A Machine Learning Approach for Addiction Detection Using Phase Amplitude Coupling of EEG Signals

Currently, the detection of opioid addiction is done by biological tests, but these tests are time-consuming, and their result can be changed by applying some tricks. Using bio-signals such as EEG for detecting opioid abuse can be a good alternative to the current biological tests. In this study, we are aimed to employ EEG signal to detect opioid addiction. The dataset of this study consist of a 19-channel resting-state EEG signal recorded from 22 opioid addicts and 22 healthy normal individuals (without a history of substance abuse). Extracted features of EEG signal include phase-amplitude coupling (PAC) among delta, theta, alpha1, alpha 2, beta1, beta2, and gamma frequency bands. Informative features that can discriminate addicted groups from normal groups are selected by statistical tests and the Minimum Redundancy Maximum Relevance (mRMR) technique. Then selected features are fed to the k-nearest neighbors (KNN) classifier, which is evaluated by Leave-one-out cross-validation. The proposed algorithm classified the addicted and normal group with 93.18% accuracy, 100% sensitivity, and 86.36% specificity. Analyzing the results indicates that delta-beta1 coupling and FZ channel had the most participation in the selected features. The obtained results show that the proposed technique based on EEG signals PAC can be useful in addiction detection.