Principal Investigator: Prof. Taous-Meriem Laleg-Kirati

Poster Presenter: Jiahao HU


Uncertainty quantification for EEG data




Electroencephalography (EEG) has proven valuable in detecting seizures owing to its impressive temporal resolution. The advancements in deep learning offer a robust means to investigate the temporal and spatial relationships among EEG channels. While numerous studies have applied deep learning to address seizure detection, they often overlook uncertainty information. In this paper, we introduce a framework featuring three uncertainty estimation networks for the seizure classification task. Our model enables Bayesian detection of seizures and supplies uncertainty details regarding model predictions. We have conducted extensive experiments to substantiate the effectiveness of our approach.