Electroencephalogram Recognition of Alzheimer's Brain with SecVibratPSO-SVM Frame

Abstract

In order to investigate the abnormalities of brain connectivity in Alzheimer's disease (AD), and to improve the EEG recognition rate of AD, brain network of linear coherence was constructed to use background electroencephalogram (EEG) signals from patients with AD patients and normal elderly control group. In this study, we extracted 8 brain network features from each frequency band(Delta, Theta, Alpha and Beta). Statistical analysis was used to investigate whether there were significant differences between the characteristics of the AD group and the control group, and further support vector machine (SVM) classification analysis was conducted to identify the differences between the two groups. As an intelligent optimization algorithm, particle swarm optimization (PSO) and its improved algorithm (SecvibratPSO) provide a new possibility for effective feature screening of brain networks. PSO and SecvibratPSO were applied to screen feature combinations of brain networks in single band and in the cross band combinations. The simulation results showed that the two algorithms could determine the optimal feature combination of brain network and improved the EEG recognition rate of AD, but the accuracy of the SecvibratPSO algorithm was higher, reaching 0.8891. These results indicated that the SecvibratPSO algorithm is an effective method to identify the abnormal topological structure of AD brain network.