Classification of alcoholic subjects using multi channel ERPs based on channel optimization and Probabilistic Neural Network

Abstract

The Alcoholism is an addictive disorder, which causes social, physical, psychiatric and neurological damages on individuals. In this paper, Global Field Synchronization (GFS) measurements of multi channel ERP (Event Related Potential) signals in Delta, Theta, Alpha, Beta and Gamma frequency bands are used as discriminating feature vectors in the classification of alcoholic and non-alcoholic control subjects. GFS measurements show the functional connectivity of neurocognitive networks in the patient's brain as a response to a given stimuli type. A channel optimization algorithm that improves recognition accuracy by selecting channels with the most significant attributes is applied during Global Field Synchronization prior to classification stage. Probabilistic Neural Network is used as the classifier. The proposed system successfully classifies alcoholic and non-alcoholic subjects with accuracy over 80%.