EEG-Based Alcoholism Detection: A Channel-Wise Approach Using 1D CNN

Alcohol dependence is a global health issue, causing severe impacts on the respiratory and circulatory systems and potentially leading to death. There is a growing need for effective diagnostic techniques for alcohol use disorder (AUD). Electroencephalogram (EEG) signals, which measure brain activity, have proven useful in identifying neurological disorders, including alcoholism. Traditional diagnostic methods often rely on subjective evaluations, whereas deep learning offers an automated approach for analyzing EEG data. This study presents a deep learning approach for classifying alcoholrelated EEG data using a 1D Convolutional Neural Network (CNN). The model analyzes frequency-based features from Power Spectral Density (PSD) analysis to detect alcohol-induced brain activity patterns. It achieved a peak classification accuracy of 98.95 % on channel 16 using the 1 Hz to 50 Hz frequency band. Performance varied across frequency bands with Delta, Theta, Alpha, and Beta bands achieving accuracies between 71.58 % and 76.84 %. A general challenge in EEG-based alcohol detection is the difficulty in accurately identifying alcohol-induced changes in brain activity. Traditional methods often face issues like low classification accuracy, subjectivity, and complexity, making them less reliable for consistent and accurate alcohol detection in clinical settings. To solve this problem, a novel approach was developed that focuses on simplifying the detection process while achieving high classification accuracy. An effective and straightforward deep learning technique was applied that allowed us to achieve optimal classification performance. Additionally, the bestperforming EEG channel (FP1) was identified, and the impact of different frequency ranges on classification accuracy was analyzed, enhancing the model's effectiveness for clinical use.