Theoretical Analysis of CNNs for Automatic Seizure Detection in EEG Signals
Date:
Poster presentation of honors undergraduate thesis research. Demonstrated the 1D CNN architecture for EEG seizure detection, Butterworth filtering pipeline, Lipschitz stability analysis, and frequency domain interpretation of learned features (22 Hz beta-wave band). Results: 97% accuracy, 0.99 AUC on the University of Bonn EEG dataset.