INTERNET OF MEDICAL THINGS IN INPATIENT NEUROLOGY: CONTINUOUS MONITORING, AUTOMATED SEIZURE DETECTION, AND EARLY WARNING SYSTEMS

Abstract

The interval between the onset of neurological deterioration in a hospitalised patient and its recognition by clinical staff represents a window of preventable harm — a window that conventional nursing observation schedules, constrained by staffing ratios and cognitive load, are often too wide to close. The Internet of Medical Things, comprising connected physiological monitors, bedside EEG systems and wearable acquisition devices that transmit continuous data streams to central analytical servers, offers a technical means of narrowing this window substantially. Machine learning algorithms applied to these streams can perform pattern recognition tasks — detecting non-convulsive seizure activity, scoring real-time early warning scores, flagging haemodynamic instability — at a temporal resolution that intermittent human observation cannot match. A review of 33 studies published between 2018 and 2024 finds that IoMT-based multi-parameter early warning systems predict neurological intensive care escalation with AUC of 0.86–0.93, reduce time-to-intervention for deterioration events by a mean of 44 minutes, and detect non-convulsive status epilepticus with sensitivity of 88–96% at clinically acceptable false alarm rates. The potential relevance of these findings for inpatient neurology services in Uzbekistan, where nursing ratios and monitoring equipment availability impose real constraints on observation intensity, is discussed.