ENHANCING SPINAL CORD INJURY MANAGEMENT USING AI-BASED, NON-INVASIVE DETECTION OF AUTONOMIC DYSREFLEXIA

Abstract

This paper introduces an AI-driven solution for detecting and monitoring Autonomic Dysreflexia (AD) in individuals with spinal cord injuries. Current AD detection methods are often limited and lack noninvasive monitoring capabilities. Our proposed model overcomes this by combining skin nerve activity (SKNA) signals with a deep neural network (DNN). The DNN was meticulously trained on a curated dataset from controlled colorectal distension experiments, which induced AD events in rats with spinal cord surgery. The system demonstrates impressive performance, with an average classification accuracy of 93.9% ± 2.5% and a precision of 95.2% ± 2.1%. It also achieves a balanced F1 score of 94.4% ± 1.8% and a low average false-negative rate of 4.8% ± 1.6%, minimizing the chances of missing an AD event. The system's robustness was validated on unseen data, where it maintained its high accuracy and low false-negative rate, confirming its reliability and generalizability. This AI-powered solution marks a significant advancement in noninvasive, real-time AD monitoring, with the potential to greatly improve patient outcomes and AD management