Method and system for analyzing EEG signals of mild cognitive impairment based on multi-band PCMI-EBN

Abstract

The present invention relates to the technical field of neural network model construction, and in particular to a method for analyzing EEG signals of mild cognitive impairment based on multi-band PCMI-EBN. The method collects historical data samples of EEG signals, performs denoising processing, and then uses the sorting conditional mutual information method to extract coupling features to construct a causal brain network PCMI-EBN, and uses a multi-band kernel ELM method based on multi-band PCMI-EBN signals to achieve high-precision classification of different EEG signals. The beneficial effect of the present invention is that the method of the present invention helps to solve the problem of difficulty in early diagnosis of mild cognitive impairment, and can be widely used in health management guidance and clinical diagnosis and treatment in hospitals and communities. As a medical aid, clinical diagnosis reference and treatment auxiliary technology for mild cognitive impairment, the problem of difficulty in early diagnosis of mild cognitive impairment can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of neural network model building, and in particular to a method and system for analyzing EEG signals of mild cognitive impairment based on multi-band PCMI-EBN. Background Art

[0002] Dementia is the fifth leading cause of death worldwide, and Alzheimer's disease (AD) accounts for 60% to 80% of dementia cases worldwide. The main characteristics of AD are decreased memory and cognitive ability, which seriously affect the normal life of patients. This is a major challenge to the global health and social care systems. At present, a large proportion of Alzheimer's patients are not discovered in time in the early stages of the disease, making later treatment more difficult. Therefore, it is crucial for patients to diagnose AD as early as possible and intervene in time.

[0003] Mild cognitive impairment (MCI) is an early stage of AD. Among the elderly population, about 15% to 20% suffer from MCI, which is characterized by a decline in cognitive ability that is greater than expected for the individual's age, but does not seriously affect daily activities. Every year, 10-15% of MCI patients transform into AD. Compared with age-matched healthy individuals, MCI patients transform into AD significantly faster, with an increase of about 1-2%. Current medical technology can control the condition of MCI patients, and the cure rate is relatively high. Therefore, studying the early screening method of MCI is of great significance for the early intervention of AD.

[0004] At present, the methods for confirming whether a patient has AD or MCI mainly include neuropsychological examinations, such as the Simple Mental Scale Examination and the Daily Living Ability Assessment Scale Examination; neuroimaging examinations, such as magnetic resonance imaging, computed tomography, positron emission tomography electroencephalogram, etc.; hematological examinations; cerebrospinal fluid testing; genetic testing and other methods. Studies have shown that the MMSE examination commonly used in clinical psychology evaluation has a high rate of missed detection of MCI. Because clinical diagnosis is mostly based on the subjective judgment of doctors, there may be certain defects. Therefore, the use of other scientific and efficient non-clinical methods for confirmation can make the diagnosis more scientific and objective, and play a certain role in improving the accuracy of diagnosis. Most of the methods using neuroimaging techniques such as MRI or PET for diagnosis are expensive and highly invasive. In comparison, the low cost, non-invasiveness and portability of EEG technology make it a good choice. In addition, studies have shown that EEG can detect some brain function abnormalities that cannot be detected by other methods in advance, that is, it can show some abnormalities that have not reached the diagnostic standard stage of other methods, which is suitable for the prediction of MCI.

[0005] In the prediction research of AD or MCI based on EEG data, most researchers extract multiple features of EEG, process the extracted feature sets using methods such as feature selection, and then use machine learning methods to achieve classification. In terms of extracting EEG features, the permutation conditional mutual information (PCMI) method considers both the coupling strength and coupling direction of the neural network, and describes the relationship between the data more comprehensively. In different frequency bands, the coupling features extracted by the PCMI method can reflect the significant differences in the EEG signals of MCI and healthy people. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a method and system for analyzing EEG signals of mild cognitive impairment based on multi-band PCMI-EBN, so as to achieve high-accuracy classification of EEG signals and solve any of the above and potential problems in the prior art.

[0007] The technical solution adopted by the present invention is: a method for EEG analysis of mild cognitive impairment based on multi-band PCMI-EBN, the method comprising the following steps:

[0008] S1) collecting EEG signal historical data samples, performing denoising processing, and extracting EEG signal data in multiple frequency bands of each EEG signal historical data sample;

[0009] S2) constructing a PCMI-EBN of the causal effect brain network of the single-band EEG signal data for each EEG signal history data sample;

[0010] S3) performing sparse processing on the single-band PCMI-EBN obtained in S2) according to the set sparsity value, and then obtaining the kernel matrix of the single-band under the sparsity value;

[0011] S4) repeating S2)-S3) and then fusing to obtain the kernel matrix of multiple frequency bands of all EEG signal historical data samples;

[0012] S5) According to the kernel matrices of multiple frequency bands of all EEG signal historical data samples obtained in S4), a kernel ELM classifier of multiple frequency bands is obtained by fusing them, and the kernel matrices of multiple frequency bands of all EEG signal historical data samples obtained in S4) are input into the kernel ELM classifier of multiple frequency bands, and the classification of normal signals and abnormal signals is completed according to the output results, so as to realize EEG analysis of mild cognitive impairment.

[0013] Further, the specific steps of S1) are:

[0014] S1.1) removing the noise frequency bands below 0.5 Hz and above 32 Hz in the collected EEG signal historical data samples, and then using stationary wavelet transform and independent component analysis methods to perform denoising;

[0015] S1.2) using a segmented aggregation approximation method to compress the EEG signal historical data samples processed by S1.1);

[0016] S1.3) Use the short-time Fourier transform method to extract EEG signal data of multiple frequency bands for each EEG signal historical data sample in the compressed data.

[0017] Furthermore, the number of the multiple frequency bands is four, namely: δ: 1-3 Hz, θ: 4-7 Hz, α: 8-13 Hz and β: 14-30 Hz.

[0018] Further, the specific steps of S2) are:

[0019] S2.1) taking the sorted conditional mutual information PCMI between the channel where the EEG signal data in a single frequency band is located and all other remaining channels as the coupling feature, and constructing a brain network matrix according to the number of EEG signal channels, where the number of EEG signal channels is a, and a is a positive integer greater than 0;

[0020] S2.2) The constructed brain network matrix is ​​converted into a binary matrix by a threshold transformation method, and the binary matrix is ​​recorded as PCMI-EBN.

[0021] Further, the specific steps of S3) are:

[0022] S3.1) Setting the sparsity value,

[0023] S3.2) performing sparse processing on the PCMI-EBN of the single-band EEG signal data obtained in S2.2) according to the sparsity value set in S3.1) to obtain a single-band kernel matrix under the sparsity value.

[0024] Furthermore, the sparsity value in S3.1) is 12% to 40%, and the sampling step is 2%.

[0025] Further, the specific steps of S4) are:

[0026] S4.1) Repeat S2)-S3) to obtain the kernel matrix of a single frequency band under the sparsity value of multiple frequency bands of all EEG signal historical data samples;

[0027] S4.2) The following formula (1) is used to fuse the kernel matrix of a single frequency band in the kernel matrix of multiple frequency bands of all EEG signal historical data samples. The formula is as follows:

[0028]

[0029] In the formula, k i is the graph kernel when the sparsity value is i, λ i is the weight of the graph core when the sparsity value is i, and N is the number of graph cores with different sparsity.

[0030] Further, the specific steps of S5) are:

[0031] S5.1) Input the single-band kernel matrix of the multiple-band kernel matrices of all EEG signal historical data samples obtained in S4) into the ELM classifier at once, and obtain multiple single-band kernel ELM classifiers after training, and calculate g(x), the formula is as follows:

[0032]

[0033] Among them, g(x) is the output; k(x,x i ) is the value in the kernel matrix of x and the i-th training sample; k n×n is the training set graph kernel matrix; n is the number of samples in the training set; T is the sample category in the training set; Cr is the penalty factor, Cr = 10;

[0034] S5.2) Then, multiple single-band kernel ELM classifiers of each electrical signal historical data sample are fused, and the kernel matrix of multiple frequency bands of all EEG signal historical data samples obtained in S4) is input into the kernel ELM classifier of multiple frequency bands to obtain g(x):

[0035]

[0036] Where b is the number of frequency bands, ki is the kernel used in the i-th frequency band model, and λ' i is the weight and satisfies the following formula:

[0037]

[0038] If the value of g(x) is 1, it is a normal data sample. If the value of g(x) is -1, it is an abnormal data sample. The classification of normal and abnormal signals can be completed.

[0039] The present invention also provides a mild cognitive impairment EEG analysis system based on multi-band PCMI-EBN, the analysis system comprising: an EEG data input module, a preprocessing module, a PCMI-EBN display module and a classification result display module;

[0040] The EEG data input module is used to collect EEG signal history data and input the collected history data;

[0041] The preprocessing module is used to perform denoising on the input historical data, extract brain wave signals within a certain frequency band, and send them to the PCMI-EBN display module;

[0042] The PCMI-EBN display module is used to construct a PCMI-EBN brain network with the EEG data of each frequency band and calculate the kernel matrix of each sample under multiple sparsity;

[0043] The classification result display module is used to classify the kernel matrix obtained by the PCMI-EBN display module and confirm normal and abnormal signals according to the classification results.

[0044] A readable storage medium includes a memory storing a program and a processor, wherein the processor executes the above-mentioned method for analyzing EEG signals of mild cognitive impairment based on a multi-band PCMI-EBN kernel.

[0045] The beneficial effects of the present invention are as follows: due to the adoption of the above technical solution, the present invention uses the sorting conditional mutual information method to extract coupling features to construct a causal brain network PCMI-EBN, and uses the kernel ELM method based on multi-band PCMI-EBN signals to achieve high-precision classification of different EEG signals. The method of the present invention helps to solve the problem of difficulty in early diagnosis of cognitive impairment, and can be widely used in health management guidance and clinical diagnosis and treatment in hospitals and communities. As a medical aid, clinical diagnosis reference and treatment auxiliary technology for mild cognitive impairment, it solves the problem of difficulty in early diagnosis of mild cognitive impairment. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The present invention is a flowchart of the method for analyzing EEG signals of mild cognitive impairment based on multi-band PCMI-EBN.

[0047] Figure 2 The PCMI-EBN kernel method flow chart of the present invention is shown as follows:

[0048] Figure 3 It is a block diagram of the mild cognitive impairment EEG signal analysis system based on multi-band PCMI-EBN of the present invention. DETAILED DESCRIPTION

[0049] The following is a detailed description of the method and application of EEG signal analysis for mild cognitive impairment based on multi-band PCMI-EBN of the present invention in conjunction with the embodiments and drawings.

[0050] The present invention uses the sorting conditional mutual information method to extract coupling features to construct a causal effect brain network PCMI-EBN, and uses the kernel ELM method based on multi-band PCMI-EBN signals to achieve high-precision classification of different EEG signals.

[0051] like Figure 1 As shown, the present invention is based on the multi-band PCMI-EBN mild cognitive impairment EEG signal analysis method, the specific steps of the method are:

[0052] 1) Collect historical EEG signal data, preprocess it and extract EEG signals in characteristic frequency bands;

[0053] 1.1) Remove the EEG data with noise frequency bands below 0.5 Hz and above 32 Hz from the collected EEG signal history, and then use stationary wavelet transform and independent component analysis methods to denoise;

[0054] 1.2) Use the segmented aggregation approximation method to compress the historical EEG signal data processed by S1.1).

[0055] 1.3) Use the short-time Fourier transform method to extract EEG signals in four frequency bands: delta (1-4Hz), theta (4-8Hz), alpha (8-13Hz) and beta (13-30Hz) from the compressed data.

[0056] 2) Constructing PCMI-EBN, a causal brain network based on sorting conditional mutual information, according to the EEG signals of the extracted characteristic frequency bands;

[0057] 2.1) For each frequency band, the PCMI between the pairwise EEG signals is calculated as the coupling feature;

[0058] The model for calculating the PCMI value between the X channel and the Y channel, that is, between the X node and the Y node in the brain network is as follows:

[0059]

[0060] Where δ is the observed value derived from the δ-step delayed state of Y, and δ = 3-15, PCMI X→Y represents the functional connection strength between the X node and the Y node. If the number of EEG signal channels is N, the constructed brain network is a matrix of size N×N, where the element in the i-th row and j-th column is PCMI X→Y ;

[0061] 2.2) The constructed brain network matrix of size N×N is converted into a binary matrix through threshold transformation, and the binary matrix is ​​recorded as PCMI-EBN. The transformation model of the binary matrix is ​​as follows:

[0062]

[0063] Among them, th is the threshold, which is determined by the sparsity using the relative threshold method, G is a binary matrix, representing the connection structure of the brain network, G ij= 0 means that the connection from the i-th brain region to the j-th brain region is not retained, G ij =1 means to keep the connection.

[0064] 3) In order to eliminate the influence of sparsity on the final classification result, after calculating the graph kernel under multiple network sparsities, the graph kernels under multiple sparsities are fused to obtain the graph kernel matrix;

[0065] 3.1) The obtained PCMI-EBN is processed at multiple sparsities, and the sparsity is defined as follows:

[0066]

[0067] Considering the integrity of the network structure at the physiological significance level, the present invention performs sampling within the sparsity range of 12% to 40%, with a sampling step of 2%, to obtain brain networks at various sparsities. Therefore, for n EEG signal samples, at a fixed sparsity, such as 12%, n brain network graphs can be obtained;

[0068] 3.2) The graph kernel can be used to measure the similarity between two graphs. Under multiple network sparsities, the graph kernel between the brain networks of every two samples is calculated.

[0069] Take graph G and graph H as examples to calculate the graph core between them:

[0070] (i) For graph G and graph H, construct a set of sub-network groups respectively:

[0071]

[0072] Where V i j It is composed of the i-th node and all nodes whose shortest distance to the i-th node is less than j. is the corresponding edge, n is the total number of nodes, and h is a preset parameter, usually 1 to 4. The same logic applies.

[0073] (ii) Sub-network Group and Calculate the symmetric semi-positive definite matrix C separately:

[0074] Let x 0 =e n×1 ,

[0075] Let μ = e l×1 , Where A is the adjacency matrix of the graph and e is an all-one vector.

[0076] (iii) Compute the graph kernel between graphs G and H:

[0077]

[0078] According to the above process, we can obtain a graph kernel of size n×n at each sparsity.

[0079] 3.3) In order to eliminate the influence of sparsity on the final classification result, the graph kernels under multiple sparsities are fused to obtain the graph kernel matrix;

[0080] The model that integrates different sparsity kernels is as follows:

[0081]

[0082] where k i is the graph kernel when the sparsity is i, λ i is the weight, and N is the number of selected graph kernels with different sparsity.

[0083] 4) The kernel matrix calculated on multiple frequency bands is put into the kernel ELM classifier for classification. The experiment divides the sample set into three parts: training set, validation set, and test set. The test set is divided by the ten-fold cross-validation method to test the generalization ability of the model; the validation set is divided by the leave-one-out validation method to tune the hyperparameters in the training model.

[0084] 4.1) First, for the given kernel matrix in each frequency band, determine the kernel ELM classifier for a single frequency band, that is, obtain 4 single-frequency band kernel ELM classifiers for signals in multiple frequency bands:

[0085]

[0086] Where N is the total number of samples; g(x) is the output; k(x,x i ) is the value in the kernel matrix of x and the i-th training sample; k n×n is the training set graph kernel matrix; n is the number of samples in the training set; T is the sample category in the training set; Cr is the penalty factor. When the penalty factor is too large, it may lead to poor fitting and low classification accuracy; when the penalty factor is too small, it may lead to overfitting. Here, Cr=10.

[0087] 4.2) The ELM models trained in different frequency bands are fused. The formula of the fused ELM model is as follows:

[0088]

[0089]

[0090] Where N = 4, ki is the kernel used in the i-th frequency band model, λ i is the weight and satisfies the following formula:

[0091]

[0092] 4.3) The value of g(x) is 1 or -1. Based on the value of g(x), the classification of normal and abnormal signals can be completed.

[0093] like Figure 3 As shown, the present invention also provides a mild cognitive impairment EEG analysis system based on multi-band PCMI-EBN, the analysis system comprising: an EEG data input module, a preprocessing module, a PCMI-EBN display module and a classification result display module;

[0094] The EEG data input module is used to collect EEG signal history data and input the collected history data;

[0095] The preprocessing module is used to perform denoising on the input historical data, extract brain wave signals within a certain frequency band, and send them to the PCMI-EBN display module;

[0096] The PCMI-EBN display module is used to construct a PCMI-EBN brain network with the EEG data of each frequency band and calculate the kernel matrix of each sample under multiple sparsity;

[0097] The classification result display module is used to classify the kernel matrix obtained by the PCMI-EBN display module and confirm normal and abnormal signals according to the classification results.

[0098] A readable storage medium includes a memory storing a program and a processor, wherein the processor executes the above-mentioned method for analyzing EEG signals of mild cognitive impairment based on a multi-band PCMI-EBN kernel.

[0099] The method for analyzing EEG signals of mild cognitive impairment based on multi-band PCMI-EBN of the present invention, on the premise of obtaining the EEG data of the subject, adopts the method for analyzing EEG signals of mild cognitive impairment based on multi-band PCMI-EBN, obtains the classification result of EEG, and realizes the recognition and judgment of the brain state of mild cognitive impairment by analyzing the result. Through the method of the present invention, accurate classification of EEG data under mild cognitive impairment and healthy brain state can be realized, wherein the accuracy rate of classification result based on multi-band PCMI-EBN atlas on the test set can reach 88.57%, which is higher than the current classification result based on the data set, so the method of the present invention can realize the discrimination under the state of mild cognitive impairment.

[0100] Here is a research example:

[0101] EEG data of 61 subjects were collected, including 32 normal patients and 29 MCI patients, all aged over 55 years old, and all subjects had primary or higher education background. Those with a history of major mental illness, substance abuse, head trauma, serious medical illness, and dementia were excluded. All subjects were interviewed by neuropsychiatry according to the Peterson MCI criteria, and a mild mental state examination (MMSE) score of 21-26 was considered MCI, and a score greater than 27 was considered normal.

[0102] First, the collected raw EEG data is preprocessed to obtain denoised EEG data and EEG data of four frequency bands. Then, the PCMI-EBN brain network is constructed based on the EEG data of each frequency band, and the kernel matrix of each sample under multiple sparsity is calculated on this basis. The kernel matrix calculated on each single frequency band is put into the single frequency band kernel ELM classifier for training, and multiple single-channel ELM classifiers with excellent classification effect are obtained. Each single frequency band classifier is fused according to the fusion algorithm to obtain a multi-band kernel ELM classifier, and the kernel matrix obtained on multiple frequency bands is put into the multi-band kernel ELM classifier for classification. Based on the ten-fold cross validation and leave-one-out validation method, the method of the present invention can realize the accurate classification of EEG data under mild cognitive impairment and normal state, and the accuracy rate can reach 88.57%.

[0103] The above is a detailed introduction to the method and system for analyzing EEG signals of mild cognitive impairment based on multi-band PCMI-EBN provided in the embodiments of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application; at the same time, for those skilled in the art, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

[0104] For example, certain words are used in the specification and claims to refer to specific components. Those skilled in the art should understand that hardware manufacturers may use different nouns to refer to the same component. This specification and claims do not use differences in names as a way to distinguish components, but use differences in the functions of components as the criteria for distinction. As mentioned throughout the specification and claims, "including" and "comprising" are open-ended terms, so they should be interpreted as "including / including but not limited to". "Approximately" means that within an acceptable error range, those skilled in the art can solve the technical problem within a certain error range and basically achieve the technical effect. The subsequent description of the specification is a preferred embodiment of the present application, but the description is for the purpose of illustrating the general principles of the present application, and is not used to limit the scope of the present application. The scope of protection of the present application shall be determined by the definition of the attached claims.

[0105] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a product or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such a product or system. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the product or system including the elements.

[0106] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0107] The above description shows and describes several preferred embodiments of the present application, but as mentioned above, it should be understood that the present application is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the application concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art do not deviate from the spirit and scope of the present application, and should be within the scope of protection of the claims attached to the present application.

Claims

1. A method for EEG analysis of mild cognitive impairment based on multi-band PCMI-EBN, characterized in that: The following steps are involved: S1) collecting EEG signal historical data samples, performing denoising processing, and extracting EEG signal data in multiple frequency bands of each EEG signal historical data sample; The specific steps are: S1.1) Remove the noise frequency bands below 0.5 Hz and above 32 Hz in the collected EEG signal historical data samples, and then use the stationary wavelet transform and independent component analysis method to denoise; S1.2) Using the segmented aggregation approximation method to compress the EEG signal historical data samples processed by S1.1); S1.3) extracting EEG signal data of multiple frequency bands of each EEG signal historical data sample in the compressed data using a short-time Fourier transform method; The multiple frequency bands are four frequency bands, namely: δ: 1-3 Hz, θ: 4-7 Hz, α: 8-13 Hz and β: 14-30 Hz; S2) constructing the PCMI-EBN of the causal effect brain network of the single-band EEG signal data for each EEG signal history data sample; The specific steps are: S2.1) taking the ranked conditional mutual information PCMI between the channel where the EEG signal data in a single frequency band is located and all other remaining channels as the coupling feature, and constructing a brain network matrix according to the number of EEG signal channels, where the number of EEG signal channels is a, and a is a positive integer greater than 0; S2.2) converting the constructed brain network matrix into a binary matrix by threshold transformation method, and recording the binary matrix as PCMI-EBN; S3) performing sparse processing on the single-band PCMI-EBN obtained in S2) according to the set sparsity value to obtain a single-band kernel matrix under the sparsity value; The specific steps are: S3.1) Setting the sparsity value, S3.2) performing sparse processing on the PCMI-EBN of the single-band EEG signal data obtained in S2.2) according to the sparsity value set in S3.1), and obtaining a kernel matrix of the single-band under the sparsity value; S4) repeating S2)-S3) and then fusing to obtain the kernel matrix of multiple frequency bands of all EEG signal historical data samples; The specific steps are: S4.1) Repeat S2)-S3) to obtain the kernel matrix of a single frequency band under the sparsity value of multiple frequency bands of all EEG signal historical data samples; S4.2) The following formula (1) is used to fuse the kernel matrix of a single frequency band in the kernel matrix of multiple frequency bands of all EEG signal historical data samples. The formula is as follows: （1） In the formula, is the kernel when the sparsity value is i, is the weight of the graph core when the sparsity value is i, and N is the number of graph cores with different sparsity values; S5) according to S4), a kernel matrix of multiple frequency bands of all EEG signal historical data samples is obtained, the kernel matrix of multiple frequency bands of all EEG signal historical data samples obtained in S4) is input into a multi-band kernel ELM classifier, and the classification of normal signals and abnormal signals is completed according to the output results, so as to realize EEG analysis of mild cognitive impairment; The specific steps are: S5.1) Input the single-band kernel matrices of the multiple-band kernel matrices of all EEG signal historical data samples obtained in S4) into the single-band ELM classifier in sequence, and then train them to obtain multiple single-band kernel ELM classifiers. The formula is as follows: ， （2） Among them, g(x) is the output; k(x,x i ) is the value in the kernel matrix of x and the i-th training sample; is the training set graph kernel matrix; n is the number of samples in the training set; T is the sample category in the training set; Cr is the penalty factor, Cr=10; S5.2) Then, multiple single-band kernel ELM classifiers of each EEG signal historical data sample are fused to obtain a multi-band kernel ELM classifier, and the kernel matrix of multiple bands of all EEG signal historical data samples obtained in S4) is input into the multi-band kernel ELM classifier to obtain : （3） Where b is the number of frequency bands, is the kernel used in the i-th frequency band model, is the weight of the i-th single-band classifier and satisfies ; if The value of is 1, which is a normal data sample. If the value of is -1, it is an abnormal data sample, and the classification of normal and abnormal signals can be completed.

2. The method according to claim 1, characterized in that The sparsity value in S3.1) is 12%~40%, and the sampling step is 2%.

3. A mild cognitive impairment EEG analysis system based on multi-band PCMI-EBN is used to implement the mild cognitive impairment EEG analysis method as described in any one of claims 1-2, characterized in that: The analysis system comprises: an EEG data input module, a preprocessing module, a PCMI-EBN display module and a classification result display module; Wherein, the EEG data input module is used to adopt EEG signal historical data samples and input the collected EEG signal historical data samples to the preprocessing module; The preprocessing module is used to perform denoising on the input EEG signal historical data samples, extract EEG signal data in multiple frequency bands of each EEG signal historical data sample, and send the data to the PCMI-EBN display module; The PCMI-EBN display module is used to construct a causal brain network PCMI-EBN of a single-band EEG signal data of each EEG signal historical data sample, and to construct a causal brain network PCMI-EBN of a single-band EEG signal data of each EEG signal historical data sample to obtain a multi-band kernel matrix; The classification result display module is used to classify the kernel matrix obtained by the PCMI-EBN display module, and classify the EEG signal history data samples as normal or abnormal signals according to the classification results.

4. A readable storage medium, comprising a memory storing a program, and a processor, wherein the processor executes the method for analyzing EEG signals for mild cognitive impairment as described in any one of claims 1 to 2.

Images