Automated Classification of Alcoholism Using Discrete Cosine Harmonic Wavelet-Packet Transform

Alcoholism, characterized by excessive alcohol consumption, leads to addiction and life-threatening health complications. This not only affects an individual's physical health but also their mental and social well-being. While traditional self-reported survey-based methods often lack reliability, neuroimaging studies provide more objective and accurate data. This study introduces the Discrete Cosine Harmonic Wavelet-Packet Transform (DCHW-PT) as an innovative method for automated alcoholism detection using EEG data. To the best of our knowledge, this is the first implementation of DCHW-PT in this domain. The framework leverages DCT’s computational simplicity, producing precise real coefficients, unlike the DFT. It retains HWT’s benefits, such as built-in decimation and interpolation, whereas DWT requires anti-aliasing and anti-imaging filters. WPT enhances multi-resolution analysis, and the shift-invariant nature of DCHW-PT provides an effective solution for detecting transient EEG signals associated with alcoholism. Extracted features—including Hjorth parameters (Activity, Mobility, Complexity), kurtosis, standard deviation, mean, energy, and skewness—offer a comprehensive statistical EEG description. Dimensionality reduction is achieved via a t-test, and model performance is assessed using accuracy, sensitivity, specificity, and F1-score, with 10-fold cross-validation. The Ensemble-Subspace classifier achieves 98.3% accuracy, with sensitivity (98.59%) and specificity (98.01%), surpassing traditional methods and aligning with leading EEG-based alcoholism detection algorithms. Ultimately, this study demonstrates the effectiveness of DCHW-PT in alcoholism detection, setting a foundation for future research in advanced alcoholism diagnostic applications.