Accurate low-delay QRS detection algorithm for real-time ECG acquisition in IoT sensors

• Published in: Internet of things (Amsterdam. Online) Vol. 31; p. 101537
• Main Authors: Kim, Sebin, Kim, Chaehyun, Yoo, Youngwoo, Kim, Young-Joon
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.05.2025
• Subjects: Electrocardiogram (ECG); Healthcare; Internet-of-Things; QRS complex detection; R-peak detection; Real-time; Wearable devices; Internet-of-Things; QRS complex detection; Real-time; Healthcare; Electrocardiogram (ECG); Wearable devices; R-peak detection
• ISSN: 2542-6605; 2542-6605
• DOI: 10.1016/j.iot.2025.101537

QRS detection is crucial for heart function diagnosis and sports science. This paper presents a real-time QRS detection algorithm designed for low-cost wearable embedded platforms, enabling novel applications such as closed-loop stimulation for acute diseases, precise monitoring in sports science, and home health monitoring. This algorithm locates the R-peak in real-time, with a mean delay of 0.405 s, throughout the MIT-BIH dataset. We achieve high accuracy with minimal compromise to computational power or delay, using a two-step, find and validate method. Initially, we identify potential QRS candidates by detecting zero-crossing points through filtering and convolution processes. Next, we validate these candidates by comparing them with previous R-R intervals (RRI), adaptively comparing values to minimize T-wave errors and reject adjacent noise components. We introduced a novel algorithm based on RRI periodicity, simplifying the computational load while enhancing detection accuracy. By using the MIT-BIH dataset, we detected the QRS complexes with a 99.75% accuracy. Furthermore, we embedded the algorithm into an Arm Cortex-M4 microcontroller unit (MCU) with a 64 MHz clock, maintaining identical accuracy. We demonstrate live-stream QRS detection by generating MIT-BIH waveforms using a function generator and processing them with the MCU's on-chip 10-bit analog-to-digital converter (ADC), achieving 99.71% accuracy. Finally, we validate our work with a miniaturized flexible electrocardiogram (ECG) sensor in a form factor of a bandage, wirelessly linked to a smartwatch for real-time ECG monitoring and R-peak detection. A cloud connectivity network is established concluding that this work is suitable for practical monitoring applications.