A Combination of Deep Neural Network and Fuzzy Clustering for EEG-Based Alcoholism Diagnosis

Alcohol dependence is a medical condition with various biological, genetic and environmental risk factors. This study tries to provide a new system for classifying alcoholics and non-alcoholics through a deep learning structure and a fuzzy clustering approach from resting state EEG signals. The suggested framework includes various phases: (A) signal preprocessing and segmentation, (B) data augmentation, (C) deep feature extraction via CNN, and (D) data classification through fuzzy clustering. This framework was exposed to different evaluations based on performance metrics. The entire proposed framework by integrating EEG preprocessing and segmentation and data augmentation yielded the best classification performance for alcoholism detection with accuracy of 94.76%, sensitivity of 94.10%, specificity of 95.31% and F1-score of 94.81%. Experimental results showed that the proposed framework outperforms other basic and deep algorithms for alcohol addiction detection from EEG data. Also, the experiments showed that the proposed method is largely resistant to noise and can obtain acceptable results. When applied for external validation on unseen EEG data, it was able to detect alcoholic EEG with 100% accuracy, 100% sensitivity, and 100% specificity, demonstrating the robustness and generalizability of the proposed framework.