Abstract

The growing prevalence of chronic conditions, such as dementia, diabetes, and hypertension, driven by an ageing global population, together with increasing demand for personalised healthcare, has significantly accelerated the development of everyday wearable sensor technologies. These devices, playing a crucial role in preventive medicine, chronic disease management, and personalised healthcare and typically equipped with accelerometers, gyroscopes, GPS, and magnetometers, enable continuous, real-time monitoring of physiological parameters, including cardiovascular activity, body temperature, blood pressure, metabolic indicators, oxygen saturation, posture, and gait, as well as biochemical markers derived from body fluids. However, the vast volume and complexity of data generated by wearable sensors pose significant challenges in terms of interpretation, decision-making, and clinical integration.

Large Language Models (LLMs), such as those based on transformer architectures, offer new opportunities to enhance the intelligence and usability of wearable medical systems. By combining wearable sensing technologies with LLMs, healthcare systems can move beyond raw data analysis toward intelligent, context-aware, and human-centric medical decision support. The integration of wearable medical sensors with LLMs represents a significant step toward intelligent, personalised, and patient-centric healthcare. While wearable sensors provide continuous and rich physiological data, LLMs add powerful reasoning, contextual understanding, and natural language interaction capabilities. Together, they have the potential to revolutionise remote monitoring, clinical decision-making, and preventive healthcare, provided that ethical, technical, and regulatory challenges are carefully addressed. LLMs can transform complex sensor data into clinically actionable insights by integrating physiological signals with patient history, medical guidelines, and contextual information.

This research explores opportunities (e.g. interpretation, reasoning, guidance, personalised recommendations) and challenges (e.g. data overload, communication with a corpus of LLMs, integration with patient history, ethics, privacy, security) of integrating LLMs into wearable sensor technologies in remote patient monitoring.