Optimal EEG Channel Selection for Vascular Dementia Identification Using Improved Binary Gravitation Search Algorithm

The aim of the present study was to select optimal channels that may help in detecting the abnormalities in the electroencephalogram (EEG) of vascular dementia (VaD) patients. Spectral entropy \( (SpecEn) \), approximation entropy \( (ApEn) \) and permutation entropy \( (PerEn) \) have been extracted from the EEG background activity of 5 VaD, 15 patients with stroke-related mild cognitive impairment (MCI) and 15 healthy control subjects during a working memory (WM) task. EEG artifacts were removed using automatic independent component analysis and wavelet denoising technique (AICA-WT). In order to reduce the computational time, improved binary gravitation search algorithm (IBGSA) channel selection was used to find the most effective EEG channels for VaD patients’ detection. Eight channels were found suitable to extract EEG markers that help to detect dementia in the early stages. Moreover, k-nearest neighbors (kNN) was used after the IBGSA technique. The IBGSA technique increased the kNN classification accuracy from 86.67 to 90.52%. These results suggest that IBGSA consistently improves the discrimination of VaD, MCI patients and control normal subjects and it could be a useful feature selection to help the identification of patients with VaD and MCI.