EpiEEG-60: A Balanced Epileptic Seizure and Normal EEG Dataset

Description

This dataset comprises raw EEG recordings from 60 individuals, including 30 neurologically normal subjects and 30 subjects diagnosed with epilepsy. All recordings were labeled by an expert neurologist from the National Institute of Neuroscience and Hospital (NINS). Participants include both male and female subjects, aged between 17 and 70 years. The dataset follows a structured directory hierarchy. The main folder, Ready_data, contains two subdirectories: Normal and Abnormal. Each subdirectory includes subject-wise folders (e.g., Patient_01, Patient_02, etc.), with each folder corresponding to a single EEG recording session. EEG signals were acquired using a Nihon Kohden EEG system. For each subject, seven associated files are provided: .EEG, .21E, .CMT, .CN2, .BFT/TRD, .EVT, and .LOG. The .EEG file contains continuous raw EEG waveform data, while the remaining files store acquisition parameters, channel configuration, electrode information, event annotations, trend data, and system logs required for accurate decoding and interpretation. The average recording duration is approximately 302 minutes per subject. Each recording contains 43 channels, including EEG, EOG, BIO, and miscellaneous reference channels. Signals were sampled at 500 Hz with a bandwidth of 0–250 Hz and stored as raw, unprocessed signals (RawNihon object type). Clinically annotated event markers—such as Eyes Open, Eyes Closed, Hyperventilation, Photic Stimulation, and Recording Start—are temporally aligned with the EEG signals and provided via event annotation files. These annotations support clinical interpretation and supervised learning tasks, including seizure analysis.

Institutions

• Bangladesh University of Business and Technology

  Dhaka District, Dhaka
• National Institute of Neurosciences & Hospital

  Dhaka, Sher-E-Bangla Nagar

Categories

Licence