An AI-Driven Interpretable Multiview Feature Learning Approach for EEG Based Epileptic Seizure Detection.

Area of Science:

• Neurology and Biomedical Signal Processing
• Artificial Intelligence in Healthcare

Background:

• Epilepsy is a chronic neurological disorder impacting quality of life and potentially causing irreversible brain damage.
• Electroencephalography (EEG) signal analysis is vital for monitoring epilepsy, enabling early seizure detection and intervention.
• Effective seizure detection relies on identifying interpretable features from EEG signals to improve clinical outcomes.

Purpose of the Study:

• To propose a novel interpretable multi-view feature learning (IMV-FL) approach for enhanced EEG-based seizure detection (ESD).
• To improve the accuracy and interpretability of seizure detection by integrating time and frequency domain features.
• To provide an effective tool for clinical and healthcare settings for epilepsy management.

Main Methods:

• Converted time-domain EEG signals to frequency-domain representations using Discrete Fourier Transform (DFT).
• Extracted spatial and temporal morphological features using ResNet and Long Short-Term Memory (LSTM) models, with feature compression via Deep Neural Network (DNN).
• Employed Mutual Information-Based Feature (MIBF) selection and a Stacking Ensemble Classifier (SAEC) for unified classification, enhanced by SHapley Additive exPlanations (SHAP) for interpretability.

Main Results:

• The proposed IMV-FL framework demonstrated superior performance compared to single-view methods, achieving an average 3% improvement.
• Outperformed state-of-the-art techniques by 2% in classification accuracy, sensitivity, specificity, and F1-score.
• Validated effectiveness on the CHB-MIT Scalp and Bonn EEG datasets.

Conclusions:

• The interpretable multi-view feature learning approach offers a significant advancement in EEG-based seizure detection.
• This method provides enhanced accuracy and clinical interpretability, crucial for effective epilepsy monitoring and management.
• The framework serves as a promising tool for improving patient outcomes in clinical and healthcare environments.