阅读说明：本技术 一种基于本征模函数相干分析的事件诱发电位提取装置 (Event evoked potential extraction device based on eigenmode function coherent analysis ) 是由 谭波 李凌 于 2021-09-24 设计创作，主要内容包括：本发明公开了一种基于本征模函数相干分析的事件诱发电位提取装置,该发明的主要技术特点包括：通过脑电测试系统采集包含自发脑电和事件刺激下诱发脑电混合信号；将脑电混合信号经过预处理后输入本征模函数序列获取单元进行分解并获得合格的IMFs序列；然后将上述IMFs分量序列输入到IMFs相干分析及筛选单元,并结合刺激前自发脑电信号的特性,计算每个IMFs分量与该信号刺激前自发脑电的相干值,筛选并排除与自发脑电信号相干性最大的IMF分量；再将剩下的分量进行重组,所得重组信号即为事件诱发信号。本发明是基于脑电信号自身的局部特征来快速有效地提取到高质量的ERP信号,对先验知识和重复刺激次数无固定要求,相对传统叠加平均方法提取更高效和更准确。(The invention discloses an event evoked potential extraction device based on eigenmode function coherent analysis, which is mainly technically characterized by comprising the following steps: acquiring an electroencephalogram mixed signal containing self-generating electroencephalograms and induced electroencephalogram under the stimulation of an event through an electroencephalogram test system; preprocessing the electroencephalogram mixed signal, inputting the preprocessed electroencephalogram mixed signal into an eigenmode function sequence acquisition unit for decomposition, and acquiring a qualified IMFs sequence; then inputting the IMFs component sequence into an IMFs coherence analysis and screening unit, calculating the coherence value of each IMFs component and the spontaneous electroencephalogram before stimulation of the signal by combining the characteristics of the spontaneous electroencephalogram before stimulation, and screening and eliminating the IMF component with the maximum coherence with the spontaneous electroencephalogram; and recombining the rest components to obtain a recombined signal, namely the event-induced signal. The method is used for quickly and effectively extracting the high-quality ERP signals based on local characteristics of the electroencephalogram signals, has no fixed requirements on prior knowledge and repeated stimulation times, and is more efficient and accurate in extraction compared with the traditional superposition averaging method.)

1. An event evoked potential extracting device based on eigenmode function coherent analysis, the device comprising: the system comprises an electroencephalogram mixed signal acquisition and preprocessing unit, an eigenmode Function Sequence (IMFs) acquisition unit, a coherent analysis and screening unit of the eigenmode function sequence and an event-induced electroencephalogram signal recombination unit;

the electroencephalogram mixed signal acquisition and preprocessing unit acquires electroencephalogram mixed signals before and after event stimulation, and carries out preprocessing such as baseline calibration, average reference, band-pass filtering and the like on the electroencephalogram mixed signals; then transmitting to an eigenmode function sequence acquisition unit;

the eigenmode function sequence acquisition unit performs empirical mode decomposition on each channel of the preprocessed electroencephalogram mixed signals to acquire all intermediate signals of the electroencephalogram mixed signals of each channel; then judging whether each intermediate signal is a qualified IMF component by using two constraint conditions of the IMF component in the empirical mode algorithm, screening the qualified IMF component of each channel and forming an eigenmode function sequence of the channel; then transmitting the sequence to a coherent analysis and screening unit of an eigenmode function sequence;

the coherent analysis and screening unit of the eigenmode function sequence takes an event stimulation point as a boundary point, respectively extracts a section of an eigenmode function sequence with the same length behind the event stimulation point obtained by the electroencephalogram signal before the event stimulation point (namely, an spontaneous electroencephalogram signal used as a reference) and the eigenmode function sequence obtained by the eigenmode function sequence obtaining unit, and calculates and sorts the coherent values of all eigenmode function components in the spontaneous electroencephalogram signal before the event stimulation point and the eigenmode function sequence after the stimulation, screens out the IMF component with the maximum coherence behind the event stimulation point, and judges whether to carry out secondary empirical mode decomposition or not according to the coherent analysis condition of the IMF component;

if the second-stage EMD decomposition is needed, calculating the correlation value of the eigenmode function sequence after the second-stage decomposition and the spontaneous electroencephalogram reference signal before the event stimulation point again, screening out the eigenmode function component with the maximum correlation value, and setting the eigenmode function component to zero; simultaneously, transmitting the two-stage IMF components after the processed event stimulus point to an event-induced electroencephalogram signal recombination unit;

the event-induced electroencephalogram signal recombination unit firstly eliminates the residual components of the received eigenmode function components of all levels, and utilizes the remaining eigenmode function components to carry out electroencephalogram signal recombination to obtain the channel event-induced electroencephalogram signal.

2. The eigenmode function coherence analysis-based event evoked potential extraction apparatus as claimed in claim 1, wherein the method for determining the intermediate signal as qualified eigenmode function component in said eigenmode function sequence acquisition unit is: 1) in the intermediate signal sequence, the number N of extreme points and the number of zero-crossing points must be equal or the difference cannot exceed 1 at most, and the extreme points comprise maximum values and minimum values;

2) and an upper envelope f formed by the local maximum and local minimum of the signal sequence at any time point_max(t) and the lower envelope f_minThe average value of (t) must be 0, i.e. the waveform of the whole sequence must be locally symmetrical, satisfying Mean (f)_max(t)+f_min(t))＝0。

3. The device for extracting event evoked potential based on eigenmode function coherent analysis as claimed in claim 1, wherein said unit for coherent analysis and screening of eigenmode function sequence determines whether to perform secondary EMD decomposition according to the coherent analysis condition of eigenmode function component:

1) in the first-stage EMD decomposition, if the IMF component with the maximum coherence value found by the coherence analysis and screening unit of the eigenmode function sequence is the 1 st IMF high-frequency component, the component is directly set to zero without performing the second-stage EMD decomposition;

2) if the 1 st IMF high-frequency component after the first-stage EMD decomposition is not the IMF component with the maximum coherent value obtained by screening, the 1 st IMF high-frequency component is input into a coherent analysis and screening unit of the eigenmode function sequence to carry out second-stage EMD decomposition, the second-stage IMF component with the maximum coherent value is screened again by using the coherent analysis and screening unit of the eigenmode function sequence, and the IMF components with the maximum coherent value after the first-stage EMD decomposition and the second-stage EMD decomposition are respectively set to be zero.

Technical Field

The invention relates to a method and a device for extracting event-related potential, belongs to the technical field of biomedical signal processing, and particularly relates to a method and a device for extracting event evoked potential based on eigenmode function coherent analysis.

Background

During the activity of cerebral cortex, a large number of neurons generate postsynaptic potentials synchronously, and the postsynaptic potentials are summed up to form an Electroencephalogram (EEG) with complex and irregular components. The strength of the brain electrical signals is weak, and single neuron electrical activity cannot be recorded on the scalp, and only synchronous discharge of neuron groups can be recorded. Normal spontaneous brain electrical activity is usually between a few microvolts and tens of microvolts, and when one or more (or various) stimuli are applied to a certain part of the sensory system or brain, the applied stimuli or the removed stimuli may cause changes in the electrical potential of the brain area, i.e., event-related potentials (ERP), which reflect changes in the neural electrophysiology of the brain during the cognitive process. The ERP wave amplitude induced by general stimulation is about 2-10 mV, which is much smaller than the spontaneous potential (EEG); therefore, the ERP is usually buried in the spontaneous electroencephalogram, and the ERP and the spontaneous electroencephalogram form the relationship between small signals and large noise, thereby causing great difficulty in extracting and researching the ERP.

Reportedly, ERPs are usually derived from the postsynaptic potential of cortical pyramidal neurons, and are produced by ion flow through the cell membrane in response to neurotransmitters that bind to postsynaptic cell receptors. When postsynaptic potentials occur simultaneously, the field potential sum is detected on the scalp. Therefore, a certain idea is provided for extracting the components of the early ERP: overlapping a plurality of sections of electroencephalograms with the same stimulation, wherein the spontaneous electroencephalograms or noises are randomly changed and have high or low levels, so that positive and negative counteractions can be generated when the electroencephalograms or noises are overlapped with each other; the ERP signals are constant, so that the ERP signals cannot be offset, the amplitude of the ERP signals is increased continuously, and the ERP signals are displayed after being superposed for a certain number of times. However, the method not only needs a large number of repeated stimulation tests (usually the times are more than 100), but also causes excessive manpower and resource consumption, and ideal induced electroencephalogram is not obtained in each experiment, so that the randomness in the actual operation process is high, the tested emotion can be changed due to continuous repeated stimulation, and further the research result of ERP is influenced.

In recent years, many other methods for extracting ERP have been proposed, such as AR model, neural network algorithm, independent component analysis, principal component analysis, and the like. However, because of the adaptability and fatigue of the brain nerve to each stimulation, parameters such as the amplitude and the latency of the ERP caused by each stimulation are not completely the same, and most methods cannot sufficiently reflect information such as the latency and the amplitude of a single evoked potential. In addition, the inherent limitations and applicability of each method are different, so that no one of the current methods is well adapted to all types of event-evoked potential extraction algorithms.

Disclosure of Invention

In order to overcome some defects of the existing ERP extraction technology, the uncertainty of extraction effect caused by repeated stimulation is reduced, and the extraction precision and the practicability of the ERP are improved. The invention provides an event evoked potential extraction method and device based on eigen mode function coherent analysis, wherein the method makes full use of the time scale characteristics of data to adaptively decompose signals, and decomposes original electroencephalogram signals into a plurality of connotative modal components (IMFs components) to reflect the local characteristics of the signals; and then, coherent analysis on a frequency domain is carried out by combining with spontaneous electroencephalogram to screen out spontaneous components in the stimulated electroencephalogram signals, and the spontaneous components are eliminated so as to accurately and quickly extract the ERP signals of each stimulation. The method has no fixed requirements on prior knowledge and repeated stimulation times, is more efficient and accurate in extraction compared with the traditional superposition averaging method, still has a good effect when extracting the ERP signal of single-channel single stimulation, and provides help for subsequent ERP signal research and application.

The invention is realized by the following technical scheme: an event evoked potential extracting device based on eigenmode function coherent analysis, the device comprising: the system comprises an electroencephalogram mixed signal acquisition and preprocessing unit, an eigenmode Function Sequence (IMFs) acquisition unit, a coherent analysis and screening unit of the eigenmode function sequence and an event-induced electroencephalogram signal recombination unit;

the electroencephalogram mixed signal acquisition and preprocessing unit acquires electroencephalogram mixed signals before and after event stimulation, and carries out preprocessing such as baseline calibration, average reference, band-pass filtering and the like on the electroencephalogram mixed signals; then transmitting to an eigenmode function sequence acquisition unit;

the eigenmode function sequence acquisition unit performs empirical mode decomposition on each channel of the preprocessed electroencephalogram mixed signals to acquire all intermediate signals of the electroencephalogram mixed signals of each channel; then judging whether each intermediate signal is a qualified IMF component by using two constraint conditions of the IMF component in the empirical mode algorithm, screening the qualified IMF component of each channel and forming an eigenmode function sequence of the channel; then transmitting the sequence to a coherent analysis and screening unit of an eigenmode function sequence;

the coherent analysis and screening unit of the eigenmode function sequence takes an event stimulation point as a boundary point, respectively extracts a section of an eigenmode function sequence with the same length behind the event stimulation point obtained by the electroencephalogram signal before the event stimulation point (namely, an spontaneous electroencephalogram signal used as a reference) and the eigenmode function sequence obtained by the eigenmode function sequence obtaining unit, and calculates and sorts the coherent values of all eigenmode function components in the spontaneous electroencephalogram signal before the event stimulation point and the eigenmode function sequence after the stimulation, screens out the IMF component with the maximum coherence behind the event stimulation point, and judges whether to carry out secondary empirical mode decomposition or not according to the coherent analysis condition of the IMF component;

if the second-stage empirical mode decomposition is needed, calculating the correlation value of the eigenmode function sequence after the second-stage decomposition and the spontaneous electroencephalogram reference signal before the event stimulation point again, screening out the eigenmode function component with the maximum correlation value, and setting the eigenmode function component to zero; simultaneously, transmitting the two-stage IMF components after the processed event stimulus point to an event-induced electroencephalogram signal recombination unit;

the event-induced electroencephalogram signal recombination unit firstly eliminates the residual components of the received eigenmode function components of all levels, and utilizes the remaining eigenmode function components to carry out electroencephalogram signal recombination to obtain the channel event-induced electroencephalogram signal.

Further, the method for determining the intermediate signal as the qualified eigenmode function component in the eigenmode function sequence obtaining unit is as follows:

1) in the intermediate signal sequence, the number N of extreme points and the number of zero-crossing points must be equal or the difference cannot exceed 1 at most, and the extreme points comprise maximum values and minimum values;

2) and an upper envelope f formed by the local maximum and local minimum of the signal sequence at any time point_max(t) and the lower envelope f_minThe average value of (t) must be 0, i.e.of the entire sequenceThe waveform must be locally symmetric to satisfy Mean (f)_max(t)+f_min(t))＝0；

Further, the method for judging whether to perform secondary decomposition according to the coherent analysis condition of the eigenmode function component in the coherent analysis and screening unit of the eigenmode function sequence comprises the following steps:

1) in the first-stage EMD decomposition, if the IMF component with the maximum coherence value found by the coherence analysis and screening unit of the eigenmode function sequence is the 1 st IMF high-frequency component, the component is directly set to zero without performing the second-stage EMD decomposition;

2) if the 1 st IMF high-frequency component after the first-stage EMD decomposition is not the IMF component with the maximum coherent value obtained by screening, inputting the 1 st IMF high-frequency component into a coherent analysis and screening unit of the eigenmode function sequence for second-stage EMD decomposition, screening out the second-stage IMF component with the maximum coherent value again by using the coherent analysis and screening unit of the eigenmode function sequence, and simultaneously respectively setting the IMF component with the maximum coherent value after the first-stage EMD decomposition and the second-stage EMD decomposition to zero;

the invention provides an event evoked potential extraction device based on eigen mode function coherent analysis, which makes full use of the time scale characteristics of data to adaptively decompose electroencephalogram signals, and decomposes original electroencephalogram mixed signals into a plurality of connotative modal components (IMFs components) reflecting the local characteristics of the signals; and then, coherent analysis is carried out by combining with spontaneous electroencephalogram, and the ERP signal of single-channel single stimulation is extracted accurately and quickly. Compared with the traditional superposition averaging method, the extraction efficiency, accuracy and robustness are improved, the repeated stimulation times are not too high, the extracted ERP components are more obvious, and the curve is smoother. In addition, the whole process of the extraction and processing of the ERP can be effectively integrated in one device, so that the method has better integration and portability, and can provide certain help for the follow-up research of the ERP.

Drawings

FIG. 1 is a block diagram showing the detailed operation and structure of the apparatus of the present invention

FIG. 2 is an IMFs sequence screened by empirical mode decomposition FC3 channel electroencephalogram signal

FIG. 3 is a graph comparing the effect of extracting FC3 channel event evoked potential according to the present invention and the traditional superposition averaging

FIG. 4 is a schematic diagram of the present invention for extracting the event evoked potential and original EEG signal of 64 channels of the whole brain

Detailed Description

In order to facilitate better understanding and implementation of the present invention for those of ordinary skill in the art, the present invention will be further described with reference to the accompanying drawings and examples, which are used for explaining the present invention and are not limited thereto. The following description of the embodiments of the present invention with reference to the accompanying drawings illustrates the following steps:

1. electroencephalogram mixed signal acquisition and preprocessing unit

A. Firstly, recording electroencephalogram mixed signals (the sampling rate is 1000Hz) before and after an event stimulation (such as picture stimulation) through a 64-channel electroencephalogram test system, simultaneously setting the number of repeated stimulation times to be 65 times, setting the recording time of each stimulation time to be 1200ms (including 200ms before the stimulation and 1000ms after the stimulation), simultaneously carrying out pretreatment such as baseline calibration, average reference, band-pass filtering and the like on the recorded electroencephalogram mixed signals, and then obtaining a section of three-dimensional signals (namely 64 x 1200) including an event evoked potential.

2. Eigenmode Function Sequences (IMFs) acquisition unit

B. Extracting the whole EEG signal (such as X (t)) of one channel single stimulation after pretreatment, inputting the signal into an eigenmode Function Sequence (IMFs) acquisition unit for empirical mode decomposition processing to obtain a plurality of intermediate signals R (t) contained in the channel (FIG. 2 is intermediate signals obtained by EMD decomposition of an FC3 channel);

C. and (3) judging whether the intermediate signal is a qualified IMF component or not by using a constraint condition of the IMFs component in the empirical mode algorithm, and finally screening an IMF component sequence of which the channel meets the condition (the FC3 channel is decomposed into 5 qualified IMFs and 1 Res in the drawing). The empirical mode decomposition comprises the following specific steps:

1) finding all local maximum and minimum points of the EEG signal (EEG (t)) input into the channel;

2) carrying out envelope fitting on the extreme points; fitting all maximum value points into an upper envelope line Up _ envelop (t) by utilizing a spline interpolation function; similarly, fitting all minimum value points to obtain a lower envelope line Low _ envelop (t), and then obtaining a Mean envelope signal Mean _ EEG (t) of the upper envelope line and the lower envelope line;

3) subtracting the Mean envelope Mean _ EEG (t) from the input original signal EEG (t) to obtain an intermediate signal R (t); repeating the above process to obtain several intermediate signals R (t).

4) Judging whether each intermediate signal R (t) meets two constraint conditions of the IMF component, if so, obtaining a qualified IMF component, and if not, returning to the previous step for repeated iteration;

5) if the intermediate component R (t) satisfies the constraint condition, it becomes the first high frequency component and is marked as IMF₁And mean envelope mean (t) versus IMF₁Is a low frequency component. Then, taking mean (t) as the input original signal again to continue to decompose to obtain the next IMF, and stopping decomposition until the residual component is a monotonic function or a constant. Typically the residual component is denoted as res (t).

3. Coherent analysis and screening unit for IMFs component

D. Taking an event stimulation point as a demarcation point, and extracting an EEG signal EEG _ before (t) 200ms before the stimulation point as a reference signal (namely a spontaneous EEG signal); simultaneously intercepting the IMFs (t) component after the stimulation with the same length as a target signal of coherent analysis (namely an electroencephalogram mixed signal containing an event evoked potential);

E. and (3) carrying out coherent analysis on a frequency domain on the target signals IMFs (t) and the reference signals EEG _ before (t), calculating a coherent value between each IMF component and the spontaneous electroencephalogram signals, and screening out the IMF component with the maximum coherence after the first-level EMD decomposition, wherein the IMF component with the maximum coherence is the 1 st IMF high-frequency component, and the second-level EMD decomposition is not required.

4. Event-induced electroencephalogram signal recombination unit

F. And (4) setting the IMF component with the maximum coherence obtained by screening to be zero, and simultaneously removing the residual component Res (t) of the empirical mode decomposition. And superposing and recombining the electroencephalogram signals by utilizing the IMFs component after each stage of screening, wherein the obtained recombined signals are the event-induced electroencephalogram signals of the channel.

G. For multiple repeated stimulation tests (such as 65 stimulation times), the above process may be repeated to obtain the event evoked potential generated by each stimulation of the channel, and the average value is the event evoked potential averaged by the channel under multiple repeated stimulation. Meanwhile, the steps can be repeatedly completed for all channels of the whole brain, and the event-induced electroencephalogram signals of 64 channels of the whole brain can be obtained.

In order to further illustrate the beneficial effects of the invention, we compare the ERP signal extracted by the device of the invention and the traditional superposition averaging method. As shown in fig. 3, compared with the conventional superposition averaging method, in the embodiment of the invention, a good extraction effect can be achieved with a few times of repeated stimulation (most of ERP components are present around 25 times) when the channel FC3 is extracted, and the extraction efficiency and the accuracy of the invention are both higher (the coherence value of the ERP signal extracted by the invention and the reference standard induced electroencephalogram is about 0.8953, while the coherence value obtained after the conventional superposition averaging is performed for 60 times is only 0.7955). In addition, as shown in fig. 4, the extraction processing is performed on the electroencephalogram signals of 64 channels of the whole brain, and the result shows that basically all channels can extract certain ERP components, so that the method has strong universality and practicability. The invention can effectively integrate the whole process of the ERP extraction processing in one device, has better integration and portability, and provides certain help for the subsequent ERP research.