AI-Driven Anomaly Detection in Graphene-Based Wearable Sensors for Real-Time Remote Patient Monitoring
Sundarraj Saranya, CH Hussaian Basha, S. Vasanth, M. Ragul Vignesh, N.N. Baalakumar, M. Saravanan · 2025
Graphene based sensors have contributed to the wearables movement by providing exceptional electrical, mechanical, biocompatible properties, with harmonic changes in technology's ability to continually monitor a patient's health in real time. This paper presents a novel framework of integrating artificial intelligence (AI) driven anomaly detection mechanisms on graphene based wearable devices for accurate and fast anomaly detection of abnormal physiological events. Because the system has high sensitivity of graphene sensors to the vital signals (ECC,$\text{SpO}_{2}$and temperature), the acquired signals have minimal signal loss and precise data acquisition. We try to train an embedded lightweight deep learning model from the labeled physiological datasets on the deep learning model such that on device inference can be done on the embedded edge device with real time response and limited latency. In addition, techniques of adaptive learning are used to learn anomaly thresholds personalized with respect to patientbased baselines to minimize false positives. Together, the devised system not only solves the problems of power consumption, data privacy and device scalability, but also the problem of early warning by providing means for remote or home patients care. The framework is extensively simulated and prototyped and the detection accuracy of more than 95% is demonstrated across many clinical cases.