Alcoholism Classification Based on EEG Spectrograms Using Hybrid Convolutional Neural Network-Machine Learning Models
Year: 2025
Type: article
Abstract: Alcoholism has a significant impact on brain activity, making early detection essential for effective intervention. Conventional diagnostic methods are often subjective and time-consuming, creating a need for automated approaches. This research presents a hybrid framework that integrates CNN for extracting features from EEG spectrograms with various machine learning classifiers to enhance classification accuracy. Its main contribution is the creation of a CNN–Machine Learning pipeline that merges deep feature representations with conventional classifiers such as SVM, KNN, RF, NN, SGD, DT, NB, and QDA for detecting alcoholism based on EEG data. This approach utilizes CNN’s ability to capture complex spatiotemporal patterns while systematically comparing different algorithms on the same extracted features. A total of 7,200 EEG spectrogram images were preprocessed through resizing, normalization, and labeling. CNN was used to extract 512-dimensional feature vectors, which were then classified using the selected machine learning models. Model performance was assessed through accuracy, precision, recall, F1-score, specificity, and ROC-AUC to ensure a thorough evaluation. Findings revealed that the CNN–SVM model delivered the highest results, attaining 82.41% accuracy, 78.04% precision, 90.21% recall, an F1-score of 83.69%, 74.62% specificity, and a ROC-AUC of 90.77%. QDA followed with the highest recall at 91.72% and a ROC-AUC of 89.77%, while RF delivered competitive stability. NN, SGD, and NB recorded lower metrics, showing limited ability to leverage CNN features. In conclusion, the hybrid CNN–Machine Learning framework effectively classifies alcoholic and non-alcoholic EEG spectrograms, demonstrating strong potential as an automated, accurate, and scalable tool for clinical screening.
Authors: Yuri Pamungkas, Myo Min Aung, Gao Yulan
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: diamond
OpenAlex ID: W7160891607