic parameters output by the field characteristic determination model, the field characteristic parameters being used to realize the brain function target regulation data; the field characteristic determination model being trained based on sample brain function target regulation data, sample target key physiological parameters and sample field characteristic parameters; and determining corresponding transcranial stimulation parameters based on the field characteristic parameters.

[0166] In another aspect, the present application also provides a non-transitory computer readable storage medium, which stores a computer program, the computer program being executed by a processor to implement the method for determining individual transcranial stimulation parameters based on target regulation effect provided by the above-mentioned methods, the method comprising: determining brain function target regulation data; inputting the brain function target regulation data and corresponding target key physiological parameters into a field characteristic determination model to obtain field characteristic parameters output by the field characteristic determination model, the field characteristic parameters being used to realize the brain function target regulation data; the field characteristic determination model being trained based on sample brain function target regulation data, sample target key physiological parameters and sample field characteristic parameters; and determining corresponding transcranial stimulation parameters based on the field characteristic parameters.

[0167] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0168] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0169] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for determining individualized transcranial stimulation parameters based on the effect of target modulation, characterized in that, The method comprises the following steps: determining brain function target regulation data; the brain function target regulation data is the target regulation effect of the expected change or expected enhancement of brain function by non-invasive neural regulation means; inputting the brain function target regulation data and the corresponding target key physiological parameter into a field characteristic determination model to obtain a field characteristic parameter output by the field characteristic determination model; the field characteristic determination model is trained based on sample brain function target regulation data, sample subject key physiological parameters and sample field characteristic parameters; determining the corresponding transcranial stimulation parameter based on the field characteristic parameter; the target key physiological parameter refers to the key physiological parameter of an individual who needs to be regulated in brain function, and the key physiological parameter refers to a preset physiological characteristic affecting the brain function target regulation data; the sample brain function target regulation data is the expected change or expected enhancement effect of the brain function of a subject participating in an experiment, the sample subject key physiological parameter refers to a preset physiological characteristic affecting the brain function target regulation data of a subject participating in an experiment, and the sample field characteristic parameter refers to a field characteristic parameter for determining the sample brain function target regulation data; the field characteristic determination model is trained based on sample brain function target regulation data, sample subject key physiological parameters and sample field characteristic parameters, and comprises: inputting the sample field characteristic parameter into a forward module of an original field characteristic determination model to obtain Gaussian noise data output by the forward module; the Gaussian noise data is obtained by gradually adding noise to the sample field characteristic parameter and performing forward diffusion; inputting the Gaussian noise data into a reverse module of the original field characteristic determination model to obtain predicted field characteristic parameters recovered by the reverse module; the predicted field characteristic parameters are obtained by performing an inverse process on the Gaussian noise data; determining a loss function of the predicted field characteristic parameters and the sample field characteristic parameters, and adjusting parameters of the original field characteristic determination model by minimizing the loss function; repeating the step of inputting the sample field characteristic parameter into the forward module of the original field characteristic determination model until the original field characteristic determination model converges, to obtain the field characteristic determination model; wherein the sample brain function target regulation data and the corresponding subject key physiological parameter are control conditions of the inverse process.

2. The method of determining individualized transcranial stimulation parameters based on target modulation effects according to claim 1, wherein, the field characteristic determination model is trained based on sample brain function target regulation data, sample subject key physiological parameters and sample field characteristic parameters, and comprises: inputting the sample brain function target regulation data and the sample field characteristic parameter into a pre-training matching model of an original field characteristic determination model to obtain an encoding vector of the sample field characteristic parameter output by the pre-training matching model; inputting the encoding vector into a forward module of the original field characteristic determination model to obtain Gaussian noise data output by the forward module; the Gaussian noise data is obtained by gradually adding noise to the encoding vector and performing forward diffusion; The Gaussian noise data is input into a reverse module of the original field feature determination model, to obtain a predicted field feature parameter code vector recovered by the reverse module; the predicted field feature parameter code vector is obtained by performing an inverse process on the Gaussian noise data; A loss function of the predicted field feature parameter code vector and a code vector of the sample field feature parameter is determined, and parameters of the original field feature determination model are adjusted by minimizing the loss function; The steps of inputting the sample brain function target regulation data and the sample field feature parameter into the pre-training matching model of the original field feature determination model are repeated until the original field feature determination model converges, to obtain the field feature determination model; The sample brain function target regulation data and the corresponding subject key physiological parameter are control conditions of an inverse process; the field feature determination model comprises a decoding module, and the decoding module is configured to decode a predicted field intensity distribution code vector output by the reverse module to obtain a predicted field intensity distribution.

3. A device for determining individualized transcranial stimulation parameters based on the effect of target modulation, characterized in that, Comprise: A target regulation data determination module configured to determine brain function target regulation data; The brain function target regulation data is a target regulation effect of an expected change or an expected enhancement of brain function generated by a non-invasive neural regulation means; A field feature parameter determination module configured to input the brain function target regulation data and a corresponding target key physiological parameter into a field feature determination model, to obtain a field feature parameter output by the field feature determination model; the field feature determination model is trained based on sample brain function target regulation data, sample subject key physiological parameters, and sample field feature parameters; A stimulation parameter determination module configured to determine a transcranial stimulation parameter based on the field feature parameter; The target key physiological parameter refers to a key physiological parameter of an individual who needs to be regulated in brain function, and the key physiological parameter refers to a preset physiological characteristic that affects the brain function target regulation data; The sample brain function target regulation data is an expected change or an expected enhancement effect of brain function of a subject participating in an experiment, the sample subject key physiological parameter refers to a preset physiological characteristic that affects the brain function target regulation data of the subject participating in the experiment, and the sample field feature parameter refers to a field feature parameter determined to generate the sample brain function target regulation data; Further comprising a field feature determination model training module configured to: Input the sample field feature parameter into a forward module of an original field feature determination model, to obtain Gaussian noise data output by the forward module; the Gaussian noise data is obtained by performing forward diffusion on the sample field feature parameter to which noise is gradually added; Input the Gaussian noise data into a reverse module of the original field feature determination model, to obtain a predicted field feature parameter recovered by the reverse module; the predicted field feature parameter is obtained by performing an inverse process on the Gaussian noise data; Determine a loss function of the predicted field feature parameter and the sample field feature parameter, and adjust parameters of the original field feature determination model by minimizing the loss function; The forward module step of inputting the sample field feature parameter into the original field feature determination model is repeated until the original field feature determination model converges, and the field feature determination model is obtained. The sample brain function target regulation data and the corresponding target key physiological parameter are control conditions of the reverse process.

4. An electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, The processor executes the computer program to implement the method for determining individual transcranial stimulation parameters based on target regulation effects according to any one of claims 1 to 2; and the electronic device comprises a transcranial stimulation device.

5. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the method for determining individual transcranial stimulation parameters based on target regulation effects according to any one of claims 1 to 2.

6. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the method for determining individual transcranial stimulation parameters based on target regulation effects according to any one of claims 1 to 2. The computer program is executed by the processor to implement the method for determining individual transcranial stimulation parameters based on target regulation effects according to any one of claims 1 to 2.

Images