Comparative Analysis of Microcontrollers in Wearable Devices for Real-Time Arrhythmia Classification: An Embedded Machine Learning Approach

In Bangladesh, cardiovascular diseases (CVDs) pose a significant health challenge, where the cost of treatment is a major barrier to widespread healthcare access. This study addresses this issue by focusing on wearable device-based arrhythmia pre-detection to reduce diagnostic costs. Designing such wearable devices requires choosing the best microcontroller unit (MCU) for real-time data gathering and inference. In this study, a motherboard prototype is developed to facilitate the comparative analysis of commonly available MCUs in Bangladesh: nRF52840, Raspberry Pi RP2040, STM32F446RE, and ESP32. Evaluating MCU performance parameters using a shallow neural network trained on locally obtained datasets reveals insightful findings. The nRF52840 stands out for its minimal power consumption, averaging 33.46 m W, making it an energy -efficient choice. The STM32F446RE demonstrates the fastest execution time, processing in an impressive 125 ms. Moreover, considering pricing factors, the nRF52840 offers a moderate cost option. This research not only aids in selecting the most suitable M CU for wearable devices but also provides a comprehensive understanding of the trade-offs involved. Findings enable the design of wearables balancing performance, power efficiency, and affordability, thereby enhancing accessible and cost-effective healthcare options for addressing cardiovascular disease challenges in remote and economically constrained environments like Bangladesh.