Design of a Smart and Adaptive Wearable Device for Early Detection of Cardiovascular Abnormalities with Secure and Sustainable IoT Integration

Cardiovascular diseases are a leading cause of global morbidity and mortality, making early detection of abnormalities, such as arrhythmias, crucial for effective intervention. This paper presents a wearable device design utilizing Tiny Machine Learning (TinyML), optimized for resource-constrained devices, enabling on-board training and learning to reduce reliance on cloud processing and enhance real-time decision-making. The system combines heart rate, accelerometer, and environmental sensors for continuous monitoring. We compare three processing scenarios: cloud-based processing, traditional TinyML with pre-trained models, and on-board training and learning. While the cloud-based system achieves the highest initial accuracy (0.94), it faces significant latency and high-power consumption. The traditional TinyML model reduces latency but remains static, with an accuracy of 0.91. The proposed on-board training and learning system offers an initial accuracy of 0.88 but adapts over time, improving monitoring accuracy. This approach provides faster inference speeds (125 ms) and lower energy consumption than cloud-based solutions. Additionally, it reduces environmental impact by eliminating cloud reliance. The results highlight that on-board learning with TinyML offers personalized health monitoring, improved accuracy, low power usage, and enhanced security, making it a sustainable, energy-efficient, and secure solution for real-time healthcare monitoring.