CN114041803A - TMS pulse artifact removing device based on interval median filtering - Google Patents

Abstract

The invention provides a TMS pulse artifact removing device based on interval median filtering, and belongs to the technical field of biomedical signal detection and processing. The device includes: the electroencephalogram signal acquisition and preprocessing unit acquires an electroencephalogram signal through an electroencephalogram test system, preprocesses the acquired electroencephalogram signal, and then sequentially passes through the wavelet modulus maximum processing unit, the interval median filtering unit and the wavelet reconstruction unit to obtain the electroencephalogram signal without TMS pulse artifacts. The method can effectively and quickly remove the influence of TMS electromagnetic pulse artifacts in the electroencephalogram signals, and keep more stable and accurate electroencephalogram signals, is suitable for processing the electroencephalogram signals stimulated by single-channel single-pulse TMS, greatly reduces the dependence of other methods such as ICA on multi-channel repeated stimulation, and is beneficial to the subsequent further analysis and research of transcranial magnetic stimulation and electroencephalogram data.

Description

TMS pulse artifact removing device based on interval median filtering

Technical Field

The invention belongs to the technical field of biomedical signal detection and processing, and particularly relates to a TMS pulse artifact removing method and device integrating wavelet transformation and interval median filtering.

Background

Transcranial Magnetic Stimulation (TMS) is a non-invasive electromagnetic neuromodulation technique that uses a transient Magnetic field to pass through the skull to act on the central nervous system of the brain and then generate an induced current to activate the cortex to affect intracerebral metabolism and neuroelectrical activity. The TMS is divided into three stimulation modes, namely single-pulse TMS (sTMS), double-pulse TMS (pTMS) and repetitive TMS (rTMS), according to different stimulation pulses, and the application ranges of the different stimulation modes are different; the former two modes are mainly used for conventional electrophysiological examination and the study of nerve facilitation or inhibition, while rTMS is more used for diagnosing or treating some mental diseases by producing different neurophysiological effects, such as epilepsy, depression, seminal emission syndrome, obsessive compulsive disorder, post-traumatic stress disorder, and the like, and also has certain effects on improving the nerve rehabilitation treatment aspects such as sleep quality and the like. At present, three technical characteristics of painlessness, no wound and no damage of TMS are widely applied to a plurality of neuroscience and clinical research directions.

In addition to the above-described uses for psychiatric disease treatment, TMS is also commonly used to actively intervene in normal brain function to explore the relationship between normal brain and behavioral activity, revealing causal mechanisms between brain activity and task completion. However, the sudden change electromagnetic pulse generated in the TMS stimulation process can cover a plurality of important spontaneous electroencephalogram components and TMS evoked components, and further, the subsequent analysis and research of the spontaneous electroencephalogram and TMS evoked electroencephalogram signals are seriously influenced. The traditional method for solving the problem is to extract an Independent Component of the electromagnetic pulse signal and remove the Independent Component through Independent Component Analysis (ICA), but the ICA is found to be implemented only by acquiring multi-channel electroencephalogram signals when the Independent Component is extracted in actual Analysis; and a small amount of electromagnetic pulse artifacts still remain in the extracted independent components, and the interference of TMS electromagnetic pulses cannot be truly and completely removed. Therefore, how to remove the influence of the electromagnetic pulse artifact from the brain electrical mixed signal after TMS stimulation with high quality is a problem to be solved.

Disclosure of Invention

In order to overcome the defects and problems of the conventional ICA method in removing the TMS pulse artifact, the efficiency and the accuracy of removing the TMS electromagnetic pulse signal are further improved. The invention provides a TMS pulse artifact removing device based on interval median filtering, which effectively removes TMS pulse artifacts in electroencephalogram mixed signals by fusing wavelet transformation and interval median filtering. Firstly, acquiring electroencephalogram data including before and after TMS stimulation through an electroencephalogram test system and a TMS stimulation system, and finishing preprocessing; then extracting a single-channel electroencephalogram signal, performing multilayer wavelet decomposition on the channel electroencephalogram mixed signal by utilizing a one-dimensional wavelet mode maximum algorithm, and extracting wavelet coefficients under different scales; and simultaneously, automatically detecting all extreme points of wavelet coefficients of each layer, positioning a demoulded maximum value point (namely a TMS pulse position), searching adjacent 5 extreme points by taking the maximum value point as a center, recording a data interval of the wavelet coefficients corresponding to the extreme values, and performing interval median filtering processing by taking the data interval as target data of median filtering. And finally, reconstructing the filtered wavelet coefficients into the electroencephalogram signals without TMS pulse artifacts by using a wavelet reconstruction function, namely, the electroencephalogram signals without the TMS pulse artifacts. The method organically integrates the advantages of wavelet modulus maximum and interval median filtering and is applied to removing electromagnetic pulse artifacts in TMS stimulation; compared with the conventional ICA removal method, the method has the advantages that the TMS pulse stimulation point can be quickly and accurately positioned, electromagnetic pulse artifacts caused by TMS stimulation can be effectively removed, and the single-channel TMS-stimulated brain electrical signals can be processed; meanwhile, compared with an ICA removal method, the method has higher efficiency and accuracy, and the waveform of the reconstructed electroencephalogram signal is smoother and cleaner, so that the method has important significance on the fusion research of TMS and electroencephalogram signals.

The invention is realized by the following technical scheme:

a TMS pulse artifact removing device based on interval median filtering comprises: the electroencephalogram signal acquisition and preprocessing unit firstly acquires an electroencephalogram signal, preprocesses the acquired electroencephalogram signal, and then sequentially passes through the wavelet modulus maximum processing unit, the interval median filtering unit and the wavelet reconstruction unit to obtain the electroencephalogram signal without TMS electromagnetic pulse artifacts;

the electroencephalogram signal acquisition and preprocessing unit acquires electroencephalogram mixed signals before and after TMS stimulation, and carries out drift and power frequency interference removal processing on the electroencephalogram mixed signals to obtain preprocessed electroencephalogram mixed signals;

the processing method in the wavelet mode maximum processing unit comprises the following steps:

step 1: extracting an electroencephalogram mixed signal of one channel, and performing wavelet decomposition on the electroencephalogram mixed signal by utilizing wavelet transformation to obtain an approximate wavelet coefficient cD and a detail wavelet coefficient cA under different scales;

step 2: detecting extreme points of wavelet coefficients under all scales by using a wavelet modulus maximum linear search algorithm, screening out a maximum value of a modulus from the extreme points, simultaneously retrieving a data interval adjacent to the extreme points by taking a time point corresponding to the maximum value as a center, recording wavelet coefficients corresponding to the extreme points, and taking the interval as a target data interval of interval median filtering processing;

further, the method for searching the wavelet coefficient extreme point under each scale by the wavelet modulus maximum linear search algorithm comprises the following steps:

S₁search time interval [ t ] on decomposition scale J₀,t₁]The amplitude of the wavelet coefficient at a time t is arbitrarily selected as an initial maximum value | W_maxL, |; meanwhile, setting a maximum threshold T (usually the threshold is a linear average value of the wavelet coefficients) of the wavelet coefficient modulus at the scale according to specific requirements;

S₂from t₀Starting at time t ═ t₀(ii) a Compare | W (t) | with | W_maxThe magnitude of the value, | W (t) | > | W_maxIf W (T) is greater than T, let W_maxW (t), and recording the corresponding time t; if W (t) is less than or equal to W_maxIf w ≦ T, then the comparison is carried out until the next time, i.e. T ≦ T₀+ 1; wherein | represents a modulo operation;

S₃repeating the above process until the search time is foundInterval [ t ]₀,t₁]Maxima of the modulus of the inner wavelet coefficients.

The processing method in the interval median filtering unit comprises the following steps:

constructing a median filter according to the target data obtained by the wavelet modulus maximum processing unit, performing interval median filtering on pulse data intervals in the detail wavelet coefficients under each scale by using the median filter, and performing data reconstruction on the detail wavelet coefficients subjected to the interval median filtering to obtain new wavelet coefficients;

further, the interval median filter construction and filtering method is as follows:

for the target data interval [ t₀,t₁]If the time series of detail wavelet coefficients is x₁,x₂，...，x_nTaking the window length m, m being an odd number, and extracting m numbers, e.g. x, from the input sequence_i-v，…,x_i-1，…x_i，…x_i+1，…x_i+v，…x_i-vWherein i is the central position of the window, satisfies

Sequencing the m sequence values from small to large according to the numerical values, wherein the sequence after sequencing is as follows: y is_i-v，…,y_i-1，…,y₁，…,y_i+1，…,y_i+vThen, the median filter takes the value corresponding to the middle sequence number as the output of the interval median filter, and is expressed as: y is_i＝Med{y_i-v,…,y_i,…,y_i+vIn which i ∈ Z,

med {. is } represents taking the median operation.

The processing method in the wavelet reconstruction unit comprises the following steps:

and reconstructing the new wavelet coefficient obtained by the interval median filtering unit into a new electroencephalogram signal by using a wavelet reconstruction function, namely the new electroencephalogram signal for filtering the TMS pulse artifact.

The invention has the advantages and beneficial effects that:

the invention provides a TMS pulse artifact removing device based on interval median filtering, which organically integrates the advantages of a modulus maximum algorithm of wavelet analysis and the interval median filtering to efficiently remove electromagnetic pulse artifacts in TMS-stimulated electroencephalogram mixed signals. The invention not only makes full use of the multi-scale characteristics of wavelet mode maximum analysis, decomposes the EEG signal containing pulse artifacts into multi-scale wavelet coefficients, rapidly and accurately positions the pulse stimulation points of each layer of wavelet coefficients, but also integrates the most effective median filtering for filtering pulse interference to process the pulse data interval of the detail wavelet coefficients and the approximate wavelet coefficients, effectively removes the electromagnetic pulse artifacts and retains the effective EEG signal. The method can efficiently and stably remove the interference of the TMS electromagnetic pulse on the electroencephalogram signal without complicated calculation and analysis; in addition, compared with the conventional ICA multi-channel extraction method, the method is not limited by the number of channels and the experimental times, and can directly process single-channel single TMS-stimulated electroencephalogram signals; the electroencephalogram signal waveform extracted by the method is cleaner, and the method has important significance for analyzing, processing and researching the fused TMS and electroencephalogram.

Drawings

FIG. 1 is a block diagram of an embodiment of the present invention for removing TMS pulse artifacts;

FIG. 2 is an electroencephalogram signal of a frontal lobe cortex anterior-posterior channel FC2 stimulated by a single-pulse TMS in the embodiment;

FIG. 3 is a comparison graph of detail wavelet coefficients and approximation wavelet coefficients before (left column) and after (right column) the interval median filtering process in an embodiment;

FIG. 4 is a comparison graph of brain electrical signals before and after TMS pulse for removing FC2 channel in the embodiment of the invention.

Detailed Description

In order to facilitate better understanding and implementation of the present invention for those skilled in the art, the following detailed description is provided for the advantages of the present invention with reference to the accompanying drawings; the electroencephalogram data before and after single-channel single-pulse TMS stimulation are extracted and processed in the embodiment of the invention, and the method specifically comprises the following steps:

electroencephalogram signal acquisition and preprocessing unit

A. Firstly, acquiring electroencephalogram mixed signals before and after TMS (single pulse TMS) stimulation on a brain by an electroencephalogram test system and a TMS stimulation system (in the embodiment, a 64-channel electroencephalogram test system is adopted to record the electroencephalogram mixed signals before and after single pulse TMS stimulation); and (4) preprocessing the recorded electroencephalogram mixed signal for drift removal, power frequency interference removal and the like.

Wavelet modulus maximum processing unit:

B. extracting a single-channel electroencephalogram signal (in the embodiment, the electroencephalogram signal of a channel FC 2) from the preprocessed multi-channel electroencephalogram signal, performing multi-scale decomposition on the channel electroencephalogram mixed signal by using wavelet transform (in the embodiment, db3 wavelet is used for performing 3-layer wavelet decomposition), and extracting approximate wavelet coefficients and detail wavelet coefficients (such as cD1, cD2, cD2, cA3 and the like) under different scales;

C. after all extreme points of wavelets of each layer are automatically detected by using a wavelet modulus maximum linear search algorithm, the maximum point of the modulus (i.e. the pulse interference position) is located, meanwhile, the maximum point is used as the center to search the adjacent extreme points and record the pulse data interval of the wavelet coefficient corresponding to the extreme value, and the pulse data interval is used as the target data of median filtering processing (in the embodiment, the number of other extreme values near the maximum extreme value detected by the wavelet coefficient of each scale is between 3 and 5).

Interval median filtering unit

D. A median filter is constructed according to specific data characteristics of the target data interval, local interval median filtering is performed on the pulse data interval positioned in the approximate wavelet coefficient and the detail wavelet coefficient of each layer by using the median filter, and the detail wavelet coefficient after filtering is reconstructed into a new wavelet coefficient (the filtered cD1, cD2, cD3 and cA3 are used for reconstructing subsequent electroencephalogram signals in the embodiment).

Wavelet reconstruction unit

E. And finally, filtering and reconstructing the processed new wavelet coefficient into a new electroencephalogram signal by using a wavelet reconstruction function, wherein the obtained signal is the electroencephalogram signal without the TMS pulse artifact (in the embodiment, the db3 wavelet is used for carrying out wavelet decomposition). In addition, the multichannel brain electrical signals (channel number data points, namely 60 data points 1000 brain electrical data) without TMS stimulation pulses can be obtained by only repeating the above process to process the multiple channels.

In order to further explain the beneficial effects of the invention, firstly, the processed brain electrical signal of the invention is compared with the original brain electrical signal after TMS stimulation, the embodiment is exemplified by the brain electrical signal of FC2 channel, as can be seen from fig. 2, TMS pulse stimulation has a large influence on the original brain electrical signal of FC2 channel; as shown in figure 3, the influence of TMS pulse signals on wavelet coefficients of each layer can be effectively filtered by the wavelet decomposition and interval median filtering unit in the device, and then relatively stable wavelet coefficients are obtained to reconstruct new electroencephalogram signals. The comparison between the 1 st line and the 3 rd line of fig. 4 shows that the device of the present invention can effectively filter the TMS electromagnetic pulse artifact, and the reconstructed new brain electrical signal contains almost no pulse information. In addition, in order to further clarify the advantages of the device of the present invention, we compare the effect of the device of the present invention in filtering the TMS pulse signal with that of the conventional ICA method, as shown in fig. 4, line 2, it can be seen that the reconstructed electroencephalogram signal after the pulse signal is filtered by the ICA method still has a small amount of pulse information, and as shown in fig. 4, line 3, the device of the present invention can not only effectively filter the TMS pulse signal, but also the reconstructed electroencephalogram signal after the pulse signal is removed is more smooth and stable, and especially more electroencephalogram signals are retained in the data interval near the pulse stimulation point. Besides the higher efficiency and accuracy, the device has the other outstanding advantage that the device can well process the electroencephalogram signals after single-channel single-pulse TMS stimulation, further breaks through the limitation of the ICA method by the number of channels, and plays an important role in promoting the fusion research of TMS and EEG.

It should be emphasized that the above description of preferred embodiments is given by way of illustration and not by way of limitation, and that those skilled in the art, in light of the teaching of this disclosure, may make alterations and modifications without departing from the scope of the invention as defined by the appended claims.

Claims (2)

1. A TMS pulse artifact removing device based on interval median filtering comprises: the electroencephalogram signal acquisition and preprocessing unit acquires electroencephalogram signals, preprocesses the acquired electroencephalogram signals, and then sequentially passes through the wavelet modulus maximum processing unit, the interval median filtering unit and the wavelet reconstruction unit to obtain electroencephalogram signals without TMS pulse artifacts;

the electroencephalogram signal acquisition and preprocessing unit acquires electroencephalogram signals before and after TMS stimulation, and carries out drift removal and power frequency interference removal processing on the electroencephalogram signals to obtain electroencephalogram mixed signals;

the processing method of the wavelet modulus maximum processing unit comprises the following steps:

step 1: extracting an electroencephalogram mixed signal of one channel, and performing multi-scale wavelet decomposition on the electroencephalogram mixed signal by utilizing wavelet transformation to obtain an approximate wavelet coefficient cD and a detail wavelet coefficient cA under different scales;

step 2: detecting extreme points of wavelet coefficients under all scales by using a wavelet modulus maximum linear search algorithm, screening out a maximum value of a modulus from the extreme points, simultaneously retrieving an adjacent extreme point by taking a time point corresponding to the maximum value as a center, recording a data interval of the wavelet coefficients corresponding to the part of the extreme points, and taking the interval as a target data interval of interval median filtering processing, wherein the number of the adjacent extreme points is determined according to the duration time of a TMS pulse, and the single-pulse TMS can generally take 4 to 8 extreme points;

the method for searching the wavelet coefficient extreme point under each scale by the wavelet mode maximum linear search algorithm comprises the following steps:

S₁search time interval [ t ] as on decomposition scale J₀,t₁]The amplitude of the wavelet coefficient at a time t is arbitrarily selected as an initial maximum value | W_maxL, |; meanwhile, setting a modulus maximum threshold T of the wavelet coefficient under the scale according to specific requirements;

S₂from t₀Starting at time t ═ t₀(ii) a Compare | W (t) | with | W_maxThe magnitude of the value, | W (t) | > | W_maxIf W (T) is greater than T, let W_maxW (t), and recording the corresponding time t; if W (t) is less than or equal to W_maxIf w ≦ T, then the comparison is carried out until the next time, i.e. T ≦ T₀+ 1; wherein | represents a modulo operation;

S₃repeating the above process until finding the search time interval [ t ]₀,t₁]Maximum of the inner wavelet coefficient modulus; step S may also be repeated for different scales₁And S₂The maximum value sequence of the wavelet coefficient modulus under different scales can be obtained.

The processing method in the interval median filtering unit comprises the following steps:

constructing an interval median filter according to the target data obtained by the wavelet modulus maximum processing unit, performing interval median filtering on pulse data intervals in the detail wavelet coefficients under each scale by using the interval median filter, and performing data reconstruction on the detail wavelet coefficients subjected to the interval median filtering to obtain new wavelet coefficients;

the processing method in the wavelet reconstruction unit comprises the following steps:

and reconstructing the new wavelet coefficient obtained by the interval median filtering unit into a new electroencephalogram signal by using a wavelet reconstruction function, namely the new electroencephalogram signal for filtering the TMS pulse artifact.

2. The device for removing the TMS pulse artifacts based on interval median filtering of claim 1, wherein the interval median filter is constructed and filtered by the following method:

for the target data interval [ t₀,t₁]If the time series of detail wavelet coefficients is x₁,x₂，...，x_nTaking the window length m, m being an odd number, and extracting m numbers, e.g. x, from the input sequence_i-v，…,x_i-1，…x_i，…x_i+1，…x_i+v，…x_i-vWherein i is the central position of the window, satisfies

Sequencing the m sequence values from small to large according to the numerical values, wherein the sequence after sequencing is as follows: y is_i-v，…,y_i-1，…,y₁，…,y_i+1，…,y_i+vThen, the median filter takes the value corresponding to the middle sequence number as the output of the interval median filter, and is expressed as: y is_i＝Med{y_i-v,…,y_i,…,y_i+vIn which i ∈ Z,

med {. is } represents taking the median operation.

Images