Multivariate EEG Dataset for Seizure Classification

This dataset comprises preprocessed electroencephalography (EEG) segments derived from the Neural Engineering Data Consortium curated Temple University Hospital EEG Seizure Corpus (TUSZ). It has been systematically prepared to facilitate supervised machine learning. The dataset is restricted to three clinically relevant classes: bckg (background, non-seizure EEG activity), fnsz (focal non-specific seizure), and gnsz (generalized seizure). Each instance corresponds to a standardised 5-second EEG segment formatted for deep learning workflows, ensuring consistency in temporal resolution and input structure. The original corpus was developed and released by Temple University’s Neural Engineering Data Consortium (NEDC) and is publicly accessible.