An Efficient 1D CNN Architecture for Multi-Channel EEG Seizure Detection

Epilepsy is one of the most common neurological conditions, impacting more than 1% of the global population. Early and accurate seizure prediction is vital for enhancing patients' lives and has become a significant research focus. Despite this, existing approaches often overlook the intricate dynamics of multi-channel electroencephalography (EEG) signals, the primary tool for detecting seizures. In this work, we designed an efficient 1D CNN architecture tailored for EEG-based seizure detection on a multi-channel EEG system optimized for computational efficiency. It employs depthwise-separable convolutions to reduce model complexity while preserving temporal features of EEG signals. The proposed model is validated on CHBMIT seizure database over multi-channel analysis attains an accuracy, F1, precision, recall and specificity of 95%, 91%, 94%, 89% and 87% respectively. Our findings confirm that the proposed method delivers superior performance compared to other state-of-art models.