BRIDGING EFFICIENCY AND INTERPRETABILITY: EXPLAINABLE EDGE AI FOR CLINICAL DECISION-MAKING IN MEDICAL TELEMETRY

Abstract

The increasing adoption of medical telemetry and remote patient monitoring has transformed healthcare by enabling continuous observation of physiological signals in real-world environments. In this context, Edge Artificial Intelligence (Edge AI) has emerged as a promising approach for processing data locally, reducing latency and dependence on cloud infrastructures, particularly in resource-constrained settings. However, most edge-based models operate as black boxes, limiting their transparency and hindering clinical adoption. This study presents a systematic review, following PRISMA guidelines, of Explainable Artificial Intelligence (XAI) methods applied to Edge AI systems in medical telemetry. The analysis highlights a persistent trade-off between computational efficiency and interpretability, with most approaches favoring performance over explainability. From a clinical perspective, the findings emphasize that telemetry data alone is insufficient to support medical decision-making, requiring integration with clinical reasoning and contextual patient information. Additionally, the study discusses the risks associated with algorithmic and cognitive biases, as well as the challenges posed by limited infrastructure and professional training, particularly in developing healthcare systems. The results reveal a critical gap in the development of integrated, lightweight, and interpretable models suitable for real-world deployment. Addressing this gap is essential for advancing trustworthy, scalable, and clinically relevant artificial intelligence in healthcare.