Deep learning based diagnosis of alcohol use disorder (AUD) using EEG

alcohol addiction, alcohol use disorder (AUD), deep learning, CNN, classification

Alcohol use disorder (AUD) involves people who have difficulty controlling their drinking habits. This results in significant distress and also affects functioning normally in their daily life. The challenge in screening AUD patients using subjective measures is the dependency of this method on self-assessment, which is unreliable in the case of alcoholics as they may lie or not correctly remember facts because access to alcohol use can affect memory. The solution is to use neuroimaging modalities such as electroencephalography (EEG), which looks into the brain patterns and does not involve self-rating. This study proposes a deep learning (DL) method to classify alcoholics and healthy controls. The proposed deep learning method applies EEG feature extraction automatically and classifies the participants into relevant groups. The participants included 30 AUD patients (mean age 56.70 15.33 years) and 15 healthy controls (mean 42.67 15.90 years) who were recruited to acquire EEG data. The data were recorded during 10 minutes of eyes closed (EC) and eyes open (EO) conditions. The proposed analysis utilizes 1D CNN to fit and evaluate the classification model. From EEG data, features were extracted and classified using a convolutional neural network (CNN). The results show that the CNN has achieved the performance rendering a classification accuracy of (93%), specificity (89 % ), and sensitivity (89 % ) with an f1 score of 0.94 for the AUD group. In addition, for the healthy control group, the specificity of (100%), the sensitivity of (100%), and the f1 score of 0.91 are achieved. In conclusion, the results implicated significant neurophysiological differences between alcoholics and control.