Study On Task State Eeg Signal Recognition Method For Patients With Mild Cognitive Impairment

As China enters the aging stage,Alzheimer’s Disease(AD)patients are increasing.So far,there has been no cure for AD.As a result,people have changed their research focus to the early stage of AD: the diagnosis of Mild cognitive Impairment(MCI).The research in this field has attracted increasing attention from relevant researchers at home and abroad.At present,he is an effective method of rapid real-time diagnosis of MCI through Electroencephalogram(EEG)signal analysis.Because of his own advantages,deep learning has become a research hotspot in the field of EEG signal classification,and it is extremely important to extract features suitable for deep learning model classification from EEG signals.In this paper,from the two aspects of eeg signal feature extraction and deep learning,the multi-spectral image transformation method is improved,a new eeg signal feature extraction method is proposed,and the multi-input convolutional neural network is improved correspondingly.Firstly,a method of multi-spectral image transformation is improved,and a method of eeg feature extraction based on multi-spectral image fusion is proposed.In this method,the brain is divided into several different brain regions,and the eeg signals of different brain regions are converted into multiple multi-spectral images by combining the multi-spectral image transformation method,and then the multi-spectral images of multi-brain regions are fused into a multi-spectral image with higher quality and richer information content by using the pixel-level image weighted fusion method.Secondly,the multi-input convolutional neural network is improved.In order to solve the problem that the classical convolutional neural network can not obtain ideal diversified features,the multi-input convolutional neural network is introduced.For eeg multispectral images,a single scale convolution kernel is proposed to construct the network structure,so as to make it more suitable for the classification of eeg multispectral images,that is,the single scale multi-input convolutional neural network classification method for eeg signals.Finally,the task-state electroencephalogram signals of amnestic Mild Cognitive Impairment(aMCI)patients with type 2 diabetes mellitus(T2DM)and Normal Control(NC)patients were taken as experimental data sets for experimental analysis.The new feature extraction method and the improved classification method are compared with the existing correlation algorithms respectively,and the results show that the two methods proposed in this paper can effectively improve the classification performance of eeg signals of the two types of samples.