Alzheimer disease auxiliary detection system and method based on electroencephalogram signal channel selection and gating circulation unit

Abstract

The invention belongs to but is not limited to the technical field of artificial intelligence, and discloses an Alzheimer's disease diagnosis method based on electroencephalogram signal channel selection and a gating circulation unit, which selects an electroencephalogram channel through a particle swarm algorithm, reduces redundant information in electroencephalogram signal data, and improves the accuracy of electroencephalogram data analysis; feature extraction is carried out on the electroencephalogram signal data through principal component analysis, and the importance and distinction degree of features are improved; and finally, inputting the electroencephalogram signals into a gated circulation unit neural network for classification prediction, thereby reducing the number of electroencephalogram channels and reducing the dimension of data input. And the calculation efficiency is improved while the classification performance is ensured. According to the method, the classification performance of the Alzheimer's disease patients and the healthy subjects is improved, and accurate and efficient classification of the Alzheimer's disease patients and the healthy subjects is achieved.

Description

Technical Field

[0001] The present invention belongs to but is not limited to the field of artificial intelligence technology, and in particular relates to an Alzheimer's disease auxiliary detection system and method based on EEG signal channel selection and gated recurrent units. Background Art

[0002] Dementia is a general term for a series of symptoms that are caused by diseases that affect the brain and have an impact on memory, thinking, behavior and emotions. Alzheimer's disease and vascular dementia are the most common types of dementia, accounting for 90% of dementia cases. Alzheimer's disease is a progressive neurodegenerative disease that often occurs in people over the age of 65 and is usually characterized by memory loss and cognitive dysfunction. Patients gradually lose their understanding and ability to carry out daily activities, including language ability, spatial orientation ability and executive function. At different stages of the disease, patients also show symptoms such as emotional instability, behavioral changes and decreased ability to take care of themselves. As the number of Alzheimer's patients continues to increase, it has brought a huge economic burden to many countries. It is of great significance to develop a system and method that can effectively diagnose Alzheimer's disease.

[0003] Traditional machine learning methods have good performance in EEG signal classification tasks, but when faced with high-dimensional data, they often face challenges in extracting features and limited model expression capabilities. With the continuous development of deep learning, it is possible to automatically learn higher-level abstract feature representations from high-dimensional data and process such complex data. Recurrent neural networks have become a powerful tool for processing EEG signal time series classification tasks. Summary of the invention

[0004] In view of the problems existing in the prior art, the present invention provides a method and system for diagnosing Alzheimer's disease based on EEG signal channel selection and gated recurrent units; an algorithm combining principal component analysis feature extraction with gated recurrent unit classification is used to achieve the classification task of healthy subjects and Alzheimer's patients.

[0005] The present invention is implemented as follows: a method for diagnosing Alzheimer's disease based on EEG signal channel selection and gated recurrent unit, comprising the following steps:

[0006] S1, 19 EEG signal channels are selected through the particle swarm algorithm, the selected EEG channel combination is used as input data, and each input data is standardized;

[0007] S2, use principal component analysis to retain the main features in the input data and convert high-dimensional data into low-dimensional data; then directly perform singular value decomposition on the matrix of the input data, truncate the obtained singular values ​​according to the threshold of principal component analysis, and use the retained principal components as the final input of the model;

[0008] S3, after feature extraction, a gated recurrent unit neural network is used to perform the classification task. The EEG data retained by the principal component analysis is first input into the model, and then hierarchical feature extraction is performed by stacking two gated recurrent units.

[0009] Furthermore, the particles in S1 select the best EEG channel combination from the 19 EEG channels. When initializing the particles, the unselected EEG channels are recorded as 0 and the selected EEG channels are recorded as 1. The performance index of channel selection is the EEG signal energy E i :

[0010]

[0011] Among them, i is each band, is the square value for each band i.

[0012] Furthermore, the selected EEG channel data is standardized:

[0013]

[0014] Among them, z is the standardized value, x is the original data, μ is the mean of the original data, and σ is the standard deviation of the original data. After standardization, the EEG data will be scaled to a distribution with a mean of 0 and a standard deviation of 1.

[0015] Furthermore, the EEG signal data is input into S2, and the threshold is determined according to the cumulative contribution rate through principal component analysis, and the number of effective principal components is determined by the threshold:

[0016]

[0017] Among them, L i Cumulative contribution rate, λ i is the eigenvalue of the covariance matrix, and δ is the selected threshold; the larger the threshold δ is, the more effective principal components there are and the more information they carry.

[0018] Furthermore, the singular value decomposition is directly applied to the original data matrix A∈R m×n Perform the calculation:

[0019] A=UΣV T (4)

[0020] Among them, U and V are both orthogonal matrices, U is called a left singular matrix, and U=(u1,u2,…,u m )∈R m×m , V is called a right singular matrix, and V=(v1,v2,…,v m )∈R n×n , u1 is called the left singular vector, vi It is called the right singular vector. Σ is an m×n matrix consisting of non-negative diagonal elements in descending order. Σ=diag(σ1,σ2,…,σ r ) is a diagonal matrix, σ i are the singular values ​​of the matrix A, and σ1≥σ2≥…≥σ r ≥0,r=min(m,n);

[0021] The obtained singular value σ i The principal components are truncated according to the threshold and used as the final input of the neural network model.

[0022] Another object of the present invention is to provide an Alzheimer's disease diagnosis method based on EEG signal channel selection and gated recurrent unit and an Alzheimer's disease diagnosis system based on EEG signal channel selection and gated recurrent unit, comprising:

[0023] The EEG channel selection module selects 19 EEG signal channels through the particle swarm algorithm, uses the selected EEG channel combination as input data, and performs standardization on each input data;

[0024] The feature extraction module uses principal component analysis to retain the main features in the input data and convert high-dimensional data into low-dimensional data. It then directly performs singular value decomposition on the matrix of the input data, truncates the obtained singular values ​​according to the threshold of principal component analysis, and retains the principal components as the final input of the model.

[0025] The classification prediction module uses a gated recurrent unit neural network to perform the classification task after feature extraction. The EEG data retained by the principal component analysis is first input into the model, and then hierarchical feature extraction is performed by stacking two gated recurrent units.

[0026] Another object of the present invention is to provide a computer device, the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the Alzheimer's disease diagnosis method based on EEG signal channel selection and gated recurrent unit.

[0027] Another object of the present invention is to provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the Alzheimer's disease diagnosis method based on EEG signal channel selection and gated recurrent units.

[0028] Another object of the present invention is to provide an information data processing terminal, which includes the Alzheimer's disease diagnosis system based on EEG signal channel selection and gated cycle unit.

[0029] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0030] First, the present invention selects EEG channels through particle swarm algorithm, reduces redundant information in EEG signal data, and improves the accuracy of EEG data analysis. Then, the EEG signal data is subjected to feature extraction through principal component analysis, which improves the importance and discrimination of the features. Finally, it is input into the gated recurrent unit neural network for classification prediction. The neural network model extracts the input EEG data features in two stages to realize the transmission and processing of information; the number of EEG channels is reduced, and the dimension of data input is reduced. While ensuring the classification performance, the computational efficiency is improved.

[0031] Second, the technical solution of the present invention mainly solves the following technical problems in the existing auxiliary diagnosis methods for Alzheimer's disease in industrial applications, and has achieved significant technical progress:

[0032] 1. Improve the diagnostic accuracy of Alzheimer's disease: The existing diagnostic methods for Alzheimer's disease process EEG signals in a relatively simple way and cannot fully extract complex temporal and spatial features, resulting in low diagnostic accuracy. The present invention combines the particle swarm algorithm (PSO) for EEG channel selection, uses principal component analysis (PCA) and singular value decomposition (SVD) for feature extraction, and then uses the gated recurrent unit (GRU) neural network for classification, which can fully extract the characteristic signals of Alzheimer's disease and significantly improve the accuracy and sensitivity of diagnosis.

[0033] 2. Reduce data redundancy and improve computing efficiency: In traditional EEG signal diagnosis methods, data from all channels are usually processed, resulting in a large amount of redundant information that increases the computational complexity. The present invention uses a particle swarm algorithm to select the EEG signal channel with the most characterization capability, reducing the amount of unnecessary data and thus reducing data redundancy. On this basis, principal component analysis and singular value decomposition further reduce the dimension and optimize the data, so that the system can quickly obtain diagnostic results with less computing resources.

[0034] 3. Enhance the depth and temporal correlation of feature extraction: Existing feature extraction methods often have difficulty capturing deep and temporal correlation features in EEG signals, which affects the diagnosis results. The present invention can effectively capture the temporal correlation in time series data and extract higher-level features layer by layer by introducing a stacked gated recurrent unit (GRU) neural network. This approach ensures that the system can more accurately identify the differences between Alzheimer's patients and healthy individuals, enhancing the accuracy and stability of diagnosis.

[0035] 4. Improve the intelligence and adaptability of the diagnostic system: Traditional diagnostic systems often rely on manually set feature extraction rules, lack adaptability, and are difficult to cope with the complex pathological characteristics of different patients. The present invention automatically learns the complex features in EEG signals through a deep learning algorithm. Based on the gated recurrent unit (GRU) model, the system can automatically adapt to the signal characteristics of different patients, reduce manual intervention, and achieve a more intelligent diagnostic process.

[0036] The present invention has achieved significant technological progress in EEG signal processing, feature extraction and classification algorithms, greatly improving the accuracy, efficiency and intelligence level of the Alzheimer's disease diagnosis system. It has broad industrial application prospects, especially in medical diagnosis auxiliary systems and large-scale Alzheimer's disease screening, and can provide clinicians with more accurate and reliable diagnostic auxiliary tools. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of a method for diagnosing Alzheimer's disease based on EEG signal channel selection and gated recurrent unit provided by an embodiment of the present invention;

[0038] Figure 2 is a combination diagram of EEG channels selected by the particle swarm algorithm provided by an embodiment of the present invention;

[0039] Figure 3 1 is an accuracy curve and a loss curve diagram of a gated recurrent unit neural network provided by an embodiment of the present invention;

[0040] Figure 4 is a comparison diagram of the parameters of the gated recurrent unit neural network under different EEG channel combinations provided by an embodiment of the present invention;

[0041] Figure 5 is a comparison chart of the accuracy of the gated recurrent unit neural network under different EEG channel combinations provided by the embodiment of the present invention;

[0042] Figure 6 It is a structural diagram of an Alzheimer's disease diagnosis system based on EEG signal channel selection and gated recurrent unit provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] Example 1: Early screening system for Alzheimer's disease based on EEG signals

[0045] In medical institutions or elderly care centers, early screening of Alzheimer's disease is crucial and can significantly improve the effects of early intervention and treatment. The Alzheimer's disease diagnosis system based on EEG signal channel selection and gated recurrent unit (GRU) of the present invention can be deployed in hospitals or care centers for routine EEG signal detection and analysis of the elderly. During the screening process, the patient wears an EEG signal acquisition device, and the system automatically selects the most representative EEG signal channel through a particle swarm algorithm (PSO) for real-time data acquisition. Then, the system reduces the data dimension through principal component analysis (PCA) and singular value decomposition (SVD), extracts representative features, and classifies the data through a GRU neural network to quickly obtain a diagnosis result. The system can identify individuals with Alzheimer's disease in a short time, thereby providing clinicians with an early diagnosis basis and helping patients receive intervention and treatment in a timely manner.

[0046] This screening system is suitable for deployment in hospitals, community clinics and elderly care institutions to achieve early warning and routine physical examinations of Alzheimer's disease and improve the efficiency of disease screening.

[0047] Example 2: Remote Monitoring and Diagnosis Platform for Alzheimer's Disease Patients

[0048] In telemedicine and home care, the EEG signal diagnosis system of the present invention can be used for long-term monitoring and diagnosis of patients with Alzheimer's disease. The patient wears an EEG signal acquisition device at home and transmits the EEG data to the telemedicine platform wirelessly. The platform analyzes the data through the channel selection algorithm and GRU neural network of the invention, monitors the patient's EEG activity in real time, and performs automatic classification diagnosis.

[0049] The system can not only monitor the condition of Alzheimer's patients in real time, but also provide intelligent early warnings based on data trends, reminding medical staff or family members to take necessary care measures. In addition, the platform can automatically generate diagnostic reports for doctors to review remotely, reducing the need for patients to go to the hospital for frequent examinations and improving their quality of life.

[0050] This remote monitoring system is particularly suitable for Alzheimer's patients who require long-term care. It enables home diagnosis, remote monitoring and telemedicine support, and is widely used in industrial fields such as home-based elderly care, smart nursing, and telemedicine platforms.

[0051] like Figure 1 As shown, an Alzheimer's disease diagnosis method based on EEG signal channel selection and gated recurrent unit provided by an embodiment of the present invention comprises the following steps:

[0052] S1, 19 EEG signal channels are selected by particle swarm algorithm, and the selected EEG channel combination is used as input data; and each input data is standardized to ensure that all features are on the same scale, which is convenient for extracting EEG features related to the classification task; the EEG channel combination selected by the particle swarm algorithm is as follows Figure 2 shown.

[0053] EEG channel selection helps eliminate the interference of redundant channels, reduce the amount of data processing calculations, and reduce computing costs. EEG signal energy directly reflects brain activity and provides overall information about EEG signals. The overall characteristics of EEG signals can be captured through EEG signal energy.

[0054] The specific process of S1 is that particles select the best EEG channel combination from 19 EEG channels. When initializing particles, the unselected EEG channels are recorded as 0 and the selected EEG channels are recorded as 1. The performance indicator of channel selection is the EEG signal energy E i :

[0055]

[0056] Among them, i is each band, is the square value of each band i. EEG signals have different amplitude ranges. In order to extract EEG features related to classification, the selected EEG channel data needs to be standardized:

[0057]

[0058] Among them, z is the standardized value, x is the original data, μ is the mean of the original data, and σ is the standard deviation of the original data. After standardization, the EEG data will be scaled to a distribution with a mean of 0 and a standard deviation of 1.

[0059] S2, use principal component analysis to retain the main features of the input data, convert high-dimensional data into low-dimensional data, and improve the efficiency of data analysis; then directly perform singular value decomposition on the matrix of the input data, truncate the obtained singular values ​​according to the threshold of principal component analysis, and use the retained principal components as the final input of the model;

[0060] Feature extraction is to concentrate the features of EEG signals, which can effectively reduce the complexity of data and improve the classification performance of the model. Feature extraction is divided into two parts: principal component analysis and singular value decomposition. The goal is to reduce the data dimension of the input EEG signal data through feature extraction.

[0061] The specific process of S2 is to input EEG signal data, determine the threshold according to the cumulative contribution rate through principal component analysis, and determine the number of effective principal components through the threshold:

[0062]

[0063] Among them, L i Cumulative contribution rate, λ i is the eigenvalue of the covariance matrix, and δ is the selected threshold. The larger the threshold δ is, the more effective principal components there are, and the more information they carry. Principal component analysis cannot avoid the impact of rounding errors when calculating the covariance matrix, while singular value decomposition can directly calculate the covariance matrix A∈R m×n Perform the calculation:

[0064]

[0065] Among them, U and V are both orthogonal matrices, U is called a left singular matrix, and U=(u1,u2,…,u m )∈R m×m , V is called a right singular matrix, and V=(v1,v2,…,v m )∈R n×n ,u1 is called the left singular vector,v i It is called the right singular vector. Σ is an m×n matrix consisting of non-negative diagonal elements arranged in descending order. Σ=diag(σ1,σ2,…,σ r ) is a diagonal matrix, σ i are the singular values ​​of the matrix A, and σ1≥σ2≥…≥σ r ≥0, r=min(m,n). The obtained singular value σ i The principal components are truncated according to the threshold and used as the final input of the neural network model.

[0066] S3, after feature extraction, a gated recurrent unit neural network is introduced to perform classification tasks. First, the EEG data retained by principal component analysis is input into the model. Secondly, in order to help the model better understand and process EEG signal time series data, hierarchical feature extraction is performed by stacking two gated recurrent units. This neural network model extracts the input EEG data features in two stages to achieve information transmission and processing. The accuracy curve and loss curve of the gated recurrent unit neural network are shown in Figure 1. Figure 3 shown.

[0067] The parameters of the gated recurrent unit neural network under different EEG channel combinations are as follows: Figure 4 As shown; the accuracy of the gated recurrent unit neural network under different EEG channel combinations is shown in Figure 5 shown.

[0068] like Figure 6As shown, an Alzheimer's disease diagnosis method based on EEG signal channel selection and gated recurrent unit provided by an embodiment of the present invention includes:

[0069] The EEG channel selection module selects 19 EEG signal channels through the particle swarm algorithm, uses the selected EEG channel combination as input data, and performs standardization on each input data;

[0070] The feature extraction module uses principal component analysis to retain the main features in the input data and convert high-dimensional data into low-dimensional data. It then directly performs singular value decomposition on the matrix of the input data, truncates the obtained singular values ​​according to the threshold of principal component analysis, and retains the principal components as the final input of the model.

[0071] The classification prediction module uses a gated recurrent unit neural network to perform the classification task after feature extraction. The EEG data retained by the principal component analysis is first input into the model, and then hierarchical feature extraction is performed by stacking two gated recurrent units.

[0072] This method is mainly based on electroencephalogram (EEG) channel selection and gated recurrent unit (GRU) neural network to realize the diagnosis of Alzheimer's disease. First, in the channel selection stage, 19 EEG signal channels are screened by particle swarm algorithm (PSO). PSO simulates group behavior to find the optimal solution. It can quickly and effectively select the most diagnostic channel combination from multiple channels, reduce unnecessary data redundancy, and improve the training efficiency and accuracy of the model. Through this process, the system obtains a set of EEG signal channels that best characterize Alzheimer's patients and normal people.

[0073] Next, we enter the data standardization stage to standardize the selected EEG signal channel data. The purpose of this process is to scale the data to the same scale to eliminate the impact of numerical differences between different channels and ensure that each channel data contributes equally to model training. The standardized data eliminates differences in measurement units and numerical ranges, making subsequent feature extraction more accurate and consistent.

[0074] In the feature extraction stage, the method uses principal component analysis (PCA) to reduce the dimensionality of the input data. PCA converts high-dimensional data into low-dimensional data by finding the main feature directions of the data. This process retains the main feature information in the input data while removing noise and redundant data. Through this dimensionality reduction process, the input data of the model is simplified, the computational burden is reduced, and the training speed and generalization ability of the model are effectively improved.

[0075] After PCA, the system further decomposes the data matrix through singular value decomposition (SVD). SVD decomposes the data matrix into singular values ​​and corresponding singular vectors, truncates the singular values ​​according to the threshold of principal component analysis, and retains those singular values ​​that can represent the main information. The purpose of this step is to further refine the feature selection, make the main components in the data more prominent, reduce unnecessary information interference, and improve the data's ability to represent classification tasks.

[0076] Next, the system entered the classification stage, with the Gated Recurrent Unit (GRU) neural network as its core. GRU is a variant of the recurrent neural network (RNN) that can better process sequence data and remember long-term dependent information than the standard RNN. The processed EEG data is input into the GRU model, and hierarchical feature extraction is performed by stacking two GRU layers. The first layer of GRU extracts the time-correlation features in the original EEG data, while the second layer of GRU further deeply mines the advanced features in the data to ensure that the model can more fully capture the characteristics of Alzheimer's disease.

[0077] Finally, the GRU neural network outputs a classification result, which indicates that the input EEG data belongs to an Alzheimer's patient or a healthy person. Through hierarchical feature extraction of multiple GRU units, the model can extract key features representing Alzheimer's disease from the complex time series data of EEG signals and make high-precision classification decisions. This method not only improves the accuracy of diagnosis, but also reduces the risk of misdiagnosis.

[0078] Example 1: Alzheimer's disease early screening system in hospital

[0079] In the neurology department of the hospital, especially in the geriatric ward, the Alzheimer's disease diagnosis method based on EEG signal channel selection and gated recurrent units is widely used to screen patients for early stage. The system can collect data from 19 EEG signal channels in real time when the patient undergoes a routine electroencephalogram (EEG) examination, and automatically select the most relevant channel for diagnosis through the particle swarm algorithm.

[0080] Once the EEG signal is collected, the system automatically performs data standardization and principal component analysis to compress the data dimension and eliminate redundant information. Subsequently, the data is analyzed through the gated recurrent unit (GRU) neural network, and the system can accurately identify the characteristics of Alzheimer's disease, especially when early symptoms have not yet become apparent. This method can not only improve screening efficiency, but also help doctors quickly identify potential patients and intervene in the early stages of the disease, greatly improving treatment outcomes and patient quality of life.

[0081] Example 2: Home use Alzheimer's disease detection device

[0082] In the field of health monitoring, the diagnosis method of Alzheimer's disease based on EEG signals is integrated into home smart health devices. For example, a wearable EEG detection helmet or hat is wirelessly connected to a smart device (such as a mobile phone or tablet), and users can perform simple EEG signal collection at home.

[0083] The device is equipped with built-in particle swarm algorithm and GRU neural network algorithm modules, which can automatically perform data processing, channel selection, feature extraction and classification after collecting EEG signals. The system can produce preliminary diagnosis results within a few minutes and prompt the user whether further professional medical consultation is needed. This home detection device is convenient for the elderly to conduct regular monitoring at home, helping to detect signs of Alzheimer's disease early, while also reducing the pressure on medical resources and providing a cost-effective health management solution for patients' families.

[0084] 2. Relevant evidence of the technical effects obtained by the embodiments of the present invention.

[0085] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0086] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. An Alzheimer's disease auxiliary detection system based on EEG signal channel selection and gated recurrent unit, characterized in that: The system includes: The channel selection module is used to screen 19 EEG signal channels through the particle swarm algorithm (PSO) to select the EEG signal channel with the best representation ability; A data standardization module is used to standardize the selected EEG signal channel data to eliminate the numerical differences between different channels; The feature extraction module uses principal component analysis (PCA) to extract features from the standardized data, retaining the main features and reducing the dimension; The classification module uses a gated recurrent unit (GRU) neural network to classify the extracted features and outputs the classification results of Alzheimer's patients or healthy people.

2. The system according to claim 1, characterized in that The channel selection module optimally selects 19 EEG signal channels through a particle swarm algorithm (PSO), screens out a channel combination that can best characterize the characteristics of Alzheimer's disease, reduces data redundancy and improves classification efficiency.

3. The system according to claim 1, characterized in that The feature extraction module uses singular value decomposition (SVD) to decompose the data matrix after principal component analysis, and cuts off the singular values ​​according to a preset threshold value, retaining the singular values ​​that can represent the main information as the final feature input.

4. The system according to claim 1, characterized in that The classification module consists of two stacked gated recurrent unit (GRU) layers, the first GRU layer is used to extract time-related features, and the second GRU layer further extracts high-level features to ensure high-precision Alzheimer's disease classification results.

5. A method for diagnosing Alzheimer's disease based on EEG signal channel selection and gated recurrent unit, characterized in that: The following steps are involved: S1, 19 EEG signal channels are selected through the particle swarm algorithm, the selected EEG channel combination is used as input data, and each input data is standardized; S2, use principal component analysis to retain the main features in the input data and convert high-dimensional data into low-dimensional data; then directly perform singular value decomposition on the matrix of the input data, truncate the obtained singular values ​​according to the threshold of principal component analysis, and use the retained principal components as the final input of the model; S3, after feature extraction, a gated recurrent unit neural network is used to perform the classification task. The EEG data retained by the principal component analysis is first input into the model, and then hierarchical feature extraction is performed by stacking two gated recurrent units.

6. The method for diagnosing Alzheimer's disease based on EEG signal channel selection and gated recurrent unit according to claim 1, characterized in that: The particles in S1 select the best EEG channel combination from the 19 EEG channels. When initializing the particles, the unselected EEG channels are recorded as 0 and the selected EEG channels are recorded as 1. The performance index of channel selection is the EEG signal energy E i : Among them, i is each band, is the square value for each band i.

7. The method for diagnosing Alzheimer's disease based on EEG signal channel selection and gated recurrent unit according to claim 5, wherein the selected EEG channel data is subjected to standardization processing: in, z is the standardized value, x is the original data, μ is the mean of the original data, and σ is the standard deviation of the original data. After standardization, the EEG data will be scaled to a distribution with a mean of 0 and a standard deviation of 1.

8. The method for diagnosing Alzheimer's disease based on EEG signal channel selection and gated recurrent unit according to claim 5, wherein the EEG signal data is input in S2, a threshold is determined according to the cumulative contribution rate through principal component analysis, and the number of effective principal components is determined through the threshold: in, L i Cumulative contribution rate, λ i is the eigenvalue of the covariance matrix, δ is the selected threshold; The larger the threshold δ is, the more effective principal components there are and the more information they carry.

9. The method for diagnosing Alzheimer's disease based on EEG signal channel selection and gated recurrent unit as claimed in claim 5, wherein the singular value decomposition is directly performed on the original data matrix A∈R m×n Perform the calculation: A=UΣV T (4) in, U and V are both orthogonal matrices, U is called a left singular matrix, and U = (u1, u2, ..., u m )∈R m×m , V is called a right singular matrix, and V=(v1,v2,…,v m )∈R n×n , u1 is called the left singular vector, v i It is called the right singular vector. Σ is an m×n matrix consisting of non-negative diagonal elements in descending order. Σ=diag(σ1,σ2,…,σ r ) is a diagonal matrix, σ i are the singular values ​​of the matrix A, and σ1≥σ2≥…≥σ r ≥0,r=min(m,n); The obtained singular value σ i The principal components are truncated according to the threshold and used as the final input of the neural network model.

Images