A channel-wise attention-based representation learning method for epileptic seizure detection and type classification

Epilepsy affect almost 1% of the worldwide population. An early diagnosis seizure types is a crucial patient-dependent step for treatment selection process. The proper relies on correct identification type. As such, identifying type has biggest immediate influence therapy than detection, reducing neurologist’s efforts when reading and detecting seizures in EEG recordings. Most existing detection classification methods are conceptualized following schema thus fail to perform well with unknown cases. Our work focuses patient-independent pays more attention explainability underlying mechanism our method. Using channel-wise mechanism, quantification channels contribution enabled. Therefore, results become interpretable visualization brain lobes by allowed. We evaluate model CHB-MIT recently released TUH Seizure, respectively. able classify 8 an accuracy 98.41%, directly from raw data without any preprocessing. A case study showed high correlation between neurological baselines model.