CN115281692A

CN115281692A - Closed-loop self-adaptive transcranial electrical stimulation device and method

Info

Publication number: CN115281692A
Application number: CN202210945498.3A
Authority: CN
Inventors: 唐再汐; 赵建军
Original assignee: Beijing Huanao Technology Development Co ltd
Current assignee: Beijing Huanao Technology Development Co ltd
Priority date: 2022-08-08
Filing date: 2022-08-08
Publication date: 2022-11-04

Abstract

The invention provides a closed-loop self-adaptive transcranial electrical stimulation device and a method. The method comprises the steps of collecting electroencephalogram data and paradigm behavior data, calculating a tested state evaluation result by combining multi-modal data fusion, comparing the tested state evaluation result with a preset regulation target, and regulating by adopting a preset regulation model, so that the tested state evaluation result and the preset regulation target are free from deviation. The invention can realize a closed-loop type self-adaptive transcranial electrical stimulation regulation mode.

Description

Closed-loop self-adaptive transcranial electrical stimulation device and method

Technical Field

The invention relates to a brain state regulation and control method, in particular to a closed-loop self-adaptive transcranial electrical stimulation device and method integrating high-precision transcranial electrical stimulation (HD-tDCS), electroencephalogram (EEG) and behavioral paradigm.

Background

Transcranial electrical stimulation (tES) is an important technical form of noninvasive neuromodulation in which low-intensity electrical current in a specific pattern is applied directly to a specific brain region using external electrodes to modulate neural activity. The transcranial electrical stimulation technology comprises transcranial direct current stimulation, transcranial alternating current stimulation, transcranial pulse stimulation and the like for neural regulation. The transcranial electrical stimulation neurobrain regulation and control technology is suitable for two scenes, namely non-medical and medical scenes, wherein the non-medical scene comprises two main scenes of brain function enhancement of special people such as soldiers, armed police, firemen, astronauts and the like and brain science research performed by scientific researchers; the medical scene comprises the nerve regulation and control of patients with nervous and mental diseases such as cerebral apoplexy rehabilitation, senile dementia, depression, aphasia and the like. The principle of the two scenes is basically consistent, but parameters of a specific evaluation training model, a brain state evaluation method and neural regulation and control are different. The traditional transcranial electrical stimulation regulation and control equipment can only realize the transcranial electrical stimulation function, and the stimulation effect can only be observed and evaluated through human experience. However, with the continuous improvement of scientific research and clinical requirements, the digital accurate result evaluation of the change of the brain state by the transcranial electrical stimulation is receiving attention.

The brain state evaluation technology is required before the transcranial electrical stimulation is implemented, and can be generally realized through two modes of physiological signal detection and behavior paradigm detection. The brain state evaluation technology based on electroencephalogram (EEG) is a commonly used method for evaluating brain state by detecting physiological signals, and the brain state is generally characterized by the accuracy and reaction time of behavioral expression in behavioral pattern detection. Accurate digital evaluation of brain state evaluation is generally based on EEG algorithm and behavioral result algorithm evaluation established in a normal mode to perform accurate data characterization, but accurate evaluation based on fusion of EEG and behavioral results is still blank.

The traditional transcranial electrical stimulation can be used for evaluating the stimulation effect only by means of other electroencephalogram equipment and behavioral paradigm software, and then the transcranial electrical stimulation mode is adjusted in a targeted mode according to the evaluation result. The method is very complicated in operation, long in operation flow, prone to error and incapable of meeting development requirements of intelligent neural regulation. Furthermore, the traditional transcranial electrical stimulation technology has two important limitations, namely that the traditional sponge electrode can only realize large-area stimulation of the scalp, cannot accurately control a stimulation area, and has limitations on treatment of a plurality of diseases. Secondly, stimulation is provided in an open-loop mode, namely, the stimulation is adjusted according to a preprogrammed mode of actively setting stimulation parameters, so that the evaluation effect of clinical symptoms or diseases appearing after stimulation of a patient cannot be adjusted and fed back.

The traditional brain state evaluation technology is generally a mode of separating physiological signal evaluation and paradigm evaluation, and an evaluation algorithm combining the physiological signal evaluation and the paradigm evaluation is not provided. The traditional brain state evaluation technology cannot form an evaluation result combining evaluation and a paradigm, and cannot provide accurate data guidance for closed-loop regulation. The current closed-loop regulation mode is more an off-line mode, namely, the off-line brain state evaluation is carried out after the transcranial electrical stimulation regulation is finished, and then the transcranial electrical stimulation mode is further adjusted according to an evaluation result. Even in an online closed-loop regulation mode, the traditional method is more in a mode of separating the acquisition and stimulation devices, and the acquisition and stimulation modules are synchronized through an external system. The closed-loop regulation mode cannot form an integral system, and has poor synchronization effect, complex arrangement and difficult operation.

It can be seen that in the prior art, if the above functions are implemented, a transcranial electrical stimulation device, an electroencephalogram acquisition device, a task paradigm display device, a personnel response (reaction time and response accuracy) recording device and synchronizers of the above three devices are required, 4 sets of independent hardware and 1 set of post-statistical software are required in total, and then 1 professional makes a judgment according to a statistical result. And, because the past system is "one-way time stamp marking": namely transcranial electrical stimulation time, paradigm task time and personnel reaction time are all synchronized to a computer system to form a data packet for later analysis; but can not fully automatically close the loop to reversely control the stimulation mode and the task difficulty of reverse regulation according to the change of brain electricity and the change of personnel behavior reaction; the "stimulation mode, task difficulty" can only be adjusted manually by a professional.

Disclosure of Invention

The invention provides a closed-loop self-adaptive transcranial electrical stimulation device and a method, relates to a transcranial electrical stimulation nerve regulation and control method and a brain state evaluation method, solves the problem of brain state regulation and control of the integrated integration of transcranial electrical stimulation, electroencephalogram acquisition and paradigm operation, solves the problem of building of brain state regulation and control equipment, and realizes a high-precision transcranial electrical stimulation (HD-tDCS), EEG acquisition and behavior paradigm integrated hardware system and software system aiming at the defect that the stimulation effect of traditional transcranial electrical stimulation equipment cannot be accurately evaluated and automatically fed back and controlled. The technical scheme is as follows:

an electroencephalogram acquisition and transcranial electrical stimulation integrated nerve regulation and control device comprises a transcranial electrical stimulation unit and an electroencephalogram acquisition unit which are controlled by a control unit, wherein the transcranial electrical stimulation unit and the electroencephalogram acquisition unit share the same set of electrodes through a switching circuit, and the switching circuit is controlled by the control unit; the transcranial electrical stimulation unit comprises a positive and negative high-voltage circuit, the positive and negative high-voltage circuit is connected to the electrodes through a switching circuit, the electroencephalogram acquisition unit comprises a filter circuit and an ADC (analog-to-digital converter) acquisition circuit, the filter circuit is connected to the electrodes through the switching circuit, and the electroencephalogram acquisition unit comprises a filter circuit and an ADC acquisition circuit which are connected.

The electrode comprises an electrode cup cover, a sheet electrode and an electrode cup, wherein the electrode cup cover is buckled on the upper surface of the electrode cup, an annular supporting structure used for placing the sheet electrode is arranged at the bottom of the electrode cup, electrode cup supporting legs are arranged at the lower part of the electrode cup cover, and the upper ends of the sheet electrode are pressed by the electrode cup supporting legs.

The electroencephalogram acquisition unit comprises an electroencephalogram acquisition circuit connected with the electrodes, and the electroencephalogram acquisition circuit uses an AD7768 chip or an ADS1299 chip for 8-channel synchronous sampling.

The electroencephalogram acquisition circuit adopts Daisy Chain Daisy-Chain cascade connection, and 8 channels form 16 channels, 32 channels, 64 channels and 128 channels.

The electroencephalogram acquisition circuit leads out all the positive and negative electrodes, and acquires electroencephalogram signals by performing difference between the positive and negative electrodes; or the negative terminals are connected together internally, and electroencephalogram signals are collected through the difference between the positive electrode and the unified reference.

The transcranial electrical stimulation unit comprises a high-voltage generating circuit and a back-voltage circuit connected with the high-voltage generating circuit, and the high-voltage generating circuit adopts a tps61230 boosting chip.

The transcranial electric stimulation circuit is further provided with a constant current feedback detection circuit, the constant current feedback detection circuit sends stimulation current values of the positive and negative high-voltage circuits to the control unit, the constant current feedback detection circuit is provided with a second amplifier, and the second amplifier adopts SGM8622.

The electroencephalogram acquisition unit further comprises a lead falling judgment circuit, wherein the lead falling judgment circuit is positioned between the filter circuit and the switching circuit and is used for judging whether the electrodes fall off and adopting a hysteresis comparator circuit.

The control unit is provided with a processor, and the processor adopts an STM32F383CC chip.

A closed-loop adaptive transcranial electrical stimulation method comprises the following steps:

s1: selecting different test paradigms according to different tested objects;

s2: presetting a regulation target as the grade setting of the brain state evaluation after the regulation of the tested object;

s3: presetting a regulation and control model, and determining output parameters of transcranial electrical stimulation;

s4: acquiring and processing electroencephalogram data to obtain an evaluation result of the electroencephalogram state of a tested object;

s5: behavior data are collected and paradigm evaluation is carried out;

s6: generating a fusion brain state evaluation algorithm according to the electroencephalogram precision evaluation algorithm and the paradigm evaluation algorithm, and calculating the evaluation result of the tested state;

s7: calculating the deviation between the tested state evaluation result and a preset regulation and control target;

s8: calling a regulation and control model, and regulating and controlling transcranial electrical stimulation output parameters according to the deviation value;

s9: and repeating the steps S3-S8, and circularly evaluating, testing and regulating on line until the tested state evaluation result and the preset regulation target do not have deviation.

Further, in step S2, the control target is divided into ten stages, including five stages, i.e., a severe positive deviation stage, a substantially non-deviation stage, a negative deviation stage, and a severe negative deviation stage, where each stage is divided into two stages, i.e., a high stage and a low stage.

Further, in step S3, the setting step of the output parameter is as follows:

s11: determining a pattern of transcranial electrical stimulation;

the method comprises the following steps of (1) dividing into transcranial direct current stimulation, transcranial alternating current stimulation, transcranial pulse stimulation and transcranial random stimulation modes;

s2: determining a location of transcranial electrical stimulation;

the position is selected from an international standard electroencephalogram position distribution map, 1 to N electrical stimulation positions are arranged, wherein the positive pole is 1-N-1 positions, the negative pole is 1-N-1 positions, and a regulating and controlling circuit combination is formed together, wherein N natural numbers are more than or equal to 2;

s3: determining specific parameters of transcranial electrical stimulation;

the method comprises the steps of stimulating current polarity, stimulating current duration, stimulating current amplitude, stimulating current bias, stimulating current frequency and stimulating current duty ratio;

s4: determining an evaluation mode of a regulation target;

the evaluation mode of the regulation and control target is divided into an asynchronous mode and a synchronous mode, wherein the asynchronous mode is to evaluate whether the regulation and control target is achieved after the regulation and control execution is finished, and the synchronous mode is to evaluate whether the regulation and control target is achieved while the regulation and control are carried out;

s5: determining an operation mode when a regulation target is achieved or not achieved;

stopping the current regulation when the regulation target is achieved, or continuing to regulate according to a fixed time, thereby consolidating the target; when the regulation target is not reached, the regulation can be continuously carried out according to a set time, and the regulation is stopped no matter whether the regulation target is reached or not; or directly stopping the regulation when the regulation target is not reached.

Further, in step S3, the electroencephalogram state evaluation includes data preprocessing, relevant feature analysis and extraction, machine learning, and obtaining an evaluation result of the electroencephalogram state of the subject;

the data preprocessing aims to remove bad data or data with full artifacts without changing clean data, or to apply a filter or spatial transformation to change the clean data to facilitate analysis, and comprises the processing of re-referencing, filtering and independent principal component analysis;

the correlation characteristic analysis and extraction is used for extracting basic data of electroencephalogram analysis, and comprises Fourier transform, nonlinear transform, time domain analysis and brain network analysis;

the machine learning and the tested electroencephalogram state evaluation result are obtained, and classification tasks in supervised learning are realized, wherein the classification tasks comprise processing modes of a support vector machine, a neural network framework and a self-coding network.

Further, in step S5, the evaluation of the paradigm behavior state includes: and (4) counting the reaction time of the test, counting the reaction accuracy of the test, and calculating the reaction accuracy characteristics of the reaction time of the test.

Further, in step S6, the accuracy response time of the deviation state and the non-deviation state is used as a behavior index and introduced into the discriminant model:

the first method is to perform feature fusion according to data of different modes, and the fusion mode comprises data level fusion, decision level fusion and intermediate fusion;

the second method is that the accuracy reaction is not regarded as data of different modes, and only the data is taken as general characteristics to be normalized with the electroencephalogram characteristics;

the third method is to perform algorithm fusion of different data modes through the difference between the populations of the absolute control group.

In the first method:

the data level fusion aims at fusing a plurality of independent data into a single feature vector, and then inputting the single feature vector into a classifier, wherein the specific methods comprise splicing, multiplication according to bits and addition according to bits;

the fusion of the decision level aims at respectively training classifiers by using different modal data, and then scoring the output to perform fusion on the result, wherein the specific method comprises maximum value fusion, average value fusion, bayesian rule fusion and ensemble learning;

and the intermediate layer fusion is to convert different modal data into high-dimensional characteristic expression and then fuse the high-dimensional characteristic expression with the intermediate layer of the model.

Further, in step S8, the evaluation result already meets the regulation target, and regulation may not be started; or starting the same-level consolidation regulation mode according to a preset mode.

Further, in step S5, the paradigm evaluation includes, but is not limited to, a postner attention paradigm and an n-back working memory paradigm.

The closed-loop self-adaptive transcranial electrical stimulation device and method provided by the invention are used for automatically evaluating the brain state and automatically matching different accurate regulation and control modes, so that the closed-loop self-adaptive transcranial electrical stimulation regulation and control mode is realized.

A closed-loop self-adaptive transcranial electrical stimulation device and a method thereof comprise a high-precision transcranial electrical stimulation and electroencephalogram acquisition integrated technology, an acquisition and stimulation compatible electrode, a neural state analysis and evaluation model based on electroencephalogram and paradigm fusion, an accurate neural regulation and control model and the like. Wherein, the high-precision transcranial electrical stimulation can effectively lock the stimulation area, and realize accurate regulation and control. The electroencephalogram acquisition integrated technology solves the key technical problems of circuit multistage protection, multistage switching, multistage isolation and the like. The acquisition stimulation compatible electrode can be adapted to the current and impedance requirements of two different scenes of acquisition and stimulation, and the same electrode can acquire signals and can perform two different operation functions of stimulation. The neural state analysis and evaluation model establishes a reliable algorithm model through modes such as pattern recognition, machine learning and the like based on specific scene requirements, and provides data guidance for closed-loop regulation and control. The precise neural regulation model sets targeted differential neural regulation parameters based on different evaluation results.

A closed-loop self-adaptive transcranial electrical stimulation device and a method are an integrated neural regulation and control mode, three functions of acquisition, stimulation and evaluation are formed into closed-loop self-adaptive linkage, the potential and the advantages of a transcranial electrical stimulation technology are fully exerted, an intelligent self-adaptive treatment mode of the neural regulation and control mode can be dynamically adjusted according to a brain state evaluation result, visual diagnosis and treatment data guidance is provided for special personnel, scientific research personnel and clinicians, the feedback efficiency of evaluation analysis and control effects is greatly improved, the pertinence and the effectiveness of regulation and control are improved, and the method is a novel intelligent neural regulation and control method.

Drawings

FIG. 1 is a schematic frame diagram of the nerve regulation and control device integrating electroencephalogram acquisition and transcranial electrical stimulation;

FIG. 2 is a schematic structural diagram of the electroencephalogram electrode;

FIG. 3 is a schematic diagram of an ADC acquisition circuit of the electroencephalogram acquisition unit;

FIG. 4 is a schematic diagram of the high voltage generation circuit;

FIG. 5 is a schematic diagram of the back voltage circuit;

FIG. 6 is a schematic diagram of the constant current feedback detection circuit;

FIG. 7 is a schematic flow diagram of the closed-loop adaptive transcranial electrical stimulation method;

FIG. 8 is a schematic flow chart of the brain electrical acquisition;

FIG. 9 is a schematic flow diagram of the paradigm shift system;

fig. 10 is a schematic diagram of the Posner attention paradigm;

FIG. 11 is a schematic diagram of the 2-back paradigm.

Detailed Description

As shown in fig. 1, the closed-loop self-adaptive transcranial electrical stimulation device comprises a transcranial electrical stimulation unit and an electroencephalogram acquisition unit which are controlled by a control unit, wherein the transcranial electrical stimulation unit and the electroencephalogram acquisition unit share the same set of electrodes through a switching circuit, and the switching circuit is controlled by the control unit; the transcranial electrical stimulation unit comprises a positive and negative high-voltage circuit, the positive and negative high-voltage circuit is connected to the electrodes through a switching circuit, the electroencephalogram acquisition unit comprises a filter circuit and an ADC (analog-to-digital converter) acquisition circuit, the filter circuit is connected to the electrodes through the switching circuit, and the electroencephalogram acquisition unit comprises a filter circuit and an ADC acquisition circuit which are connected.

The electroencephalogram acquisition unit further comprises a lead falling judgment circuit, and the lead falling judgment circuit is located between the filter circuit and the switching circuit and used for judging whether the electrodes fall off or not. The lead falling judging circuit is the prior art and can be realized by adopting a hysteresis comparator circuit.

The control unit is provided with a processor, and the processor adopts an STM32F383CC chip and is connected with a computer to realize data transmission.

As shown in fig. 2, the electrode comprises an electrode cup cover 1, a sheet electrode 2 and an electrode cup 3, the electrode cup 3 is installed on an electrode cap 7, the electrode cup cover 1 is buckled on the upper surface of the electrode cup 3, an annular supporting structure 8 is arranged at the bottom of the electrode cup 3 and used for placing the sheet electrode 2, and the electrode cup cover 1 is provided with an electrode cup support leg 10 and used for being matched with the annular supporting structure 8 to fix the sheet electrode 2. The lower part of the sheet electrode 2 is used for smearing electroencephalogram paste 5, the top of the sheet electrode 2 is connected with an electrode wire 6, the sheet electrode 2 can extend downwards to the lower part of the electrode cup 3 in a vertical wire outlet mode, the distance between the sheet electrode 2 and the head 9 is reduced, and the electroencephalogram signal acquisition quality and the transcranial electrical stimulation effect are improved. In this embodiment, the plurality of electrodes are mounted on the electrode cap through the electrode holders, and the electrode cap is worn on the head of a user.

As shown in fig. 3, the ADC collecting circuit is connected to the control unit, and the ADC collecting circuit of the electroencephalogram collecting unit may use AD7768 or ADs1299 chips, or use an operational amplifier and a high-precision ADC to collect data of electrode signals.

In the embodiment, AD7768 and ADS1299 chips which are 8-channel synchronous sampling chips are used, and 16-channel, 32-channel, 64-channel and 128-channel electroencephalogram acquisition circuits can be formed by Daisy Chain Daisy-Chain cascade connection. The electroencephalogram acquisition unit can lead out the positive and negative electrodes of the electrode to perform difference between the positive and negative electrodes to acquire electroencephalogram signals; the negative terminals of the electrodes can also be connected together internally, and brain electrical signals are acquired through the difference between the positive electrode and a uniform reference (RFE ELEC).

As shown in fig. 4 and 5, the positive and negative high-voltage circuits include a high-voltage generation circuit and a back-voltage circuit connected to the high-voltage generation circuit, and the high-voltage generation circuit employs a tps61230 boost chip. The back-pressure circuit is prior art and is not described in detail. The positive and negative high-voltage circuits realize free switching of cathode stimulation and anode stimulation through positive and negative voltages.

The conventional booster circuit generates fixed high voltage by dividing voltage through resistors, and the high voltage generation circuit dynamically controls the specific value of the high voltage through a single IO pin to realize the output of 5-20V. The back voltage circuit is an existing device, and achieves the effect that the input voltage is several volts, and the output voltage is negative several volts.

The positions of the electrodes on the head of a person are different, scalp grease treatment of the person is different, electroencephalogram paste treatment conditions are different, and impedance between the positive electrode and the negative electrode is greatly different. Different impedances, different high voltages required by transcranial electrical stimulation units, such as: the impedance of 3K omega is between the positive electrode and the negative electrode, the current is 2mA, and the pressure difference between the positive electrode and the negative electrode is only 6V; if impedance of 10K omega is between the positive electrode and the negative electrode, and current of 2mA, the pressure difference between the positive electrode and the negative electrode needs 20V. The battery power supply is generally a 4.2-volt lithium battery, and the transcranial electrical stimulation unit can be ensured to have higher efficiency only by dynamically adjusting the output high-voltage range according to the impedance requirement.

As shown in fig. 6, the transcranial electrical stimulation circuit is further provided with a constant current feedback detection circuit, the constant current feedback detection circuit sends stimulation current values of the positive and negative high-voltage circuits to the control unit, the constant current feedback detection circuit is provided with a second amplifier, the second amplifier adopts SGM8622, and a high-precision resistor R66 is connected between a positive input end and a negative input end of the second amplifier. When the device is used, stimulation current transmitted by the positive and negative high-voltage circuits is subjected to differential sampling through the high-precision resistor R66 to be changed into a single-ended signal, and then the single-ended signal is input into the high-precision ADC; after ADC sampling, the processor controls the positive and negative high-voltage circuits to dynamically adjust within a certain range by comparing the deviation between the set current and the actual current.

The switching circuit is controlled by the control unit and is used for switching the working mode of the electrode, because the electrode can realize transcranial electrical stimulation and electroencephalogram acquisition, but can not realize transcranial electrical stimulation and electroencephalogram acquisition at the same time, the electrode can only be connected to the transcranial electrical stimulation unit or the electroencephalogram acquisition unit through the switching circuit at the same time, and the switching circuit adopts an electronic switching switch.

The working mode of the electrode for transcranial electrical stimulation or electroencephalogram acquisition is processed by the control unit, and when the control unit controls the positive and negative high-voltage circuits to send out stimulation currents, the working mode of the electrode is switched by the switching circuit and is in an electrical stimulation working mode; when the control unit controls the positive and negative high-voltage circuits to stop sending the stimulating current, the switching circuit switches the working mode of the electrodes, which is the working mode of electroencephalogram collection.

When the device is used, the working modes of the plurality of electrodes can adopt a synchronous control mode and an asynchronous control mode, wherein the synchronous control mode refers to transcranial electrical synchronous stimulation-electroencephalogram synchronous acquisition, the electrodes simultaneously perform transcranial electrical stimulation, or the electrodes simultaneously perform electroencephalogram acquisition, and can be made into 32-128 channels. The asynchronous control mode refers to transcranial electrical synchronous stimulation-electroencephalogram asynchronization, and the electrodes can be used for acquiring electroencephalogram or performing transcranial electrical stimulation and can be made into 32-128 channels when the electrodes are not used simultaneously. The asynchronous control mode is shown in fig. 8.

As shown in fig. 7, the closed-loop adaptive transcranial electrical stimulation method is implemented by a closed-loop adaptive transcranial electrical stimulation device, and includes the following steps:

s1: selecting different test paradigms according to different tested objects;

firstly, determining a tested object according to soldiers, armed policemen, firefighters, common young adults, old people, children and the like; different test paradigms are selected according to different tested objects, so that the pertinence and the adaptability are improved.

the evaluation of the brain state after regulation and control is divided into ten grades, the evaluation is divided into five stages of serious positive deviation, basically no deviation, negative deviation and serious negative deviation, and each stage is divided into two grades, namely high grade and low grade, so that ten grades of grade setting are formed.

In these ten levels, a level of regulation is determined as a regulation target, and in principle, the regulation target should be set step by step according to the result of the last evaluation of the brain state of the subject. In special cases, regulatory targets may also be set across levels. Likewise, only level consolidation may be done without crossing levels.

determining whether the regulation is positive regulation or negative regulation according to different brain state regulation targets, and determining specific parameters of the regulation according to different brain state regulation targets. The output parameter setting sequence is as follows:

(1) Determining a pattern of transcranial electrical stimulation;

the method is divided into several modes of transcranial direct current stimulation, transcranial alternating current stimulation, transcranial pulse stimulation, transcranial random stimulation and the like.

(2) Determining a location of transcranial electrical stimulation;

the location may be selected from international standard brain electrical location distribution maps, which may range from 1 to N electrical stimuli. Wherein the positive pole is at most N-1 positions, and the negative pole is at most N-1 positions, which jointly form a circuit combination for regulation and control.

(3) Determining specific parameters of transcranial electrical stimulation;

including the polarity of the stimulation current, the duration of the stimulation current, the amplitude of the stimulation current, the bias of the stimulation current, the frequency of the stimulation current, the duty cycle of the stimulation current, and other parameters. The stimulation current polarity comprises a positive polarity and a negative polarity; the stimulation time is from 1 minute to 6-0 minutes; the stimulation current amplitude is from 300vA to 2000vA; the amplitude of the stimulating bias current and the stimulating current does not exceed +/-2000 vA and can be set; stimulation frequencies range from direct current to 250Hz; the stimulation current has a duty cycle from 0% to 100% and an amplitude from DC to 250Hz.

(4) Determining an evaluation mode of a regulation target;

the evaluation mode of the regulation and control target is divided into an asynchronous mode and a synchronous mode. The asynchronous mode is to evaluate whether the regulation and control target is achieved after the regulation and control execution is finished, and the synchronous mode is to evaluate whether the regulation and control target is achieved while regulating.

(5) Determining an operation mode when a regulation target is achieved or not achieved;

when the regulation target is achieved, the current regulation can be stopped, or the regulation can be continued according to a fixed time, so that the target is consolidated.

When the regulation target is not reached, the regulation can be continuously carried out according to a set time, and the regulation is stopped no matter whether the regulation target is reached or not. Or directly stopping the regulation when the regulation target is not reached.

S4: acquiring and processing electroencephalogram data to obtain an evaluation result of the state of tested electroencephalogram;

the process of electroencephalogram (EEG) state evaluation includes: the method comprises the steps of data preprocessing, relevant feature analysis and extraction, machine learning and acquisition of a tested electroencephalogram state evaluation result.

S5: behavior data are collected and a paradigm is evaluated;

the paradigm evaluation algorithm includes: (1) Counting the reaction time of the test, (2) counting the accuracy of the reaction time of the test, and calculating the characteristics of the reaction time and the reaction accuracy of the test according to (1) and (2).

S6: forming a fusion calculation algorithm according to the electroencephalogram data and the behavior data so as to evaluate a tested state evaluation result;

because the behavior index and the electroencephalogram index can be used for judging the state of the tested person, the behavior index and the electroencephalogram index need to be combined for comprehensive evaluation. Specifically, the accuracy response time of the deviation state and the non-deviation state can be used as a behavior index to be introduced into the deviation discrimination model.

The first approach is to perform feature fusion on data of different modalities. The basic fusion modes include data-level fusion, decision-level fusion, and intermediate fusion.

Fusion of data levels: the method aims to synthesize electroencephalogram EEG state evaluation result data and behavior paradigm result data into a single feature vector, and then input the single feature vector into a classifier. The accuracy rate reaction is a one-dimensional feature, and bitwise multiplication and bitwise addition are not suitable, so splicing can be tried, but compared with high-dimensional electroencephalogram data, the one-dimensional feature is probably submerged in the high-dimensional data feature, and the complementary effect of the high-dimensional data feature cannot be shown.

Fusion of decision levels: the method aims to train classifiers by using electroencephalogram EEG state evaluation result data and behavior paradigm result data respectively, then score (make decision) output to perform result fusion, and a specific method comprises maximum value fusion, average value fusion, bayesian rule fusion and ensemble learning, wherein a numerical experiment is required to be performed when specifically adopting which method.

Intermediate layer fusion: the electroencephalogram EEG state evaluation result data and the behavioral paradigm result data are converted into high-dimensional feature expressions, then the high-dimensional feature expressions are fused with the middle layer of the model, the method is mainly used for feature fusion of a neural network, for pilot data, the mapping of electroencephalogram high-dimensional features to space symmetric manifolds can be considered, the mapping of accuracy reaction to one-dimensional manifolds can be considered, data fusion of the two manifolds can be carried out, and then the data can enter a classifier.

The second method is to not regard the accuracy reaction of the behavior paradigm as data of different modes, and only take the data as general characteristics to perform normalization processing with the electroencephalogram discrimination characteristics, so that the operation is simple, but the effect of the accuracy reaction can be weakened.

The third method is to perform algorithm fusion of different data modes through the difference between the populations of the absolute control group. Specifically, the method is to perform experimental evaluation on a target population and an absolute control population according to a completely consistent method. And then, calculating the difference grade between the evaluation result of the electroencephalogram nerve state between the two populations and the behavior evaluation results such as accuracy, reaction time and the like according to a percentage system. The larger the difference grade, the larger the weight, and the smaller the difference grade, the smaller the weight, thereby forming an objective fusion weight proportion.

S7: calculating the deviation between the tested state evaluation result and the regulation and control target;

and comprehensively evaluating the grade of the tested state according to the comprehensive evaluation algorithm in the steps, comparing the grade with the grade requirement of a preset regulation and control target, and determining the grade difference between the evaluation result of the tested state and the preset regulation and control target.

and (5) evaluating the grade difference between the evaluation result and the regulation target according to the tested state evaluation result in the step (S6), and calling a preset regulation and control model for regulation and control. Specifically, if the evaluation result already meets the regulation target, the regulation may not be started, or the same-level consolidated regulation mode may be started in a preset manner.

The regulation and control model adopts the output parameters of the transcranial electrical stimulation in the step S3. The output parameters of transcranial electrical stimulation include: stimulation location, stimulation time, output stimulation pattern (dc stimulation, ac stimulation, pulsed stimulation, random stimulation, deep stimulation), output stimulation frequency, output stimulation polarity, output stimulation intensity, and output start/stop.

S9: and repeating the steps S3-S6, and circularly evaluating, testing and regulating on line until the tested state evaluation result has no deviation from the regulation and control target.

The evaluation mode can be divided into a synchronous evaluation mode and an asynchronous evaluation mode.

Further, in step S4, as shown in fig. 8, the electroencephalogram (EEG) state evaluation includes data preprocessing, relevant feature analysis and extraction, machine learning, and obtaining an evaluation result of the electroencephalogram state of the subject. The details are as follows:

s41: data preprocessing: preprocessing aims to remove bad or artifact-filled data without altering the clean data, or to apply filters or spatial transforms to change the clean data to facilitate analysis, and generally the main methods of processing EEG data include:

(1) Re-referencing: EEG measures the potential of an electrode, but since the individual electrode potentials cannot be measured, it is necessary to co-operate an electrode of known electrode potential (preferably zero potential) as a reference electrode. The potential of the ideal reference electrode should be zero (its potential should be unchanged). The current common reference modes include bilateral mastoid reference, average reference and bipolar reference.

(2) Filtering; eliminating high-frequency artifacts and low-frequency drift; different bands are filtered out depending on the analysis task and data characteristics. The common frequency band in analysis is between 0.5Hz and 70Hz, and the band-stop filtering is needed to eliminate the commercial power interference.

(3) Independent principal component analysis: to clean up EEG data by identifying components that separate artifacts and then subtracting these components from the data, or by analyzing component time series for dimensionality reduction purposes. Components of blinking and eye muscle closure may be removed by independent component analysis.

S42: the related feature analysis and extraction comprises the following steps: fourier transform, nonlinear transform, time domain analysis, brain network analysis.

(4) Fourier transform: observing the state change of the brain, and extracting the characteristic of large contrast ratio as a distinguishing characteristic corresponding to the energy change of waves with different frequencies.

(5) Nonlinear indexes are as follows: entropy (En) is a physical quantity that can characterize the complexity of the brain. Around entropy-related feature extraction, for example: fisher Information (FI), spectral Entropy (SpEn), shannon's Entropy (ShEn), approximate Entropy (AppEximate Entropy, apEn), etc. furthermore, sample Entropy (Sample Entropy) and Permutation Entropy (Permutation Entropy), etc. may also be used as features.

(6) Time domain analysis: EEG recordings are observed and analyzed by utilizing the properties of EEG waveforms, such as amplitude, mean, variance, skewness, kurtosis and the like, and common time domain analysis methods in EEG signal research comprise zero crossing point analysis, histogram analysis, variance analysis, correlation analysis, peak detection, waveform parameter analysis, waveform identification and the like.

(7) Brain network analysis: power-based connectivity analysis the correlation between activities at the same or different frequencies, at the same or different points in time is calculated for correlating time-frequency power between two electrodes across time. Mutual information: information sharing between two variables is detected based on a distribution of values within the variables and a joint distribution calculation of the two (or more) variables.

S43: performing two-classification or multi-classification machine learning to obtain a tested electroencephalogram analysis multi-dimensional and multi-quality discrimination model;

the task of this step belongs to the supervised learning task, and in the supervised learning, the algorithm learns from the marked data. After understanding the data, the algorithm determines which tag should be assigned to the new data by associating the pattern with the unmarked new data. The supervised learning mainly comprises the following steps: classification and regression. Classification is a technique that determines a class to which a dependent variable belongs based on one or more independent variables. The regression problem predicts values from previously observed data. Thus, the task belongs to a classification task in supervised learning.

Common classification tasks mainly include two-classification and multi-classification. For example, if it is determined only whether fatigue belongs to a two-classification task, the fatigue level is determined to be a multi-classification task. The classification task is usually realized by a traditional classifier machine learning algorithm and a neural network deep learning algorithm.

(1) Support Vector Machine (SVM): the support vector machine is a generalized linear classifier for binary classification of data in a supervised learning mode, and a decision boundary of the support vector machine is a maximum margin hyperplane for solving learning samples. When linear classification is carried out, the classification surface is taken at a place with a larger distance from two types of samples, and the nonlinear classification is changed into linear classification in a high-dimensional space through high-dimensional space transformation when the nonlinear classification is carried out. The kernel function selection problem exists in the support vector machine, and a linear kernel (linear) is commonly used for solving the linear divisible problem, and other non-linear kernels are used for solving the linear inseparable problem (such as Gauss kernel function, polynomial kernel function and the like).

(2) A neural network framework:

fully connecting a neural network: the fully-connected neural network model is a multilayer perceptron (MLP), the principle of the perceptron is to find the most reasonable and robust hyperplane among classes, and the most representative perceptron is an SVM (support vector machine) algorithm. The neural network simultaneously uses a perceptron and bionics for reference, generally speaking, animal nerves can send each neuron after receiving a signal, and each neuron receives input and then is judged according to itself, and after being activated, output signals are generated and then are collected, so that information sources are identified and classified.

A convolutional neural network: convolutional Neural Networks (CNNs) are a class of feed forward Neural Networks (fed Neural Networks) that include convolution computations and have a deep structure, and are one of the representative algorithms of deep learning (deep learning). The convolutional Neural network has a representation learning (representation learning) capability, and can perform Shift-Invariant classification (Shift-Invariant classification) on input information according to a hierarchical structure thereof, so that the convolutional Neural network is also called a Shift-Invariant Artificial Neural network (SIANN). The convolutional Neural network is different from a common Neural network in that the convolutional Neural network comprises a feature extractor consisting of a convolutional layer and a sub-sampling layer (pooling layer), thereby greatly simplifying the model complexity and reducing the parameters of the model.

(3) Self-coding network: the self-coding network is one of the unsupervised learning fields, can automatically learn characteristics from unmarked data, is a neural network taking reconstructed input information as a target, can give out better characteristic description than original data, has stronger characteristic learning capability, and replaces the original data with the characteristics generated by the self-coding network in deep learning so as to obtain better effect.

The algorithm model construction needs to consider multi-aspect characteristics and then carry out fusion. For example, characteristics such as a coherence matrix, a mutual information matrix, a phase lag matrix and the like can be extracted in a brain network; statistical features, as well as non-linear features (entropy, etc.) are extracted for the raw signal.

Because the corresponding electroencephalogram signals are multichannel signals, the obtained characteristics often exist in a matrix form. Therefore, the matrix can be used as original data to be input into the neural network for training, and meanwhile, various aspects of brain network, time-frequency domain, statistical characteristics and nonlinear characteristics are considered, so that the network training can be firstly carried out by means of a machine learning method, then fusion and classification are carried out, and finally, an optimal algorithm model is formed.

After the electroencephalogram data are processed, the evaluation result of the tested brain state can be obtained.

The paradigm shift system shown in fig. 9 includes various paradigms such as perception, attention, memory, pointer, language, decision, emotion, etc., and these paradigms have some data evaluation indexes such as accuracy and response time. The above-mentioned various paradigms can form various psychological experiment paradigms, disease diagnosis paradigms and game paradigms.

The paradigm operating system can select difficulty and accurately record stimulation display time, tested reaction time and tested reaction accuracy, so that basic data are provided for paradigm behavior state evaluation system software. The paradigm algorithm can be determined by a percentile of the population to be tested that is normal. Firstly, counting the characteristics of reaction time and reaction accuracy in a test; counting the characteristics of the reaction time and the reaction accuracy rate between the tested samples; and counting percentiles of the tested reaction time and the reaction accuracy in the same test population.

In step S5, the present invention may include a Posner attention paradigm (spatial paradigm) and an n-back working memory paradigm, among other types of paradigms.

As shown in fig. 10, the Posner paradigm consists of several trials, in each of which spatial cue stimuli (arrows to the left or right) appear first in the center of the visual field, directing the subject to look implicit (not turning the eye, keeping looking at the center of the visual field) to the spatial location to which the cue points. After the thread disappears, a short and random-length delay period is experienced. During the delay period, the subject is asked to maintain spatial attention to the pointing position. After the delay period, a circular Gabor grid standard stimulus (50% probability) or a target stimulus (50% probability) may appear at the pointed location (80% probability, valid stimulus) or at the opposite (20% probability) of the pointed location (null stimulus). Whether the stimulus is presented in an active or inactive position, the subject is asked to respond as quickly as possible to whether the stimulus is a target stimulus. The difference between the target stimulation and the standard stimulation is the density of the Gabar grid, and the difficulty in distinguishing the target stimulation from the standard stimulation can be increased by setting the density difference to be smaller, so that stronger attention is required to be adopted to complete the task. The accuracy and response time were recorded for assessment of attention levels and attention performance.

The experimental time is 5 minutes in total, the grating stripe angle is 60 degrees/120 degrees, the waiting time before the indication arrow is displayed is 2 seconds, the indication arrow display time is 0.2 seconds, the error conductivity of the indication arrow is 20 percent, the waiting time after the indication arrow disappears is 1-4 seconds, the grating display time is about 0.02 seconds, and the maximum effective reaction time is 1-3 seconds.

As shown in FIG. 11, the n-back paradigm is a classical working memory paradigm. The n-back paradigm requires the subject to compare the just-presented stimulus to the nth preceding stimulus and manipulate the cognitive load by controlling the number of stimuli between the current stimulus and the target stimulus. The task type adopts a letter matching task. The 2-back paradigm is used in this study. The experiment time is 5 minutes, the character display time is 0.5 seconds, the maximum effective response time is 2 seconds (including the letter display time), and the waiting time for displaying the next character is 1 second.

Counting the characteristics of the reaction time and the reaction accuracy in the test; counting the characteristics of the reaction time and the reaction accuracy rate between the tested samples; and counting the percentile of the tested reaction time and the reaction accuracy in the same test population.

The invention provides a closed-loop self-adaptive transcranial electrical stimulation device and a method, which comprise a high-precision transcranial electrical stimulation and electroencephalogram acquisition integrated technology, an acquisition and stimulation compatible electrode, a neural state analysis and evaluation model based on electroencephalogram and paradigm fusion, an accurate neural regulation and control model and the like. Wherein, the high-precision transcranial electrical stimulation can effectively lock the stimulation area, and realize accurate regulation and control. The electroencephalogram collection integration technology solves the key technical problems of circuit multistage protection, multistage switching, multistage isolation and the like. The acquisition stimulation compatible electrode can be adapted to the current and impedance requirements of two different scenes of acquisition and stimulation, and the same electrode can acquire signals and can perform two different operation functions of stimulation. The neural state analysis and evaluation model establishes a reliable algorithm model through modes such as pattern recognition, machine learning and the like based on specific scene requirements, and provides data guidance for closed-loop regulation and control. The precise neural regulation model sets targeted differential neural regulation parameters based on different evaluation results.

The closed-loop self-adaptive transcranial electrical stimulation device and the method are an integrated neural regulation and control mode, form closed-loop self-adaptive linkage of three functions of acquisition, stimulation and evaluation, fully play the potential and advantages of the transcranial electrical stimulation technology, dynamically adjust an intelligent self-adaptive treatment mode of the neural regulation and control mode according to a brain state evaluation result, provide visual diagnosis and treatment data guidance for special personnel, scientific research personnel and clinicians, greatly improve the feedback efficiency of evaluation analysis and regulation and control effects, and improve the pertinence and effectiveness of regulation and control, and are a novel intelligent neural regulation and control method.

Claims

1. A closed-loop adaptive transcranial electrical stimulation device, characterized in that: the transcranial electrical stimulation unit and the electroencephalogram acquisition unit share the same set of electrodes through a switching circuit, and the switching circuit is controlled through the control unit; the transcranial electrical stimulation unit comprises a positive and negative high-voltage circuit, the positive and negative high-voltage circuit is connected to the electrodes through a switching circuit, the electroencephalogram acquisition unit comprises a filter circuit and an ADC (analog-to-digital converter) acquisition circuit, the filter circuit is connected to the electrodes through the switching circuit, and the electroencephalogram acquisition unit comprises a filter circuit and an ADC acquisition circuit which are connected.

2. The closed-loop adaptive transcranial electrical stimulation device according to claim 1, wherein: the electroencephalogram acquisition unit comprises an electroencephalogram acquisition circuit connected with the electrodes, and can be constructed into 16 channels, 32 channels, 64 channels and 128 channels by adopting Daisy Chain Daisy-Chain cascade connection.

3. The closed-loop adaptive transcranial electrical stimulation device according to claim 1, wherein: the transcranial electric stimulation circuit is also provided with a constant current feedback detection circuit, and the constant current feedback detection circuit sends stimulation current values of the positive and negative high-voltage circuits to the control unit.

4. The closed-loop adaptive transcranial electrical stimulation device according to claim 1, wherein: the switching circuit controls the connection of the electrodes and the transcranial electrical stimulation unit or the electroencephalogram acquisition unit, the control unit controls the switching circuit to realize a synchronous control mode and an asynchronous control mode of the electrodes, and the synchronous control mode refers to the fact that the electrodes simultaneously perform transcranial electrical stimulation or the electrodes simultaneously perform electroencephalogram acquisition; the asynchronous control mode is that part of the electrodes perform transcranial electrical stimulation while part of the electrodes perform electroencephalogram acquisition.

5. A closed-loop adaptive transcranial electrical stimulation method comprises the following steps:

s1: selecting different test paradigms according to different tested objects;

s5: behavior data are collected and a paradigm is evaluated;

s8: calling a regulation model, and regulating transcranial electrical stimulation output parameters according to the deviation value;

6. The closed-loop adaptive transcranial electrical stimulation method according to claim 1, characterized in that: in step S3, the output parameter setting step is as follows:

s11: determining a pattern of transcranial electrical stimulation;

s2: determining a location of transcranial electrical stimulation;

the position is selected from an international standard electroencephalogram position distribution diagram, and is provided with 1 to N electrical stimulation positions, wherein the positive electrode is 1-N-1 positions, the negative electrode is 1-N-1 positions, and a regulating circuit combination is formed together, wherein N is a natural number and is more than or equal to 2;

s3: determining specific parameters of transcranial electrical stimulation;

the method comprises the following steps of stimulating current polarity, stimulating current duration, stimulating current amplitude, stimulating current bias, stimulating current frequency and stimulating current duty ratio;

s4: determining an evaluation mode of a regulation target;

stopping the current regulation when the regulation target is achieved, or continuing to regulate according to a fixed time, thereby consolidating the target; when the regulation target is not achieved, the regulation can be continuously carried out according to a set time, and the regulation is stopped no matter whether the regulation target is achieved or not; or directly stopping the regulation when the regulation target is not reached.

7. The closed-loop adaptive transcranial electrical stimulation method according to claim 1, characterized in that: in the step S3, the electroencephalogram state evaluation comprises data preprocessing, relevant feature analysis and extraction, machine learning and acquisition of a tested electroencephalogram state evaluation result;

the data preprocessing aims to remove bad data or data with full artifacts without changing clean data, or to change the clean data by applying a filter or spatial transformation to facilitate analysis, including re-referencing, filtering, and independent principal component analysis;

the machine learning and the tested electroencephalogram state evaluation result obtaining method realize classification tasks in supervised learning, and comprise processing modes of a support vector machine, a neural network framework and a self-coding network.

8. The closed-loop adaptive transcranial electrical stimulation method according to claim 1, characterized in that: in step S5, the evaluation of the paradigm behavior state includes: and counting the reaction time of the test, counting the reaction accuracy of the test, and calculating the reaction accuracy characteristics of the test time of the test.

9. The closed-loop adaptive transcranial electrical stimulation method according to claim 1, wherein: in step S6, the accuracy response time of the deviation state and the non-deviation state is used as a behavior index and introduced into a discriminant model,

the third method is to perform algorithm fusion of different data modes through the difference between the population of the absolute control group.

10. The closed-loop adaptive transcranial electrical stimulation method according to claim 1, wherein: in step S5, the paradigm evaluation includes, but is not limited to, a Posner attention paradigm and an n-back working memory paradigm.