부산대학교 도서관

This study implements Electroencephalogram (EEG) signals in a data-driven approach based on functional brain connectivity to identify brain abnormal cortical connectivity in opioid abusers. The prime objective here is to identify opioid addiction. A 19-channel resting-state EEG signal was recorded from 22 opioid addicts and 22 normal, closely matched individuals. First, brain networks associated with two groups were constructed using the Mutual Information (MI) metric in frequency bands of the delta, theta, alpha, beta, gamma, and wideband. The groups were then divided using discriminative graph features. The results from the Support Vector Machine (SVM) classifier, evaluated by the Leave-one-out cross-validation method, achieved accuracy, sensitivity, and specificity of 100% and an F-score of 1.00 in the gamma frequency band using the AdaBoost subsampling approach. The findings indicate that the implemented pipeline, based on MI quantity between EEG data channels and combined with graph metrics, can help detect opioid addiction.