Recurrence quantification analysis of EEG signals using a random forest classifier was investigated for automated epileptic seizure detection, though specific quantitative results are not available.
An automated EEG seizure detection system using RQA features and a Random Forest classifier achieves high accuracy and computational efficiency for bedside monitoring.
Epilepsy affects over 50 million people globally and its diagnosis relies heavily on visual inspection of electroencephalogram (EEG) recordings. This paper proposes an automated three-class EEG seizure detection system based on Recurrence Quantification Analysis (RQA), a nonlinear dynamical analysis technique that characterises the recurrence structure of phase space trajectories without stationarity assumptions. Six RQA features — Recurrence Rate (RR), Determinism (DET), Mean Diagonal Line Length (Lₘean), Shannon Entropy (ENT), Laminarity (LAM), and Trapping Time (TT) — are extracted from the recurrence plot of each EEG epoch. A Random Forest classifier trained under 10-fold stratified cross-validation on the Bonn University EEG database achieves an accuracy of 96. 67%, sensitivity of 96. 67%, specificity of 98. 33%, and F1-score of 96. 65% for three-class (normal, interictal, ictal) classification. The proposed method outperforms published RQA-SVM and DWT-SVM baselines and requires no prior signal decomposition, making it computationally efficient for bedside monitoring.
Kunhimangalam et al. (Mon,) conducted a other in Epileptic Seizures. Recurrence Quantification Analysis with Random Forest Classifier was evaluated on Automated seizure detection. Recurrence quantification analysis of EEG signals using a random forest classifier was investigated for automated epileptic seizure detection, though specific quantitative results are not available.