EEG-based seizure detection using Random Forest for wearable applications

Huanca J.Z.; Taco G.C.; Suarez J.T. · Przeglad Elektrotechniczny (2025) · DOI: 10.15199/48.2025.09.42

Resumen

This study presents an efficient way to analyze EEG signals for seizure detection, designed for mobile devices in resourcelimited areas such as Arequipa, Peru. Using the CHB-MIT Scalp EEG database and focusing on five EEG channels, the pipeline integrates preprocessing, feature extraction, dimension reduction and classification using the Random Forest algorithm. The proposed system achieves an accuracy of 91.05%, a sensitivity of 92.08% and a specificity of 90.05%, demonstrating its reliability in identifying seizure events. To enhance interpretation, visual representations of each processing step and the entire workflow have been developed. The lightweight design ensures compatibility with wearable devices, providing a practical and cost-effective solution for continuous seizure monitoring in remote areas.

Instituciones: Universidad Nacional de San Agustín de Arequipa