Optimal channel selection for detecting alcoholism from EEG based on permutation entropy

Multi-channel EEG signal analysis has been widely used in the research community for detecting brain-related disorders. However, the optimal channels for detecting alcoholism are still unknown. This paper presents a permutation entropy (PE) method for selecting the optimal channel to distinguish the EEG signals of alcoholics from control drinkers. The PE features were first extracted from single electrode, 2 electrode, 3 electrode, 8 electrode and 61 electrode EEG signals respectively. Experimental results showed that the average PE values of EEG from alcoholics was significantly higher than that from control subjects and that certain single channels can achieve classification of 75.5% accuracy. Up to 91.3% accuracy can be achieved when the feature dimension is eight. The results indicate that PE values extracted from the optimal selected channel of EEG signals is sufficient to accurately identify alcoholics