Detection and Early Detection of Pathological Motor Evoked Potentials: A Novel Approach for Intraoperative Neuromonitoring Assistance

Intraoperative monitoring of motor evoked potentials (MEPs) plays a crucial role in detecting potential neurological complications during surgical procedures. Traditional methods rely on real-time human interpretation, which can introduce delays or, worse, errors in response. This study proposes an early detection framework for pathological MEPs, leveraging an anomaly detection approach based on an extension of isolation forests-one of the most effective anomaly detection models. By modeling the temporal evolution of MEPs, our approach not only identifies pathological deviations but also anticipates their occurrence, enabling proactive intervention. We validated our methodology on a dataset of intraoperative MEP recordings, demonstrating improved detection performance and enhanced early warning capabilities compared to conventional approaches. The findings suggest that this framework may support real-time intraoperative monitoring and serve as a foundation for future decision-support tools, although prospective clinical validation is still required before any deployment can be considered.