CN104083163A - Method for obtaining nonlinearity parameter electroencephalogram mapping - Google Patents

Abstract

The present invention provides a method for obtaining non-linear parametric EEG topographic map, which comprises the following steps: 1) For the multi-lead EEG synchronously collected on the scalp, it is coded by dynamic non-linear coding rules to form a multi-dimensional code sequence; 2) calculate the multidimensional information incremental entropy for the multidimensional code sequence obtained in step 1); 3) map the multidimensional information incremental entropy obtained in step 2) to the simulated head according to the electrode position collected by EEG Top view, and two-dimensional interpolation, and finally draw a non-linear parametric EEG topographic map. The non-linear parametric EEG topographic map drawn by the present invention can well visualize and reflect different activity states of the brain, and does not need preprocessing such as manual removal of artifacts on the collected EEG, and the effect is very good after the actual data test.

Description

A Method of Nonlinear Parametric EEG Topography

technical field

The invention relates to an EEG signal processing method and an EEG topographic map technology.

Background technique

Electroencephalogram (Electroencephalogram, EEG) is an important means for us to understand brain activity and state, and EEG mapping is an EEG diagnostic technique widely used in clinical practice. The EEG topography technology first extracts the characteristic parameters from the EEG, and then combines the placement position of the electrodes when collecting the EEG, and draws the characteristic parameters as a scalar function determined by the position on a two-dimensional plane, so that the time domain of the EEG Waveforms become topographical maps that can be localized and quantified simultaneously.

At present, the drawing of EEG topography mainly depends on the frequency domain parameters, that is, according to the different frequency components, the EEG is divided into δ (0.5Hz-4Hz), θ (4Hz-8Hz), α (8Hz-13Hz), β( 13Hz～30Hz) and other rhythms, the absolute energy or relative energy of the corresponding rhythm constitutes the characteristic parameters in the topographic map drawing.

Current EEG topography techniques have the following limitations:

First of all, frequency domain analysis is essentially a linear analysis method, and the analysis results are not robust to non-stationary data. However, EEG signals are rich in nonlinear components and are often non-stationary. Therefore, frequency domain parameters have limited ability to capture EEG features.

In addition, the EEG signal has a low signal-to-noise ratio and often contains a variety of artifacts. Before the existing diagnostic analysis (including frequency domain analysis), experienced doctors generally need to manually remove the artifacts, which greatly reduces the convenience of EEG diagnostic technology. sex and automaticity.

In view of the above limitations, if a topographic map method that reflects the nonlinear characteristics of EEG can be established, which can well resist the inherent non-stationarity of EEG and various interferences and artifacts, it will surely provide a more accurate diagnosis of EEG in clinical practice. , Fast diagnostic approach.

Contents of the invention

The purpose of the present invention is to provide a non-linear parameter EEG topographic map method, which can effectively reflect the brain activity state based on the nonlinear characteristics of EEG signals, and can effectively resist the non-stationary characteristics of EEG signals, and does not need to manually remove artifacts Wait for preprocessing.

The purpose of the present invention is achieved like this:

Based on the collected multi-conductor EEG signal, a multi-dimensional coding sequence is formed by dynamic nonlinear coding, and the non-linear characteristic parameter of the coding sequence - information incremental entropy is extracted, and then combined with the placement position of the electrode on the scalp when collecting EEG, The characteristic parameters are used as a scalar function determined by the position, and are drawn on the simulated top view of the head after two-dimensional interpolation to realize the non-linear parametric EEG topographic map. The core of the present invention includes two parts: dynamic nonlinear coding and nonlinear feature parameter extraction.

Further, the dynamic nonlinear encoding described in the present invention includes the following steps:

For the multi-lead EEG signal {x _i :1≤i≤N} collected non-invasively on the scalp, examine its probability distribution, and calculate the first-order difference sequence {Δx _i :1≤i≤N-1}, according to its probability distribution and the symbols of the first-order difference sequence are encoded to form a new binary coded sequence {s _i : 1≤i≤N-1}. Assuming that the word length of the binary code is m, the coding rules are described as follows:

First sort {x _i :1≤i≤N} and perform 2 ^m-1 equal parts to find 2 ^m-1 -1 equal parts, recorded as but

Then correct the last digit of {s _i } according to the differential sequence,

Thus, the nonlinear coding sequence {s _i } is obtained.

The encoding rule is based on probability encoding, which breaks through the linear limitation of amplitude encoding, and can greatly weaken the disturbance caused by the original large mutation interference in the amplitude domain. At the same time, by introducing differential sequence symbols, the high-frequency components of the original signal Coding rules are also introduced to effectively resist the influence of non-stationary trend items.

Further, the multidimensional information incremental entropy calculation described in the present invention includes the following steps:

Convert binary coded sequence {s _i } to decimal sequence {y _i };

Perform 2-dimensional and 3-dimensional delayed embedding on the sequence {y _i } respectively to obtain vector sequences {B ⁽²⁾ (i)}, {B ⁽³⁾ (i)};

Calculate the probability C ⁽²⁾ and C ⁽³⁾ of the same two vectors under the 2-dimensional delayed embedding and the 3-dimensional delayed embedding respectively;

Calculate the sequence information incremental entropy value

The information increment entropy reflects the one-step predictability or information increment of the encoding sequence under the encoding rules.

Further, the EEG topographic map drawing described in the present invention includes the following steps:

According to the placement position of the electrodes collected by the EEG, the multi-dimensional information incremental entropy obtained above is constructed into a two-dimensional function determined by the position of the electrodes;

Taking the position of the electrode as the node, the two-dimensional function interpolation is performed on the simulated top view of the head;

Finally, the interpolated function values are mapped to the color domain or grayscale domain to draw a non-linear EEG topographic map of the top view of the head.

In summary, the beneficial effect of the method of the present invention is that nonlinear dynamic coding and information incremental entropy calculation are the core, which realizes the interpretation of the essential characteristics of the EEG signal, and then combines the topographic map drawing technology to finally draw a map that reflects the state of brain activity. Visualization of nonlinear EEG topography.

The encoding rules and characteristic parameter calculations in the present invention have broken through the linear field, and at the same time, it can also resist the influence of non-stationary trend items and non-stationary mutation interference artifacts by means of encoding. Therefore, this method does not need to filter the original collected data Or artificial artifact removal and other preprocessing, thus greatly increasing the practicability and automation possibility of this method. In addition, since all calculation processes in the method only involve elementary mathematical operations, it is easy to implement.

Description of drawings

Fig. 1 is a functional block diagram of the non-linear EEG topography method of the present invention.

Fig. 2 is the non-linear parametric EEG topographic map of test data 1 (a healthy person doing motor imagery with the left hand (a) and the control state (b)).

Fig. 3 is a non-linear parametric EEG topographic map of test data 2 (a healthy person doing leg movement imagery (a) and a control state (b)).

Fig. 4 is a non-linear parametric EEG topographic map of test data 3 (a healthy person doing left-hand motor imagery (a) and right-hand motor imagery (b)).

Detailed ways

In order to better understand the technical content of the present invention, specific embodiments are given and described as follows in conjunction with the accompanying drawings.

Fig. 1 is a functional block diagram of a non-linear EEG topography method in the present invention.

A method for nonlinear EEG topography, the steps comprising:

1) For the collected multi-conductor EEG signals, a multi-dimensional coding sequence is formed through dynamic nonlinear coding;

2) The multi-dimensional encoding sequence obtained in step 1 extracts the nonlinear characteristic parameter---information incremental entropy;

3) The nonlinear characteristic parameters obtained in step 2 are used as a scalar function determined by the position, and two-dimensional interpolation is performed with the electrode position as the node, and finally drawn on the simulated two-dimensional plane of the top view of the head to realize the nonlinear parameter brain Electric topographic map. Each electrode has data to give the coding sequence.

In described step 1), dynamic non-linear encoding, taking 3-bit word length binary encoding as an example, specifically includes:

1.1) Arrange the original sequence {x _i } of each lead in ascending order (or descending order) to obtain the sequence {u _i :1≤i≤N}, and obtain the three quartile points in the sequence {u _i } in turn ( That is, 25%, 50%, and 75% quantile points), which are recorded as t ₁ , t ₂ , and t ₃ in ascending order; each lead is coded and calculated independently. In order to simplify the subsequent symbols, the subscripts of the leads are not marked, so "each lead" is used in the description here.

1.2) According to the formula (1), the first N-1 data of the original sequence of each lead is binary coded with a word length of 3 bits,

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 000</span>}<span\; class="notranslate">\; :</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ \mathrm{<span\; class="notranslate">\; 010</span>}<span\; class="notranslate">\; :</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ \mathrm{<span\; class="notranslate">\; 100</span>}<span\; class="notranslate">\; :</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ \mathrm{<span\; class="notranslate">\; 110</span>}<span\; class="notranslate">\; :</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\end{array}\right.<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

1.3) Obtain the incremental sequence {Δx _i :1≤i≤N-1} again, where

Δx _i =x _i+1 -x _i (2)

Correct the lowest bit encoded in step 1.2) with the sign of the incremental sequence, i.e.

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}<span\; class="notranslate">\; *</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>\\ <span\; class="notranslate">\; *</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>\end{array}\right.<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

Thus a 3-bit binary coded sequence {s _i :1≤i≤N-1} is formed.

Described step 2) in, the extraction of non-linear parameter---information increment entropy specifically comprises:

2.1) Convert the binary code sequence {s _i :1≤i≤N} into a decimal code sequence {y _j :1≤j≤N-1}

2.2) Perform 2-dimensional and 3-dimensional delayed embedding on the sequence {y _j } respectively to obtain a vector sequence:

B ⁽²⁾ (i)=(y _i ,y _i+1 ), 1≤i≤N-2 (4)

B ⁽³⁾ (i)=(y _i ,y _i+1 ,y _i+2 ), 1≤i≤N-3 (5)

2.3) Definition are the same number of vectors as B ⁽²⁾ (i) and B ⁽³⁾ (i), respectively, then the probability that the vectors under 2-dimensional and 3-dimensional delayed embedding are identical to each other are:

${\mathrm{<span\; class="notranslate">\; C</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\frac{{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; C</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 3</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 3</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\frac{{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$

2.4) Then the information incremental entropy value is

$\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; ln</span>}\frac{{\mathrm{<span\; class="notranslate">\; C</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; C</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

In the step 3), the information incremental entropy results of each lead calculated in the step 2 are used as the interpolation node of the two-dimensional function related to the electrode position, two-dimensional interpolation is performed, and then mapped to the color domain, and finally drawn A non-linear parametric EEG topographic map is obtained (using prior art).

Below, the method is applied to the actually collected EEG, and the present invention will be further described in conjunction with the accompanying drawings.

Referring to Figure 2, it is a nonlinear topographic map of EEG drawn by our method using the EEG collected from a healthy person. Among them, sub-figure (a) is the state of left-hand motor imagery, and sub-figure (b) is the control state without motor imagery. It can be seen from the figure that during motor imagery, while the incremental information entropy of the two sides increases, the incremental information entropy of the middle, top and occipital regions decreases.

Referring to Figure 3, it is a nonlinear EEG topographic map drawn by our method using the EEG collected from another healthy person. Among them, sub-figure (a) is the state of doing leg motor imagery, and sub-figure (b) is the control state without motor imagery. It can be seen from the figure that during motor imagery, the incremental information entropy of the occipital region is significantly lower than that of the control state.

Referring to Figure 4, it is the EEG collected from another healthy person, using our method to draw a non-linear EEG topographic map. Among them, sub-picture (a) is the state of imagining the movement of the left hand, and sub-picture (b) is the state of imagining the movement of the right hand. It can be seen from the figure that when imagining different hand movements, the relationship between the top and the occipital, the information increment entropy on the left and right sides is reversed.

The above graphical results show that the non-linear parametric EEG topographic map proposed by the present invention can effectively reflect different activity states of the brain.

In the present invention, the core lies in the calculation of dynamic non-linear coding and non-linear parameter---information incremental entropy. The nature of EEG signals is nonlinear, with considerable non-stationarity, and there are many artifacts and interferences at the same time. Therefore, to use EEG to reflect brain activity and state, it is necessary to adopt a non-linear method that can effectively resist non-stationarity. In the present invention, the dynamic nonlinear coding breaks through the linear constraint of the traditional amplitude coding, and the variable resolution can be realized by adopting the probability coding, which proves to be an effective nonlinear coding. In the present invention, the real-time dynamic changes of EEG are also incorporated into the coding rules, thereby retaining the high-frequency components of the original data. The inventors believe that the dynamic nonlinear coding in the present invention can well solve the problems of non-stationarity, artifacts and interference in the current EEG signal analysis, and realize the effective preservation of the essential characteristics of the brain power system. In the present invention, the nonlinear parameter——information increment entropy reflects the information increment when the system makes one-step prediction under the premise of the aforementioned encoding, which is the description of nonlinear characteristics, thus realizing the effective expression of nonlinear dynamic characteristics .

Since the calculations involved in this method do not require the assumption of stationarity, and do not require preprocessing such as removal of artificial artifacts, and at the same time, since all calculations in this method only involve elementary mathematical operations, they are easy to implement. Therefore, it is possible to It is very convenient to realize and facilitate the automatic processing of data.

Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Those skilled in the art of the present invention can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be defined by the claims.

Claims (5)

1. A method for obtaining nonlinear parameter EEG topography, characterized in that it comprises the following three steps: 1) For the multi-lead EEG collected synchronously on the scalp, encode it through dynamic nonlinear encoding rules to form a multi-dimensional encoding sequence; 2) to the multidimensional encoding sequence obtained in step 1), calculate the multidimensional information incremental entropy; 3) For the multi-dimensional information incremental entropy obtained in step 2), map the electrode position collected by EEG to the simulated top view of the head, and perform two-dimensional interpolation, and finally draw the nonlinear parameter EEG topographic map. 2. nonlinear parameter EEG topographic map method according to claim 1, is characterized in that, in described step 1), dynamic nonlinear coding comprises: For the original sequence {x _i :1≤i≤N} of each lead EEG, examine its probability distribution, and calculate the first-order difference sequence {Δx _i :1≤i≤N-1}, according to its probability distribution and The symbols of the first-order difference sequence are encoded to form a new encoding sequence {s _i :1≤i≤N-1}; assuming that the binary encoding word length is m, the encoding rules are described as follows: Firstly, the original sequence { _xi :1≤i≤N} of each lead EEG is sorted and divided into 2 ^m-1 equal parts, and 2 ^m-1 -1 equal parts are found, which are denoted as Then the new coding sequence {s _i } Then, after correcting the last digit of {s _i } according to the differential sequence, Thus the final nonlinear coding sequence {s _i } is obtained. 3. nonlinear parameter EEG topographic map method according to claim 1, is characterized in that, in described step 2), calculates multidimensional information increment entropy, comprises the following steps: 3.1) Convert the binary non-linear coding sequence {s _i } into the decimal sequence {y _i }; 3.2) Perform 2-dimensional and 3-dimensional delayed embedding on the sequence {y _i } respectively to obtain vector sequences {B ⁽²⁾ (i)} and {B ⁽³⁾ (i)}; 3.3) Calculate the probability C ⁽²⁾ and C ⁽³⁾ that the vectors are identical in pairs under the 2-dimensional delay embedding and the 3-dimensional delay embedding respectively; 3.4) Calculate information incremental entropy value 4. The non-linear parametric EEG topography method according to claim 1, characterized in that the information increment entropy reflects the one-step predictability or information increment of the coding sequence under the coding rules. 5. the non-linear parameter EEG topographic map method according to claim 1, is characterized in that, according to the electrode placement position that EEG collects, the multi-dimensional information increment entropy that aforesaid obtains is constructed into the two-dimensional information that is determined by the electrode position. dimension function; Taking the position of the electrode as the node, the two-dimensional function interpolation is performed on the simulated top view of the head; Finally, the interpolated function values are mapped to the color domain or grayscale domain to draw a non-linear EEG topographic map of the top view of the head.