Resting State EEG Signals Recognition Method For Diagnosis Of Patients With Mild Cognitive Impairment

As an early stage of Alzheimers Disease(AD),the diagnosis of Mild Cognitive Impairment(MCI)has received increasing attention from scholars at home and abroad.Feature extraction and classification of electroencephalogram(EEG)signals is an important way to diagnose mild cognitive impairment.The combination of current feature extraction and deep learning has been widely used in the field of EEG signals analysis.However,how to extract effective features suitable for deep learning classification and how to construct an efficient deep learning classifier is urgently needed in the field of MCI EEG signals analysis.The two main issues.Therefore,this paper intends to explore new multi-spectral image transformation feature extraction methods and convolutional neural network classification methods,the specific work is as follows.First,this paper innovatively applies the multispectral image transformation method and convolutional neural network to the EEG signals classification task of amnestic Mild Cognitive Impairment(aMCI)patients of type 2 diabetes mellitus(T2DM).Secondly,aiming at the problem of insufficient feature quantity and computational resource waste in the multi-spectral image transformation method applied to the brain electrical signals classification task of aMCI patients,this paper proposes a feature fusion multi-spectral image transformation method.The method effectively solves the deficiencies of the existing methods by multi-feature fusion,multi-band combination and reference brain position related calculation processes.Then,for the problem that the convolutional neural network is applied to the brain electrical signal classification task of aMCI patients,the number of learning features is not rich,the loss of feature information and slow learning.This paper proposes a multi-view convolutional neural network based on InceptionV1 structure.The neural network can extract and learn input data in a multi-scale receptive field.At the same time,this paper analyzes the difficulty of selecting the learning rate in the stochastic gradient descent method,and may not be able to obtain the global optimal solution.The adaptive stochastic gradient descent method is proposed and applied to the optimization process of multi-view convolutional neural network method.Finally,using the EEG signals of aMCI patients with T2 DM and Normal Control(NC)as experimental data,the two new methods proposed in this paper are compared with the existing feature extraction and classification methods,and the algorithms are compared.The feasibility,validity and accuracy were verified.