Effective Connectivity of EEG Signals in Diagnosing Alcohol Use Disorder: A Machine Learning Approach

Alcohol consumption significantly impacts public health. Excessive drinking harms physical and emotional well-being, lifestyle, and social behavior, and chronic abuse alters brain structure, leading to Alcohol Use Disorder (AUD). AUD, characterized by persistent drinking and poor impulse control, is associated with neurotransmitter and metabolic changes in the brain. Diagnosis is based on DSM-5 criteria and subjective methods like surveys and psychiatric examinations, but EEG can help predict relapse and treatment effectiveness. This study aims to develop a precise method for detecting AUD using Directed Transfer Function (DTF) and Partial Directed Coherence (PDC) algorithms as input to machine learning model. The EEG dataset comprises 30 AUD and 30 healthy control subjects. The study evaluates the accuracy, sensitivity, and precision of the algorithms, with preprocessing of EEG signals from 19 channels and analysis using DTF and PDC. The data are transformed into 19×19 adjacency matrices across 64 frequency bands. The PDC method achieved higher accuracy (94.5%) compared to DTF (89.8%).