A training method for an electroencephalogram (EEG) source tracing model, an EEG source tracing method and an electronic device

By constructing the EEG dataset and using supervised learning methods to train the network model, combining Gaussky functions and fMRI data, the problem of low accuracy of EEG traceability technology is solved, and more accurate solution to EEG inverse problem and personalized brain source is achieved. position.

CN113240106BActive Publication Date: 2024-11-22SHENZHEN QUANYU INTELLIGENT TECHNOLOGY CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202110488900.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-04-30
Publication Date
2024-11-22
Estimated Expiration
2041-04-30

AI Technical Summary

Technical Problem

The existing EEG traceability technology is not accurate and it is difficult to accurately answer the EEG inverse problem. It is mainly due to irregular brain shape and complex structure. Traditional methods use simplified prior information to cause unfairness of the solution.

Method used

By constructing an EEG dataset, including neural activity source information and header electrode information, the network model is trained using a supervised learning method, and combining Gaussky functions and fMRI brain source data, a more accurate lead matrix and EEG positive problem construction is generated. model formula, thereby improving the accuracy of the EEG traceability model.

Benefits of technology

It improves the accuracy of EEG tracing and solves the EEG inverse problem, achieves more accurate neural activity source positioning, is suitable for a variety of EEG samples, and has the ability to personalize brain source positioning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113240106B_ABST
    Figure CN113240106B_ABST
Patent Text Reader

Abstract

The present application provides a training method for an EEG traceability model, an EEG traceability method and an electronic device. The training method for the EEG traceability model includes: constructing an EEG data set; wherein the EEG data set includes neural activity source information and head surface electrode information, and the head surface electrode information is generated based on the pre-acquired lead matrix, neural activity source information and EEG positive problem modeling formula; based on the EEG data set, supervised training is performed on the constructed network model until convergence to obtain an EEG traceability model. In this way, the head surface electrode information corresponding to the neural activity source information can be obtained, and then the network model can be trained using a supervised learning method. Compared with the traditional numerical method, the EEG traceability model trained by the method of the present application can improve the accuracy of EEG traceability, thereby improving the effect of solving the EEG inverse problem.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of biomedical technology, and in particular to a training method for an EEG tracing model, an EEG tracing method, and an electronic device. Background Art

[0002] Electroencephalogram (EEG) technology, as a non-invasive method for collecting neural signals, can record the surface potential signals of the human head. It has the characteristics of low cost, strong portability, and high temporal resolution. It has high value in basic research in cognitive neuroscience and clinical applications of neurological diseases such as epilepsy and depression.

[0003] Brain source localization is a technology that estimates brain source activity through head surface electrode potentials, which requires solving the following two problems.

[0004] The first is the EEG forward problem, which describes how electrical signals from a brain source are transmitted to scalp electrodes. Solving this forward problem requires first constructing a head model. This involves building a conductivity model based on the structure (solution space) and electrical conduction (forward operator) of the brain source within the human head. Numerical algorithms such as the boundary element method (BEM), the finite element method (FEM), or the finite difference method (FDM) are then used to obtain a solution to the EEG forward problem.

[0005] The second is the EEG inverse problem, which is to infer the location, intensity, and distribution of neural activity sources in the brain through inversion analysis of EEG data. However, since the number of neural sources is far greater than the number of electrodes, the solution to the EEG inverse problem is ill-posed. In order to ensure the computability of the ill-posed problem, traditional numerical methods for EEG inverse problems (such as MNE or LORETA) often use some relatively simplified settings when adding priors. Due to the irregular shape and non-uniform structure of the brain, the prior distribution of actual brain sources is very complex and cannot be represented by simple priors. Therefore, the current EEG tracing technology is not very accurate and has not achieved very good results in solving the EEG inverse problem. Summary of the Invention

[0006] The purpose of the embodiments of the present application is to provide a training method for an EEG tracing model, an EEG tracing method and an electronic device to improve the accuracy of EEG tracing and improve the effect of solving EEG inverse problems.

[0007] The present invention is achieved in that:

[0008] In the first aspect, an embodiment of the present application provides a training method for an EEG traceability model, comprising: constructing an EEG dataset; wherein the EEG dataset includes neural activity source information and head electrode information; the head electrode information is generated based on a pre-acquired lead matrix, the neural activity source information and an EEG positive problem modeling formula; and based on the EEG dataset, supervised training is performed on the constructed network model until convergence to obtain an EEG traceability model.

[0009] In an embodiment of the present application, the constructed EEG dataset includes neural activity source information and head electrode information; the head electrode information is generated using a pre-acquired lead matrix, neural activity source information, and the EEG forward problem modeling formula. This method can obtain the head electrode information corresponding to the neural activity source information, and then use supervised learning methods to train the network model. Compared to traditional numerical methods, the EEG source tracing model trained using the method of the present application can improve the accuracy of EEG source tracing, thereby improving the effectiveness of solving the EEG inverse problem.

[0010] In combination with the technical solution provided in the first aspect above, in some possible implementation methods, constructing the EEG data set includes: simulating different neural activity sources through Gaussian basis functions to obtain the neural activity source information; obtaining the head surface electrode information based on the neural activity source information, the pre-acquired lead matrix and the EEG positive problem modeling formula.

[0011] In an embodiment of the present application, when constructing an EEG data set, different neural activity sources are simulated through Gaussian basis functions, thereby simulating more types of training samples, which helps to improve the application effect of subsequent models and makes the trained EEG tracing model applicable to more EEG samples.

[0012] In combination with the technical solution provided in the first aspect above, in some possible implementations, the expression of the Gaussian basis function is: Among them, μ n (x) represents the source activity x n The Gaussian basis function centered at ω; n represents the activation value sampled from the Gaussian distribution; n = 1, 2, ..., k; σ s is the spatial standard deviation; the expression of the neural activity source information is: j = MΩ; where j represents the neural activity source information; M represents the weight of the Gaussian basis function, Ω = [μ1, μ1,…, μ k ];μ1,μ1,…,μ k represents Gaussian basis functions that simulate different sources of neural activity.

[0013] In combination with the technical solution provided in the first aspect above, in some possible implementation methods, constructing the EEG dataset includes: constructing a brain source unit prior distribution based on fMRI open data; sampling and generating the neural activity source information from the brain source unit prior distribution; and obtaining the head surface electrode information based on the neural activity source information, the pre-acquired lead matrix, and the EEG positive problem modeling formula.

[0014] Because fMRI (functional magnetic resonance imaging) brain data includes activation patterns in various brain regions related to emotion, perception, movement, decision-making, and attention, it can provide realistic prior information at various levels, thereby improving the reliability and accuracy of subsequent EEG traceability models.

[0015] In combination with the technical solution provided in the first aspect above, in some possible implementations, the lead matrix is ​​obtained by processing a twelve-layer head model, and the twelve-layer head model includes: skin layer, bone layer, cartilage layer, cerebrospinal fluid layer, cerebral gray matter layer, cerebellar gray matter layer, cerebral white matter layer, cerebellar white matter layer, brainstem layer, eyeball layer, muscle layer and fat layer.

[0016] In an embodiment of the present application, a twelve-layer head model is used in the process of calculating the lead matrix. Compared with the traditional three-layer head model and the five-layer head model, the established head model has higher accuracy, thereby making the calculated lead matrix more accurate.

[0017] In combination with the technical solution provided in the first aspect above, in some possible implementation methods, after obtaining the EEG tracing model, the method also includes: obtaining the collected head electrode information of the subject; performing unsupervised training on the EEG tracing model based on the head electrode information of the subject until convergence, to obtain an EEG tracing model suitable for the subject.

[0018] In the embodiment of the present application, after obtaining the EEG source tracing model, the collected head electrode information of the subject is obtained; then, based on the subject's head electrode information, the EEG source tracing model is unsupervisedly trained until convergence, thereby obtaining an EEG source tracing model suitable for the subject. In other words, after obtaining the EEG source tracing model, the model can also be fine-tuned based on the individual EEG, thereby achieving the purpose of personalized brain source localization. Moreover, due to the use of unsupervised learning, the model has the ability to generalize to real data.

[0019] In a second aspect, an embodiment of the present application provides an EEG tracing method, comprising: obtaining head electrode information of a target object; inputting the head electrode information of the target object into an EEG tracing model obtained by the training method of the EEG tracing model provided in the embodiment of the first aspect, and generating neural activity source information of the brain of the target object.

[0020] In the third aspect, an embodiment of the present application provides a training device for an EEG tracing model, comprising: a construction module for constructing an EEG dataset; wherein the EEG dataset includes neural activity source information and head electrode information; the head electrode information is generated based on a pre-acquired lead matrix, the neural activity source information and the EEG positive problem modeling formula; a supervised training module for performing supervised training on the constructed network model based on the EEG dataset until convergence to obtain an EEG tracing model.

[0021] In a fourth aspect, an embodiment of the present application provides an EEG tracing device, comprising: an acquisition module for acquiring the head surface electrode information of a target object; a processing module for inputting the head surface electrode information of the target object into an EEG tracing model obtained by the training method of the EEG tracing model provided in the embodiment of the first aspect, to generate neural activity source information of the brain of the target object.

[0022] In a fifth aspect, an embodiment of the present application provides an electronic device, comprising: a processor and a memory, wherein the processor and the memory are connected; the memory is used to store programs; the processor is used to call the programs stored in the memory and execute the method provided in the above-mentioned first aspect embodiment and / or second aspect embodiment.

[0023] In a sixth aspect, an embodiment of the present application provides a storage medium on which a computer program is stored. When the computer program is run by a processor, the method provided in the above-mentioned first aspect embodiment and / or second aspect embodiment is executed. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0026] Figure 2 A flowchart of the steps of a method for training an EEG tracing model provided in an embodiment of the present application.

[0027] Figure 3 Schematic diagram of a simulated neural activity source provided in an embodiment of the present application.

[0028] Figure 4 A schematic diagram of constructing an EEG dataset provided in an embodiment of the present application.

[0029] Figure 5 A schematic diagram of supervised training of a network model provided in an embodiment of the present application.

[0030] Figure 6 A schematic diagram of unsupervised training of an EEG traceability model provided in an embodiment of the present application.

[0031] Figure 7 A flowchart of the steps of an EEG tracing method provided in an embodiment of the present application.

[0032] Figure 8 A module block diagram of a training device for an EEG tracing model provided in an embodiment of the present application.

[0033] Figure 9 This is a module block diagram of an EEG tracing device provided in an embodiment of the present application.

[0034] Icon: 100-electronic device; 110-processor; 120-memory; 200-training device for EEG tracing model; 201-construction module; 202-supervised training module; 203-unsupervised training module; 300-EEG tracing device; 301-acquisition module; 302-processing module. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0036] Given that current EEG tracing technology has low precision, it has not achieved very good results in solving the EEG inverse problem. After research and exploration, the inventors of this application propose the following embodiments to solve the above problems.

[0037] See also Figure 1, a schematic structural block diagram of an electronic device 100 that uses a training method for an EEG traceability model and / or an EEG traceability method provided in an embodiment of the present application. In an embodiment of the present application, the electronic device 100 may be, but is not limited to, a personal computer (PC), a smart phone, a tablet computer, a personal digital assistant (PDA), a mobile Internet device (MID), etc. Structurally, the electronic device 100 may include a processor 110 and a memory 120.

[0038] The processor 110 is electrically connected to the memory 120 directly or indirectly to realize data transmission or interaction. For example, these elements can be electrically connected to each other through one or more communication buses or signal lines. The training device of the EEG traceability model includes at least one software module that can be stored in the memory 120 in the form of software or firmware (Firmware) or solidified in the operating system (Operating System, OS) of the electronic device 100. The EEG traceability device includes at least one software module that can be stored in the memory 120 in the form of software or firmware (Firmware) or solidified in the operating system (Operating System, OS) of the electronic device 100. The processor 110 is used to execute the executable modules stored in the memory 120, for example, the software function modules and computer programs included in the training device of the EEG traceability model, so as to realize the training method of the EEG traceability model. The processor 110 can execute the computer program after receiving the execution instruction.

[0039] The processor 110 may be an integrated circuit chip having signal processing capabilities. The processor 110 may also be a general-purpose processor, such as a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a discrete gate or transistor logic device, or a discrete hardware component, and may implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present application. In addition, the general-purpose processor may be a microprocessor or any conventional processor.

[0040] The memory 120 may be, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), and an electrically erasable programmable read-only memory (EEPROM). The memory 120 is used to store programs, and the processor 110 executes the programs after receiving execution instructions.

[0041] It should be understood that Figure 1 The structure shown is for illustration only. The electronic device 100 provided in the embodiment of the present application may also have Figure 1 Fewer or more components, or with Figure 1 Different configurations are shown. In addition, Figure 1 The components shown may be implemented by software, hardware or a combination thereof.

[0042] See also Figure 2 , Figure 2 A flowchart of the steps of the training method of the EEG source tracing model provided in the embodiment of the present application, which is applied to Figure 1 It should be noted that the training method of the EEG traceability model provided in the embodiment of the present application is not based on Figure 2 The order shown below is limited, and the method includes: step S101-step S102.

[0043] Step S101: constructing an EEG dataset; wherein the EEG dataset includes neural activity source information and head electrode information; the head electrode information is generated based on the pre-acquired lead matrix, neural activity source information and EEG positive problem modeling formula.

[0044] It should be noted that the EEG positive problem modeling formula is:

[0045] Φ=Kj+n (1)

[0046] In formula (1), Φ represents the head electrode information, which is an m×t matrix containing the measurement values ​​of m electrodes; K represents the lead matrix; j represents the neural activity source information, which is an a×t matrix representing the amplitude of a unknown source activity, t is the time length, and n represents noise.

[0047] That is, after obtaining the lead matrix and the neural activity source information, the corresponding head electrode information can be generated based on formula (1).

[0048] When constructing an EEG dataset, corresponding training data can be constructed for the activation of various emotions, perceptions, movements, decisions, attention and other processes in different areas of the human brain.

[0049] Step S102: Perform supervised training on the constructed network model based on the EEG dataset until convergence to obtain an EEG traceability model.

[0050] After constructing the EEG dataset, the dataset is fed into the constructed network model and trained until convergence. This yields the EEG source tracing model. It should be noted that the model's input data is head electrode information, and its output is neural activity source information. The supervised training process involves comparing the actual neural activity source information with the model's predictions, and then reversely optimizing the model parameters until an optimal model is obtained.

[0051] In summary, in the embodiments of the present application, the constructed EEG dataset includes neural activity source information and head electrode information; the head electrode information is generated using the lead matrix, neural activity source information, and the EEG forward problem modeling formula. In this way, the head electrode information corresponding to the neural activity source information can be obtained, and then the network model can be trained using supervised learning methods. Compared with traditional numerical methods, the EEG source tracing model trained using the method of the present application can improve the accuracy of EEG source tracing, thereby improving the effect of solving the EEG inverse problem.

[0052] In addition, it should be noted that existing algorithms for EEG tracing often require iterative solutions to EEG data, while the EEG tracing model constructed in this application can achieve immediate output after input, without the need for iteration, and is faster and has good real-time signal processing capabilities.

[0053] The above steps are described below with reference to specific examples.

[0054] In one embodiment, the EEG dataset may be a simulated dataset. Specifically, the steps of constructing the EEG dataset include: simulating different neural activity sources using Gaussian basis functions to obtain neural activity source information; and obtaining head electrode information based on the neural activity source information, a pre-acquired lead matrix, and an EEG positive problem modeling formula.

[0055] For easier understanding, see Figure 3 , Figure 3 Gaussian simulations of different sources of neural activity in the brain are shown.

[0056] The specific expression of the Gaussian basis function is:

[0057]

[0058] In formula (2), μ n (x) represents the source activity x n The Gaussian basis function centered at ω; n represents the activation value sampled from the Gaussian distribution; n = 1, 2, ..., k; σ s is the spatial standard deviation.

[0059] Correspondingly, the expression of the neural activity source information obtained by simulating different neural activity sources based on the above-mentioned Gaussian basis function is:

[0060] j=MΩ (3)

[0061] In formula (3), j represents the neural activity source information; M represents the weight of the Gaussian basis function, Ω = [μ1, μ1, …, μ k ];μ1,μ1,…,μ k represents Gaussian basis functions that simulate different sources of neural activity.

[0062] After obtaining the neural activity source information j using the above formula, multiply it by the lead matrix obtained from the head model, then project it onto the scalp and add noise to obtain the corresponding head-surface electrode information. In other words, after obtaining the neural activity source information j, the corresponding head-surface electrode information can be obtained by combining it with the lead matrix using the aforementioned EEG forward problem modeling formula. This method yields an EEG dataset with a one-to-one correspondence between the head-surface electrode information and the neural activity source information.

[0063] It should be noted that because real data often has a lot of noise, in order to adapt to the real situation, noise also needs to be added during simulation. The specific formula for generating head table electrode information is:

[0064] Φ=K(j+n j )+n φ (4)

[0065] In formula (4), Φ represents the head electrode information; K represents the lead matrix; j represents the neural activity source information; n j represents Gaussian white noise and pink noise; n φ Indicates the noise on the head surface electrodes.

[0066] For easier understanding, see Figure 4 , Figure 4 The process of generating EEG dataset is shown. Figure 4 The upper left corner shows the noise generation. Figure 4The lower left corner shows the neural activity source information j obtained by simulating different neural activity sources using Gaussian basis functions. After obtaining these two sets of data, they are combined with the lead matrix to generate the corresponding head electrode information.

[0067] In summary, in the embodiments of the present application, when constructing an EEG dataset, different neural activity sources are simulated through Gaussian basis functions, thereby simulating more types of training samples, which helps to improve the application effect of subsequent models and makes the trained EEG tracing model applicable to more EEG samples.

[0068] Optionally, the lead matrix can be obtained by processing a twelve-layer head model. The twelve-layer head model includes: a skin layer, a hard bone layer, a cartilage layer, a cerebrospinal fluid layer, a cerebral gray matter layer, a cerebellar gray matter layer, a cerebral white matter layer, a cerebellar white matter layer, a brainstem layer, an eyeball layer, a muscle layer, and a fat layer. The processing process may include numerical algorithms such as the finite element method (FEM) or the finite difference method (FDM). Since the above processing process is existing technology, it will not be described in detail here.

[0069] Traditional head models generally have three or five layers. A three-layer head model includes the brain, skull, and skin layers, while a five-layer head model includes the gray matter, white matter, cerebrospinal fluid, skull, and skin layers. Compared to traditional three- and five-layer head models, the twelve-layer head model of this application offers higher precision, resulting in a more accurate calculated lead matrix.

[0070] In one embodiment, the EEG dataset can be generated based on sampling of fMRI brain source data. Specifically, constructing the EEG dataset includes: constructing a prior distribution of brain source units based on open fMRI data; sampling and generating neural activity source information from the prior distribution of brain source units; and obtaining head surface electrode information based on the neural activity source information, a pre-acquired lead matrix, and an EEG forward problem modeling formula.

[0071] It should be noted that the fMRI brain data comes from the open-source database Neurosynth. Because fMRI brain data includes activation patterns across various brain regions related to emotion, perception, movement, decision-making, and attention, utilizing fMRI brain data can provide realistic prior information at various levels, thereby improving the reliability and accuracy of subsequently trained EEG traceability models.

[0072] In addition, when sampling and generating data, because each piece of data has a label describing the corresponding nervous system function, and each label has at least 20 pieces of data, the same number of samples are extracted from each label to construct a brain activation map with a probability distribution of various functions of the human brain. Such a probability distribution provides close to true prior information on the distribution of sources at the group level.

[0073] The processing method after extracting the neural activity source information from the fMRI brain source data can refer to the processing method after generating the neural activity source information by constructing the Gauss basis function in the above embodiment, which will not be described in detail here. Figure 4 After the method of obtaining the neural activity source information j in the lower left corner is changed to generating the neural activity source information by sampling the fMRI brain source data, Figure 4 What is shown is the process of generating EEG datasets from fMRI brain source data.

[0074] After the EEG dataset is obtained through the above steps, supervised learning can be performed on the constructed network model through step S102.

[0075] The advantage of supervised training models is that they provide an end-to-end approach where the model can directly learn the distribution of source activities. Figure 5 , Figure 5 Schematic diagram of the supervised training model. Figure 5 The corresponding formula is as follows:

[0076] K:K(ieφ=K(j+n j )+n Φ ) (5)

[0077]

[0078] Among them, formula (5) is the calculation of the forward operator of the supervised training model, and formula (6) is the calculation of the inverse operator of the supervised training model. ANN in formula (6) represents a parameterized artificial neural network. The purpose of supervised training is to find the inverse operator. The explanation of other parameters in formula (5) and formula (6) can refer to the above formulas.

[0079] Correspondingly, the loss function of the supervised training model is defined as:

[0080]

[0081] in represents the loss value, j' represents the predicted value of the neural activity source information output by the model, and j true represents the true neural activity source information in the EEG dataset; S(j) is a regularization term, which represents the prior estimate of the source activity distribution.

[0082] Optionally, after obtaining the EEG source tracing model as described above, the model can be fine-tuned for individual EEG signals to achieve personalized brain source localization. The specific steps include: obtaining the collected head electrode information of the subject; performing unsupervised training on the EEG source tracing model based on the subject's head electrode information until convergence, thereby obtaining an EEG source tracing model suitable for the subject.

[0083] Among them, the advantage of unsupervised training models is that they are based on data-driven and model-driven methods, avoiding the problem of inconsistency between artificially generated data and real data; they have the same loss function as traditional numerical algorithms and can directly inherit and learn from the latest research results of these methods. For easier understanding, please refer to Figure 6 , Figure 6 Schematic diagram of the unsupervised training model. Figure 6 The corresponding formula is as follows:

[0084] K:K(ieφ=K(j+n j )+n Φ ) (8)

[0085]

[0086] Among them, formula (8) is the calculation of the forward operator of the unsupervised training model, and formula (9) is the calculation of the inverse operator of the unsupervised training model. ANN in formula (9) represents a parameterized artificial neural network. The purpose of unsupervised training is to find the inverse operator. The explanation of other parameters in formula (8) and formula (9) can refer to the above formulas.

[0087] Correspondingly, the loss function of the unsupervised training model is defined as:

[0088]

[0089] in, represents the loss value, j' represents the predicted value of the neural activity source information output by the model; S(j) is the regularization term, which represents the prior estimate of the source activity distribution. C is the covariance matrix of the noise on the channel.

[0090] In summary, the embodiment of the present application also provides another model training method. After obtaining the EEG traceability model, the collected head electrode information of the subject is obtained; then, based on the head electrode information of the subject, the EEG traceability model is unsupervisedly trained until convergence, and an EEG traceability model suitable for the subject is obtained. That is, after obtaining the EEG traceability model, the model can also be fine-tuned according to the individual EEG, thereby achieving the purpose of personalized brain source localization, and due to the use of unsupervised learning, the model has the ability to generalize in real data.

[0091] See also Figure 7 Based on the same inventive concept, an embodiment of the present application also provides a method for tracing the source of EEG, including: step S201-step S202.

[0092] Step S201: Acquire the head surface electrode information of the target object.

[0093] Step S202: Inputting the target subject's head surface electrode information into the EEG source tracing model to generate the target subject's brain neural activity source information.

[0094] It should be noted that the EEG traceability model in step S202 is the model obtained by the EEG traceability model training method provided in the above embodiment.

[0095] In addition, the EEG tracing method provided in the embodiments of the present application can also be used in the detection of lesions such as epilepsy.

[0096] See also Figure 8 Based on the same inventive concept, an embodiment of the present application also provides a training device 200 for an EEG traceability model, which includes: a construction module 201 and a supervised training module 202.

[0097] Construction module 201 is used to construct an EEG dataset; wherein the EEG dataset includes neural activity source information and head electrode information; the head electrode information is generated based on the pre-acquired lead matrix, the neural activity source information and the EEG positive problem modeling formula.

[0098] The supervised training module 202 is used to perform supervised training on the constructed network model based on the EEG dataset until convergence, so as to obtain an EEG traceability model.

[0099] Optionally, the construction module 201 is specifically used to simulate different neural activity sources through Gaussian basis functions to obtain the neural activity source information; based on the neural activity source information, the pre-acquired lead matrix and the EEG positive problem modeling formula, the head surface electrode information is obtained.

[0100] Optionally, the construction module 201 is specifically used to construct a brain source unit prior distribution based on fMRI open data; sample and generate the neural activity source information from the brain source unit prior distribution; and obtain the head surface electrode information based on the neural activity source information, the pre-acquired lead matrix and the EEG positive problem modeling formula.

[0101] Optionally, the device further includes an unsupervised training module 203. The unsupervised training module 203 is configured to obtain, after obtaining the EEG source tracing model, the collected head surface electrode information of the subject; perform unsupervised training on the EEG source tracing model based on the head surface electrode information of the subject until convergence, and obtain an EEG source tracing model suitable for the subject.

[0102] See also Figure 9 Based on the same inventive concept, the present application also provides an EEG traceability device 300 . The device includes: an acquisition module 301 and a processing module 302 .

[0103] The acquisition module 301 is used to acquire the head surface electrode information of the target object.

[0104] The processing module 302 is used to input the head surface electrode information of the target subject into the EEG source tracing model to generate the neural activity source information of the brain of the target subject.

[0105] It should be noted that, since those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0106] Based on the same inventive concept, an embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the method provided in the above embodiment is executed.

[0107] The storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, or a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0108] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interface, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0109] In addition, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0110] Furthermore, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0111] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any actual relationship or order between these entities or operations.

[0112] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A training method for an EEG source tracing model, characterized in that: include: Constructing an EEG data set; wherein the EEG data set includes neural activity source information and head surface electrode information; the head surface electrode information is generated based on the pre-acquired lead matrix, the neural activity source information and the EEG positive problem modeling formula; the EEG data set includes training data constructed from the activation conditions of various emotions, perceptions, movements, decisions and attention corresponding processes in different areas of the human brain; Based on the EEG data set, supervised training is performed on the constructed network model until convergence to obtain an EEG traceability model; After obtaining the EEG tracing model, the method further includes: obtaining the collected head electrode information of the subject; performing unsupervised training on the EEG tracing model based on the head electrode information of the subject until convergence, so as to obtain an EEG tracing model suitable for the subject. The construction of the EEG data set includes: constructing a brain source unit prior distribution based on fMRI open data; sampling and generating the neural activity source information from the brain source unit prior distribution; obtaining the head electrode information based on the neural activity source information, the lead matrix obtained in advance, and the EEG positive problem modeling formula; wherein each data of the brain source unit prior distribution has a label describing the corresponding nervous system function, and when sampling and generating the neural activity source information, the same number of samples are extracted from each label to construct a brain activation map with equal probability distribution for various functions of the human brain.

2. The method according to claim 1, characterized in that The lead matrix is ​​obtained by processing a twelve-layer head model, wherein the twelve-layer head model includes: a skin layer, a hard bone layer, a cartilage layer, a cerebrospinal fluid layer, a cerebral gray matter layer, a cerebellar gray matter layer, a cerebral white matter layer, a cerebellar white matter layer, a brain stem layer, an eyeball layer, a muscle layer and a fat layer.

3. A method for tracing the source of EEG, characterized in that: include: Acquire the head surface electrode information of the target object; The head surface electrode information of the target object is input into the EEG source tracing model obtained by the training method of the EEG source tracing model as described in claim 1 or 2 to generate the neural activity source information of the brain of the target object.

4. A training device for an EEG source tracing model, characterized in that: include: A construction module, used to construct an EEG data set; wherein the EEG data set includes neural activity source information and head surface electrode information; the head surface electrode information is generated based on the pre-acquired lead matrix, the neural activity source information and the EEG positive problem modeling formula; A supervised training module is used to perform supervised training on the constructed network model based on the EEG data set until convergence to obtain an EEG tracing model; An unsupervised training module is used to obtain the collected head surface electrode information of the subject after obtaining the EEG traceability model; perform unsupervised training on the EEG traceability model based on the head surface electrode information of the subject until convergence, so as to obtain an EEG traceability model suitable for the subject; The construction module is specifically used to construct a prior distribution of brain source units based on fMRI open data; sample and generate the neural activity source information from the brain source unit prior distribution; obtain the head surface electrode information based on the neural activity source information, the lead matrix acquired in advance, and the EEG positive problem modeling formula; wherein each data of the brain source unit prior distribution has a label describing the corresponding nervous system function, and when sampling to generate the neural activity source information, the same number of samples are extracted from each label to construct a brain activation map with equal probability distribution of various functions of the human brain.

5. A device for tracing the source of electroencephalogram, characterized in that: include: An acquisition module, used to acquire the head surface electrode information of the target object; A processing module is used to input the head surface electrode information of the target object into the EEG source tracing model obtained by the training method of the EEG source tracing model as described in claim 1 or 2, so as to generate the neural activity source information of the brain of the target object.

6. An electronic device, characterized in that: include: A processor and a memory, the processor and the memory being connected; The memory is used to store programs; The processor is configured to run a program stored in the memory, execute the method according to claim 1 or 2, and / or execute the method according to claim 3.

Citation Information

Cited By

  • Stroke patient neural electrophysiology modeling method and system based on image completion

    CN120527012A