EEG-based Seizures Detection with Decision Tree Classifier: A Performance Assessment using ROC Analysis and Feature Importance

In order to diagnose and track epilepsy, seizure detection using electroencephalogram (EEG) signals is essential. Using Power Spectral Density (PSD) features taken from EEG signals, this study proposes a decision tree-based classification model for automated seizure detection. The model performs robustly, as evidenced by its 92.04% accuracy, 76.18% precision, 87.61% recall, and 81.50% F1-score. According to feature importance analysis, some frequency bands have a big impact on how decisions are made. The effectiveness of the suggested approach is validated by a thorough evaluation that uses a confusion matrix, ROC curve (AUC = 0.90), and top feature importance rankings. The interpretability and clinical feasibility of decision trees for real-time seizure prediction systems are highlighted in this work. Additionally, the lightweight nature of decision trees allows them to be deployed on resource-constrained devices, making them suitable for portable or wearable EEG monitoring systems. The approach also enables clinicians to better understand and trust the model’s decision-making process due to its transparent structure. Furthermore, by emphasizing specific frequency components within EEG signals, this method may assist neurologists in identifying biomarkers associated with epileptic events, potentially contributing to personalized treatment planning and improved patient outcomes.