NEUROSKY–EPI: The First Open Single–Electrode Epilepsy EEG Dataset with Context–Aware Modeling and Clinically Grounded Metadata

We introduce NEUROSKY–EPI, to our knowledge the first open single-electrode EEG dataset for epilepsy. The resource contains 25 patients and 2,032 labeled 1 s windows captured with an off-the-shelf NeuroSky MindWave Mobile 2 headset, accompanied by clinically interpretable metadata (seizure type, antiepileptic drugs, comorbidities, adherence, education, socioeconomic status). We propose a simple, reproducible evaluation for handy devices: train EEGNet using CHB–MIT windowed into 12,009 8 s segments from 24 patients on a single Fp1 channel to mirror the consumer lead, then evaluate directly on NEUROSKY–EPI band–power features. This practical setup yields ~60% test accuracy; augmenting inputs with a 16–dim context vector from an autoencoder that summarizes resting↔awake responsiveness improves performance to ~68%. Patient–level embeddings from EEGNet’s penultimate layer also show stronger unsupervised separability with context (KMeans: 54.17% → 66.67%). We will release de-identified data, code, and concise dataset/model cards to enable clinically grounded, low-compute epilepsy analytics and fairness-aware studies on single–electrode recordings.