Brain network construction method based on task-related electroencephalogram activation level, and display device

Abstract

The present invention relates to the technical field of electroencephalogram, in particular to a brain network construction method based on a task-related electroencephalogram activation level, and a display device. The brain network construction method comprises: acquiring a verification matrix on the basis of task-related electroencephalogram activation levels of channels; converting the verification matrix into a spatial matrix conforming to the physical distribution of the channels; and acquiring an edge connection matrix for associating the channels, i.e., in the spatial matrix, weighting association coefficients and physical distances of the channels. Without the application of electric stimulation, the present invention obtains the degrees of strength of connections between brain tissue nodes by means of examining the task-related electroencephalogram activation level of brain tissue to be subjected to function testing throughout an entire task time course, and combines the degrees with the physical locations of the nodes, so as to implement comprehensive and objective visualization of distribution and association conditions of functional areas of the brain, thereby being conducive to assisting a doctor in accurately judging the relationship between a lesion and functional areas of the brain.

Description

Brain network construction method and display device based on task-related EEG activation Technical Field

[0001] The present invention relates to the field of electroencephalogram (EEG) technology, and in particular to a method and a display device for constructing a brain network based on task-related EEG activation. Background Art

[0002] In the medical field, brain networks characterize the functions of various brain tissues, helping doctors accurately locate the relationship between tumors, epileptic lesions, and other brain functional areas, allowing for surgical resection. Existing technologies typically use electrodes placed on the brain tissue to capture the EEG signals generated by the brain tissue during different tasks. This analysis then allows for detailed localization of brain tissue function and the construction of brain networks.

[0003] Among them, the most widely used method for locating brain functional areas is ESM (Electric Stimulation), the industry's gold standard. ESM places positive and negative electrodes on brain tissue and applies current between them, causing that part of the brain tissue to temporarily lose function to simulate the effect of its removal. Electrical stimulation can only be applied to one pair of electrodes at a time, and each stimulation requires the patient to repeat the task. Therefore, it is difficult to have sufficient time to select all electrodes for electrical stimulation in sequence during surgery, and only negative or positive results can be recorded for some electrodes. As a result, the distribution and correlation of brain functional areas obtained are incomplete, and the brain network cannot be fully displayed. Summary of the Invention

[0004] This embodiment provides a brain network construction method and display device based on task-related EEG activation to comprehensively and completely display the distribution and correlation of brain functional areas.

[0005] In order to solve the above technical problems, in the first aspect, the present invention provides a method for constructing a brain network based on task-related EEG activation, comprising: obtaining a test matrix based on the task-related EEG activation of each channel; converting the test matrix into a spatial matrix that conforms to the physical distribution of the channel; obtaining an edge matrix for associating channels, that is, weighting the correlation coefficient of the channel with the physical distance in the spatial matrix; wherein, obtaining the test matrix based on the task-related EEG activation of each channel comprises: dividing the EEG signal segments, that is, dividing the EEG signal segments corresponding to the task moments into baseline segments and task segments according to the task start node before and after; obtaining data features, that is, calculating the data features of the baseline segment and the task segment respectively; testing the difference in the distribution of the data features of the task segment and the baseline segment to form a test matrix whose dimensions respectively characterize the channel and the task time. The spatial matrix is ​​obtained in a manner comprising: converting the test matrix into an n-order tensor, the representation of each dimension of which includes: channel row, channel column, task time, and channel group; averaging the dimension of the task time to perform dimensionality reduction to obtain a spatial matrix.

[0006] Furthermore, the element positions of the channel row dimension and the channel column dimension in the n-order tensor conform to the physical distribution of the channels; and the spatial matrix is ​​an n-1 dimensional matrix whose dimensions represent channel rows, channel columns, and channel groups, respectively.

[0007] Furthermore, the spatial matrix is ​​updated through a convolution operation, that is, each element in the spatial matrix is ​​convolved with surrounding elements to obtain the spatial matrix.

[0008] Furthermore, obtaining an edge matrix for associating channels includes: setting a weighted equation, that is, a weighted result = a*x+b*y, where x is the physical distance between channels, y is the correlation coefficient between channels, a is the distance weighting coefficient, and b is the correlation weighting coefficient; setting the correlation coefficient to any one of the Pearson correlation coefficient, mutual information, cosine similarity, and degree of same change between channels.

[0009] In a second aspect, the present invention provides a brain network display method, comprising: constructing a brain network using the brain network construction method; and displaying the edge morphology of related channels.

[0010] Furthermore, displaying the edge shape of the relevant channels includes: normalizing the element values ​​in the edge matrix; setting a representation relationship between the display effect of the edge shape and the element values ​​after normalization; and mapping the element values ​​in the edge matrix to the edge shape of the relevant channels according to the representation relationship to form a brain network image.

[0011] Furthermore, the characterization relationship includes: when the normalized element value is less than the set blank threshold, the edge shape is characterized as blank, otherwise, the edge shape is characterized as a line according to the edge equation; wherein, the edge equation is: edge shape characterization effect = element value * shape coefficient, the shape coefficient includes: color coefficient, line width coefficient, line type coefficient, and transparency coefficient.

[0012] In a third aspect, the present invention provides a brain network display device, comprising: a processor and a display; wherein the processor is used to execute the steps of a brain network display method based on task-related EEG activation; and the display is used to display the edge morphology of the relevant channels, that is, a brain network image.

[0013] In a fourth aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a brain network construction method or a brain network display method when executing the computer program.

[0014] In a fifth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a brain network construction method or a brain network display method.

[0015] Beneficial effects of the present invention:

[0016] The present invention does not require the application of electrical stimulation. By examining the task-related EEG activation of the brain tissue of the function to be tested throughout the entire task process, the strength of the connection between each brain tissue node is obtained and combined with the physical position of each node, a comprehensive and objective visualization of the distribution and correlation of brain functional areas is achieved, which helps assist doctors in accurately determining the relationship between lesions and brain functional areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will be further described below with reference to the accompanying drawings and examples.

[0018] FIG1 is a flowchart of a brain network construction method provided by some embodiments.

[0019] FIG2 is a flow chart of an algorithm for a check matrix provided by some embodiments.

[0020] FIG3 is a schematic diagram of an algorithm of a check matrix provided by some embodiments.

[0021] FIG4 is a data diagram of electrode pair stimulation results of Comparative Example 1 provided in some embodiments.

[0022] FIG5 is a schematic diagram of the brain network of Comparative Example 1 provided by some embodiments.

[0023] FIG6 is a schematic diagram of a brain network provided by some embodiments.

[0024] FIG7 is a schematic diagram of a brain network provided by some embodiments.

[0025] FIG8 is a schematic diagram of a brain network at stage 1 provided by some embodiments.

[0026] FIG9 is a schematic diagram of a brain network in stage 2 provided by some embodiments.

[0027] FIG10 is a schematic diagram of a brain network at stage three provided by some embodiments.

[0028] FIG11 is a schematic diagram of a brain network at stage four provided by some embodiments. DETAILED DESCRIPTION

[0029] To make the above-mentioned objects, features, and advantages of the present invention more readily apparent, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. The following description sets forth numerous specific details to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways than those described herein, and those skilled in the art may make similar modifications without departing from the scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0030] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. In the description of the present invention, the meaning of "plurality" is at least two, for example, two, three, etc., unless otherwise specifically defined. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0031] As shown in Figure 1, at least one embodiment provides a brain network construction method based on task-related EEG activation, including the following steps: step S1, obtaining a test matrix based on the task-related EEG activation of each channel; step S2, converting the test matrix into a spatial matrix that conforms to the physical distribution of the channels; step S3, obtaining an edge matrix for associating channels, that is, weighting the correlation coefficient of the channel and the physical distance in the spatial matrix.

[0032] It should be noted that in some embodiments, the physical distribution of channels is the relative row and column positions of the channels included in all the set acquisition electrode nodes, which also corresponds to the node positions in the brain network image. The number of rows of electrode channels is the number of channel rows, and the number of columns is the number of channel columns. The total number of channels = number of channel rows × number of channel columns.

[0033] It should be noted that the test includes multiple tasks. The node at which each task begins is defined as the task moment. Fixed or variable-length intervals are set between tasks (generally used between different types of tasks). Then, based on the time before and after the task start node, the EEG signal segments corresponding to each task moment are divided into baseline segments and task segments. That is, the interval before the task moment in each signal segment is the baseline segment, and the portion activated by the task after each task moment is the task segment. The signal segments collected for each task are the baseline segment before the task moment and the task segment after the task moment. Among them, the signal segment is a two-dimensional matrix with the number of rows being the number of channels and the number of columns being the total number of sample points. The baseline segment is a two-dimensional matrix with the number of rows being the total number of channels and the number of columns being the number of baseline sample points. The task segment is a two-dimensional matrix with the number of rows being the total number of channels and the number of columns being the number of task sample points.

[0034] As some optional implementations of obtaining the test matrix.

[0035] It should be noted that step S1, which obtains a test matrix based on the task-related EEG activation of each channel, is equivalent to a time representation and specifically includes the following steps:

[0036] Step S11, dividing the EEG signal segments, that is, dividing the EEG signal segments corresponding to the task time into baseline segments and task segments according to the task start node before and after. Then, in some embodiments, overlapping sliding window slicing is used to assist data feature extraction and activation degree testing to ensure real-time performance while being more accurate, that is, sliding window slicing is performed on the baseline segment to obtain a baseline window, and sliding window slicing is performed on the task segment to obtain a task window. Optionally, referring to Figure 3, the baseline window is a two-dimensional matrix, the number of its rows is the total number of channels, the number of columns = window length × sampling rate = number of baseline sliding windows, the task window is a two-dimensional matrix, the number of its rows is the total number of channels, the number of columns = window length × sampling rate = number of task sliding windows, the number of rows and columns is not limited, and it is sufficient to ensure that the number of rows and columns of the above-mentioned baseline window and task window correspond to each other according to actual application; the number of baseline sample points = number of baseline sliding windows × window length × sampling rate, the number of task sample points = number of task sliding windows × window length × sampling rate, the total number of sample points = number of baseline sample points + number of task sample points, and usually the window lengths of the baseline window and the task window are equal.

[0037] Step S12, respectively calculates the data features of the baseline segment and the task segment. In this embodiment 1, the data features of the baseline window and the task window are calculated. The method of extracting data features includes: filtering to remove the signal power frequency and frequency multiplication to obtain a filtered signal; extracting features to perform FFT transformation on the filtered signal and take the absolute value to obtain an envelope signal; and log transformation to perform log transformation on the data of the set response frequency band of the envelope signal to obtain data features. In some embodiments, the High Gamma value that can clearly reflect the significance of EEG activation can be selected as the data feature, and differential entropy, HFD, amplitude perception permutation entropy, LZ complexity, approximate entropy, Lie index, coherent imaginary part, weighted PLI, relative PSD, absolute PSD, IAF, IAF-prox, SASI, Asym, alpha power variability, etc. can also be used as data features.

[0038] Step S13 examines the difference in the distribution of data features between the task segment and the baseline segment to form a test matrix. To increase the credibility of the data, this embodiment incorporates historical data information while taking into account real-time data for correction. Specifically, the difference in the distribution of data features between the current and historical task windows and the data features of the current and historical baseline windows is calculated to form a test matrix corresponding to each signal segment. This difference is calculated using statistical testing methods (including any one or a combination of T-tests, KS tests, KL divergence tests, machine learning model tests, etc.). Because the baseline segment (baseline window) represents a resting state without a task, by examining the difference between the data features within each task window and the baseline window, it is possible to characterize whether changes in the data features within that task window are due to task stimulation. Specifically, the test matrix is ​​a two-dimensional matrix whose dimensions represent the channel and task time respectively; for example, as shown in Figures 2 and 3, the signal of each baseline window includes data features of the total number of channels, and the signal of each task window includes data features of the total number of channels. The T test is used to determine the difference in the distribution of the data features of the task window and the data features of the baseline window to form a test matrix, and the number of rows of the obtained test matrix is ​​the total number of channels, and the number of columns is the number of task sliding windows.

[0039] As some optional implementation methods for obtaining the space matrix.

[0040] It should be noted that step S2 converts the test matrix into a spatial matrix that conforms to the physical distribution of the channel, which is equivalent to a spatial representation. The specific acquisition method is as follows: Step S21, first convert the test matrix into an n-order tensor of each dimension that can be used to represent the channel row, channel column, task time, and channel group. The element positions of the channel row dimension and the channel column dimension in the n-order tensor are arranged according to the actual physical distribution of the electrode channel. The task time dimension is expressed as the number of task sliding windows in some embodiments, and the channel group dimension represents the association between each channel and the channel group in the embodiment of the present invention.

[0041] Specifically, the calculation method of the channel group dimension includes: classifying the channel groups according to channel position, brain functional area or custom method, and marking each channel group with a serial number, and then marking the serial number of the channel group to which each channel belongs as the element of the corresponding channel group dimension. In practical applications, for the convenience of acquisition, a large electrode sheet containing multiple channels is usually placed on brain tissue in different areas, that is, the channel groups are classified according to the channel position, and each electrode sheet is the channel group; or, the channels that have been verified to belong to the same brain functional area are defined as abstract channel groups, and the connection between nodes in each brain functional tissue can be further targeted by calculating the connection matrix. It can be seen that the channel group dimension contains the grouping information of the channels, such as the number of channels contained in each electrode sheet or channel combination, the channel position, the distance between channels, the distance between electrode sheets, the distance between channel combinations, the functional area to which the channels belong, etc., to summarize the relationship between channel groups, which is used to ensure that the connection matrix can also be constructed in scenarios containing multiple channel groups.

[0042] Specifically, step S22 is to perform dimensionality reduction by taking the average of the n-order tensor in the dimension of task time, and using the average of the corresponding position as the matrix element to obtain an n-1-dimensional spatial matrix representing the channel rows, channel columns, and channel groups in each dimension. For example, if there is only one channel group, the test matrix is ​​converted into a third-order tensor with the number of rows being the number of channel rows, the number of columns being the number of channel columns, and the depth being the number of task sliding windows. The task sliding window dimension is then averaged for dimensionality reduction, and the average of the corresponding position is used as the matrix element to obtain a two-dimensional spatial matrix representing the channel rows and channel columns in each dimension. If there are multiple channel groups, the test matrix is ​​converted into a fourth-order tensor with the dimension of the number of channel rows × the number of channel columns × the number of channel groups × the number of task sliding windows. The task sliding window dimension is then averaged for dimensionality reduction, and the average of the corresponding position is used as the matrix element to obtain a three-dimensional spatial matrix representing the channel rows, channel columns, and channel groups in each dimension. Furthermore, the different number of channel rows and columns in each channel group does not affect the calculation of the matrix. It is only necessary to fill the gaps with zeros or not use this part of the storage space. For example, in actual applications, using three electrode sheets with channel specifications of [1*4], [2*2], and [8*8] (i.e., three channel groups) will result in a spatial matrix of dimension 8*8*3. Rearranging the matrix through n-order tensors helps to simply realize the compressed extraction of information throughout the task time, so that an element in the spatial matrix contains information on the stimulation level of the entire time course, ensuring the temporal continuity of the brain network.

[0043] As an optional implementation method for obtaining the edge matrix.

[0044] It should be noted that step S3, which obtains the edge matrix for associating channels, includes setting a weighted equation: weighted result = a*x + b*y, where x is the physical distance between channels, y is the channel correlation coefficient, a is the distance weighting coefficient, and b is the correlation weighting coefficient. The coefficients a and b are set based on the expected rules for constructing brain networks. For example, the physical distance between channels of high-density electrodes is small, so in order to accurately distinguish channel calculations, the corresponding value of a is larger. The correlation between channels of deep electrodes is more important, so the corresponding value of b is larger.

[0045] Specifically, the correlation coefficient is set to any one of the Pearson correlation coefficient, mutual information, cosine similarity, and degree of same change between related channels. The number of rows in the edge matrix is ​​the number of channel rows, and the number of columns is the number of channel columns. The edge matrix is ​​a symmetric square matrix, and the elements in the i-th row and j-th column are equal in size and meaning to the elements in the j-th row and i-th column, that is, the correlation coefficient between the i-th channel and the j-th channel. The construction of the edge matrix represents the weighted weighting of the actual physical distance between electrodes and the correlation coefficient of the full-time activation between the two electrodes to control the sparsity of the brain network. The test matrix obtained by time representation is integrated into the spatial matrix obtained by spatial representation, so that the information of all electrodes involved in the test is connected to construct a complete and objective brain network.

[0046] In some embodiments, the spatial matrix is ​​updated by a convolution operation. For example, a 3*3 (or 5*5, etc.) convolution kernel is used to convolve each element in the spatial matrix with the surrounding elements (elements within the portion covered by the convolution kernel), and all elements are updated to obtain the spatial matrix. Specifically, if the electrode is not located at the edge or corner, the convolution operation is performed with the eight electrodes surrounding it; if the electrode is located at the edge, the convolution operation is performed with the five surrounding electrodes; if the electrode is located at the four corners, the convolution operation is performed with the three surrounding electrodes.

[0047] At least one embodiment further provides a brain network display method, comprising: constructing a brain network using the brain network construction method; and displaying the edge morphology of the related channels.

[0048] It should be noted that displaying the edge patterns of the relevant channels (i.e., the brain network image) involves normalizing the element values ​​in the edge matrix. Specifically, the global maximum element of the matrix is ​​taken, and a normalization parameter is set to divide the value of each element in the edge matrix by the value of the maximum element to normalize the value. Next, a relationship is established between the display effect of the edge pattern and the normalized element values. Finally, based on this relationship, the element values ​​in the edge matrix are mapped to the edge patterns of the relevant channels to form the brain network image. Among them, the characterization relationship includes: when the normalized element value is less than the set blank threshold, the edge shape is characterized as blank, otherwise, the edge shape is represented as a line according to the edge equation, and the edge equation is: edge shape characterization effect = element value * shape coefficient, the shape coefficient includes: color coefficient, line width coefficient, line type coefficient, transparency coefficient, that is to say, the edge shape characterization effect can present different edge line colors, thicknesses, line types, transparency, etc. Usually, the darker the color and the thicker the line, the higher the activation degree of the node represented by the edge, the lighter the color and the thinner the line, the lower the activation degree of the node represented by the edge, or red, orange, yellow, blue, and white are used in sequence to represent the activation degree from high to low. For example, in some embodiments, RGB triples in Matlab software are used to plot the color of edge patterns. The first number in the triplet is always 1, and the third number is always 0. Only the second number is adjusted, calculated as 1-(element at the corresponding position in the edge matrix / normalization parameter), for example, blue is [0 0 1]. The blank threshold is calculated as follows: if the second number in the calculated RGB triplet is greater than the blank threshold, the RGB triplet for that edge is set to [1 1 1], i.e., white. The color, thickness, and line type of each edge on the brain network image are all related to the corresponding element in the edge matrix, enabling global spatial visualization of the connection strength between nodes in the brain network and real-time output of a temporally continuous brain network that reflects the activation levels of a large number of electrodes.

[0049] In particular, the node positions in the brain network image correspond one-to-one to the relative positions of the elements in the spatial matrix. That is, each element in the matrix corresponds to a channel on the electrode patch. For example, the information collected by the electrode patch channel in the 3rd row and 4th column corresponds to the element in the 3rd row and 4th column of the spatial matrix.

[0050] It should be noted that the term "related channels" refers to any two channels taken from all acquisition channels for calculation. In some embodiments, adjacent channels are selected to calculate sparse brain networks. The selection of adjacent channels includes: if the electrode channel is not located at the edge or corner of all channels, the four channel nodes above, below, left, and right of the electrode channel can be selected as adjacent channels; if the electrode channel is located at the edge of all channels, the three channel nodes surrounding it can be selected as adjacent channels; if the electrode channel is located at the four corners of all channels, the two channel nodes surrounding it can be selected as adjacent channels.

[0051] At least one embodiment also provides a brain network display device, comprising a processor and a display; wherein the processor is used to execute the steps of a brain network construction method based on task-related EEG activation; and the display is used to display the edge morphology of related channels, that is, a brain network image.

[0052] In some optional embodiments, the clinical application process may be: first, based on technical means such as magnetic resonance imaging and epileptic discharge location judgment, the possible location of the lesion is evaluated as the functional area to be detected, and then a craniotomy is performed to place electrodes on the brain tissue to be detected (including standard electrodes with larger specifications and intervals, high-density electrodes with smaller specifications and intervals, microelectrodes, etc.) or insert deep electrodes inside for signal acquisition. In particular, a piece of acquisition electrode may contain multiple channels, and an electrode piece containing multiple channels or a combination of electrode channels can be placed in multiple locations in one operation. Then, optionally, the subject is awakened during the operation to perform the task, or after the electrode is implanted, the subject is asked to perform the task in the treatment phase / experimental phase (the task may be to name a picture, repeat a word, etc., and usually each task is not completed independently by a single brain area, and it is likely to be associated with the visual area, auditory area, thinking area, language area, etc. in the brain functional area in chronological order). Finally, each electrode channel collects intracranial EEG signals in real time, converts them into digital-to-analog signals, and transmits them to a computer device for processing. The computer device stores a program that executes the method for constructing a brain network based on task-related EEG activation. By simultaneously collecting EEG signals from all electrodes during a task, rather than stimulating electrode pairs sequentially, a complete network containing all nodes can be constructed.

[0053] At least one embodiment provides a computer device comprising a memory and a processor, wherein the memory stores a computer program and the processor executes the steps of a brain network construction method or a brain network display method. The computer device may be, for example, but not limited to, a computer, a tablet, or a server.

[0054] At least one embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a brain network construction method or a brain network display method. The computer-readable storage medium may be, for example, but not limited to, a USB flash drive, a hard drive, a memory card, or the like.

[0055] The technical effects of this embodiment are described below through specific examples.

[0056] Comparative Example 1.

[0057] Stimulation results for 17 pairs of electrodes using the ESM method. Refer to the data shown in Figure 4, where stim1 is the first stimulation electrode, stim2 is the second stimulation electrode, and stim1 and stim2 form a pair of electrodes; a value of 0 is negative, indicating that after electrical stimulation, the subject lacking the brain network function corresponding to this electrode pair cannot complete the current task; a value of 1 is positive, indicating that after electrical stimulation, the subject lacking the brain network function corresponding to this electrode pair can complete the current task. Matlab simulation results in the brain network diagram of Figure 5, where the horizontal and vertical coordinates represent the location coordinates of the brain network.

[0058] Example 1.

[0059] Using electrode node data from the same patient and with the same coordinates as in Comparative Example 1, a brain network construction method was adopted, but the convolution operation on the spatial matrix was not performed to construct the brain network. The resulting brain network schematic diagram is shown in FIG6 . The horizontal and vertical coordinates in FIG6 are the position coordinates of the brain network. The figure shows more connections between brain network nodes than in Comparative Example 1, but there are more vacancies in the edges. Comparing the common parts of the two, the accuracy of Example 1 is 60% compared to Comparative Example 1.

[0060] Example 2.

[0061] Using the electrode node data of the same patient and the same coordinates as in Comparative Example 1, a brain network construction method was adopted, that is, a convolution operation was performed on the spatial matrix during the construction of the brain network. The resulting brain network schematic diagram is shown in Figure 7. The horizontal and vertical coordinates in Figure 7 are the position coordinates of the brain network. The accuracy of Example 2 compared with Comparative Example 1 is 100%, and Example 2 clearly depicts more connections between brain network nodes than Comparative Example 1 through convolution, better presents the complete network, and can also characterize the intermediate state between significant activation and inactivation compared to a single negative or positive result.

[0062] Example 3.

[0063] Using the same patient and electrode node data with the same coordinates as in Comparative Example 1, a brain network display method was used, with the normalization parameter set to 11 and the blank threshold set to 0.8. Figure 8 is a schematic diagram of the brain network of Example 3 in Stage 1, Figure 9 is a schematic diagram of the brain network of Example 3 in Stage 2, Figure 10 is a schematic diagram of the brain network of Example 3 in Stage 3, and Figure 11 is a schematic diagram of the brain network of Example 3 in Stage 4. The horizontal and vertical coordinates in Figures 8-11 are the position coordinates of the brain network. The accuracy of the results of Stage 4 relative to Comparative Example 1 was 86.67%. Since subjects are prone to fatigue and thus affect the results when repeating the same task for a long time, experiments on the same task can be conducted with dozens of trials as one stage, with the same number of trials in each stage and rest periods between stages. Although the degree of electrode activation varies, the data results obtained in each stage are superimposed and corrected on the results of the historical stage. The activation weights of areas in the brain network that are not important for a certain task are reduced, and the activation weights of important areas are increased. After multiple stages of experiments, accurate and stable brain network images can be obtained. Referring to Figures 8 to 11, it can be seen that the brain network diagrams of stages one to four tend to be stable, and there is no low accuracy or unstable switching between all positive and all negative stimulation results. At the same time, some low-importance edges are screened out. Comparing the common parts of the two, Example 3 is closer to the brain network results shown in Comparative Example 1, that is, implementing tasks in stages is conducive to assisting in the credibility of the judgment results.

[0064] In summary, this embodiment combines the test matrix obtained by time representation with the spatial matrix obtained by spatial representation, so that connections are established between electrode channels and between time points, and discrete results are integrated in units of electrode channels. The activation degree of all electrodes in the task time course can be recorded and calculated through only a single task, ensuring the continuity of the time dimension. At the same time, the strength of the connection between each node in the brain network is spatially visualized based on the task-related EEG activation degree, objectively and comprehensively representing the functional stimulation of brain tissue under each electrode throughout the task, which is closer to the actual doctor's cognition. To a certain extent, it avoids the misjudgment of the importance of a single channel, and thus guides doctors to avoid the resection of functional brain areas as much as possible during surgery. In addition, this embodiment sets electrodes on or inside all brain tissues to be tested, and analyzes and calculates based on the EEG signals fed back by the task, avoiding the side effects of traditional electrical stimulation methods such as after-discharge, disease attacks, and pain.

[0065] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0066] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be determined by the appended claims.

Claims

1. A method for constructing a brain network based on task-related EEG activation, characterized in that: include: Obtain a test matrix based on the task-related EEG activation of each channel; Converting the inspection matrix into a spatial matrix that conforms to the physical distribution of channels; An edge matrix for associating channels is obtained, that is, the association coefficients of the channels and the physical distances are weighted in the spatial matrix.

2. The brain network construction method according to claim 1, characterized in that The test matrix based on the task-related EEG activation of each channel includes: Divide the EEG signal segments, that is, divide the EEG signal segments corresponding to the task time into baseline segments and task segments according to the task start node; Obtain data features, that is, calculate the data features of the baseline segment and the task segment respectively; The difference in the distribution of data features between the test task segment and the baseline segment is formed to form a test matrix.

3. The brain network construction method according to claim 2, characterized in that: The test matrix is ​​a two-dimensional matrix whose dimensions represent the channel and the task time respectively.

4. The brain network construction method according to claim 3, characterized in that: The spatial matrix is ​​obtained by: Convert the test matrix into an n-order tensor, whose dimensions include channel rows, channel columns, task times, and channel groups; The dimensions of the task time are averaged and reduced to obtain a spatial matrix.

5. The brain network construction method according to claim 4, characterized in that: The element positions of the channel row dimension and the channel column dimension in the n-order tensor conform to the physical distribution of the channel; The spatial matrix is ​​an n-1 dimensional matrix whose dimensions represent channel rows, channel columns, and channel groups respectively.

6. The brain network construction method according to claim 4, characterized in that: The calculation of the channel group dimension in the n-th order tensor includes: Classify at least one channel group according to channel location and brain functional area and label them with serial numbers; The sequence number of the channel group to which each channel belongs is marked as the element of the corresponding channel group dimension.

7. The brain network construction method according to claim 1, characterized in that Obtaining the edge matrix used to associate channels includes: Set the weighted equation, that is, weighted result = a*x+b*y, where x is the physical distance between channels, y is the correlation coefficient between channels, a is the distance weighting coefficient, and b is the correlation weighting coefficient; The correlation coefficient is set to be any one of the Pearson correlation coefficient, mutual information, cosine similarity, and degree of same change between related channels.

8. The brain network construction method according to claim 1, characterized in that: The spatial matrix is ​​updated by a convolution operation, i.e. Convolve each element in the spatial matrix with the surrounding elements to obtain the spatial matrix.

9. A brain network display method, characterized in that: include: Constructing a brain network using the brain network construction method according to claim 1; Displays the edge shape of the related channels.

10. The brain network display method according to claim 9, characterized in that: The edge forms of the display related channels include: Normalizing the element values ​​in the edge matrix; Set the relationship between the display effect of the edge shape and the element value after normalization; According to the representation relationship, the element values ​​in the edge matrix are mapped to the edge forms of the relevant channels to form a brain network image.

11. The brain network display method according to claim 10, characterized in that: The representation relationship includes: When the normalized element value is less than the set blank threshold, the edge shape is characterized as blank, otherwise, the edge shape is characterized as a line according to the edge equation; The edge connection equation is: edge connection morphological characterization effect = element value * morphological coefficient, and the morphological coefficient includes: color coefficient, line width coefficient, line type coefficient, and transparency coefficient.

12. A brain network display device, characterized in that: include: processor and display; in The processor is configured to execute the steps of the brain network construction method according to claim 9; The display is used to display the edge connection form of the relevant channels.

13. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to claim 1 are implemented.

14. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to claim 9 are implemented.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to claim 1 are implemented.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to claim 9 are implemented.

Images