EEG-based Early Detection of Epileptic Seizures

Early detection of seizure onsets is a critical task in bio-informatics that can significantly help seizure patients with a controlled lifestyle. Recent research leverages various features from patients’ electroencephalograms (EEG) for their seizure detection model, some of which also focus on detecting seizures using a reduced number of EEG channels; however, they use longer time intervals of EEG and thus miss the early detection of seizure onsets. Imbalanced, noisy and diverse EEG data for patients makes early seizure onset detection further challenging. In my research, the first research contribution provides early detection of focal epileptic seizure onsets. In our upcoming research, we plan to explore focal seizure propagation and detect the onset of fully developed generalized seizures by reducing the number of channels. During this, we aim to leverage knowledge distillation to aid easy-to-use wearable devices for seizure detection and control. Lastly, we aim to equip our model with security features such that this model is robust against data poisoning and impersonation.