Biomechanical piezoelectric sensor design with explainable transformer analytics for early disease detection and patient monitoring

This paper presents a smart healthcare monitoring system integrating wearable piezoelectric biosensors, self-powered energy harvesting, transformer-based deep learning, explainable AI, adaptive patient learning, and telehealth connectivity for continuous remote patient monitoring. The framework supports preventive healthcare, early disease detection, and real-time clinical decision-making. The primary objective is to develop and validate a scalable healthcare platform capable of accurate physiological monitoring, autonomous operation, predictive analytics, patient-specific adaptation, and rapid telemedical response in real-world conditions. The proposed multilayer architecture includes wearable sensing nodes, signal preprocessing modules, wireless communication interfaces, transformer-based temporal analytics, explainable AI components, adaptive patient modeling, and cloud-enabled telehealth dashboards. Piezoelectric sensors are evaluated for pulse monitoring, respiration tracking, joint bending, and walking activities. Performance is assessed using diagnostic, computational, and usability metrics. Experimental results demonstrate 96.9% operational efficiency, 98.8% communication reliability, and 1.43-second system latency. Sensor outputs range from 1.82 V during pulse monitoring to 6.94 V during walking, while 98.7% sensor functionality is retained after 10,000 operational cycles. Signal preprocessing improves the signal-to-noise ratio from 14.2 to 31.8 dB. The transformer model achieves 97.9% accuracy, 97.4% precision, 97.1% recall, and 0.984 ROC-AUC, demonstrating high accuracy, scalability, interpretability, and clinical applicability.