Unleashing the potential of spiking neural networks for epileptic seizure detection: A comprehensive review

Abstract

Epilepsy is a prevalent neurological condition characterized by repeated seizures resulting from structural or functional changes in the brain. Standard treatment approaches include anti-epileptic medications and surgical interventions. However, for patients with drug-resistant epilepsy, long-term EEG recordings are essential to monitor medication efficacy and make necessary adjustments to minimize seizure frequency. Continuous monitoring of such patients requires compact and comfortable wearable devices. This review explores the potential of Spiking Neural Networks (SNNs) for low-power signal processing in neuromorphic systems, specifically addressing the demand for compact and wearable devices. SNNs offer rapid computation, minimal power consumption, and superior data representation compared to traditional artificial neural networks (ANNs). With biological plausibility and event-driven hardware support, SNNs are well-suited for modeling neural activity in epilepsy. The review delves into existing literature, investigating the potential of SNNs in seizure detection. It covers various aspects, including the neuronal system, neuromorphic computing, SNN architecture, spiking neuron models, spike encoding schemes, and training methodologies. Representative studies utilizing spike-based machine learning for seizure detection are examined. Furthermore, the review addresses gaps, challenges, strengths, weaknesses, and trade-offs in SNN implementations. By spotlighting the capabilities of SNNs, this review aims to stimulate further research and innovative approaches, contributing to the advancement of epilepsy diagnosis and management.