An implementation of an AI-assisted sonification algorithm for neonatal EEG seizure detection on an edge device

Fast and accurate seizure detection is a challenging problem for neonates. This is due to a severe shortage of specialized medical professionals for EEG analysis, especially in disadvantaged communities. Fast artificial intelligence (AI) techniques have been proposed to compensate for this lack of expertise. However, such models lack explainability, which is a key feature for these models to be adopted by clinicians. AI-assisted sonification adds additional explainability to any such automated methodology, empowering the medical professional to take accurate decisions regardless of the level of expertise in EEG analysis. The feasibility of an implementation of such an algorithm on an edge device is presented and analyzed. A lightweight derived algorithm for resource-constrained implementation scenarios is also evaluated and presented, suggesting suitability for further ultra-low power, mobile and wearables implementations.