Differential Evolution Based Channel Selection Algorithm on EEG Signal for Early Detection of Vascular Dementia among Stroke Survivors

Abstract

The purpose of this investigation was to identify optimal electroencephalogram (EEG) channels for detection of early onset dementia. EEGs of five vascular dementia (VD) patients, fifteen stroke-related patients with mild cognitive impairment (MCI) and fifteen healthy subjects were recorded during cognitive impairment of working memory (WM) when eyes were closed. This paper demonstrate the combination of several technique for the analyses of multi-channel EEG signals these are Savitzky–Golay (SG) filter which was used in denoising stage, refined composite multiscale dispersion entropy (RCMDE) was used to characterize the EEG dataset. Moreover, Differential evolution-based channel selection algorithm (DEFS_Ch) was performed to identify the EEG channels with greatest efficacy for detecting early stage VD patients. In classification stage, Support vector machine (SVM) classification scheme was used befor and after applying the DEFS_Ch selection algorithm. The results revels the most suitable six channels that enhanced classification performance under cognitive load of WM condition. The DEFS_Ch algorithm raised the SVM classification accuracy from 89.52% to 95.24%, indicating that DEFS_Ch may offer a useful channel selection algorithm for consistent improvement of the identification of VD patients, MCI patients, as well as healthy control subjects.