Categorisation of EEG suppression using enhanced feature extraction for SUDEP risk assessment

Area of Science:

• Epilepsy research
• Computational neuroscience
• Biomedical signal processing

Background:

• Sudden Unexpected Death in Epilepsy (SUDEP) is a significant concern in epilepsy management.
• Identifying biomarkers for SUDEP risk remains a challenge for the scientific community.
• The duration of postictal generalized EEG suppression (PGES) is a potential indicator of SUDEP risk, but its end is difficult to determine.

Purpose of the Study:

• To address the challenge of automatically marking the end of PGES in electroencephalogram (EEG) data.
• To evaluate the effectiveness of machine learning models in identifying PGES duration for SUDEP risk assessment.

Main Methods:

• A machine learning approach was developed to analyze EEG data from patients during seizures.
• Sensitivity analysis was performed on EEG window size, feature extraction (using pyEEG), and classifiers (Gradient Boosted Decision Trees, Random Forest).
• Ten EEG-based features were extracted and analyzed with five different window sizes.

Main Results:

• The Random Forest classifier achieved a maximum Area Under the Curve (AUC) score of 76.02%.
• Key features contributing to performance included SVD Entropy, Petrosan Fractal Dimension, and Power Spectral Intensity.
• The developed methods demonstrated effectiveness in automatically identifying the end of PGES.

Conclusions:

• The proposed methods can automatically determine the end of PGES, offering a potential tool for SUDEP risk stratification.
• Future research should focus on clinical integration of these methods for predicting SUDEP risk in patients.