Wearable PCG-Based Edge Detection of Heart Valve Disorders Using Tiny Physics-Informed Neural Networks
IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 2515317-2515317
DOI: 10.1109/tim.2026.3716478 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Ngoc-Dau Mai und 2 weiteren Autor:innen. Die Arbeit nennt 69 Literaturangaben. Gefördert von: Research Grant of Pukyong National University.
Abstract (Originalsprache)
The accurate and timely detection of heart valve disorders (HVDs) is essential for reducing cardiovascular morbidity and mortality, yet the conventional diagnostic methods rely on clinical settings, bulky equipment, and expert interpretation, limiting their use for continuous monitoring and real-time decision-making. This study proposes a compact wearable system for real-time HVD detection that integrates a phonocardiogram (PCG)-based sensing platform with embedded tiny physics-informed neural networks (TinyPINNs). The system introduces four key components: 1) a lightweight chest-mounted device for continuous, noninvasive acquisition of high-fidelity PCG signals; 2) a singular superlet transform (SST)-based time–frequency representation for generating high-resolution spectrograms; 3) an ultraefficient TinyPINN architecture combining convolutional learning with physics-informed constraints for fully on-edge cardiac classification with low latency and no reliance on external computing infrastructure; and 4) integration into the Internet of Medical Things (IoMTs) framework for secure data handling, remote access, and scalable healthcare deployment. Validation on two public datasets, including the PhysioNet/CinC Challenge 2016 Dataset, achieved 91.87 % accuracy under leave-one-out cross-validation (LOOCV) and 93.94 % under 10-fold cross-validation, together with high sensitivity, specificity, precision, and $F1$ -score, while substantially reducing computational complexity compared with the state-of-the-art models. These results demonstrate the effectiveness of physics-informed edge intelligence for cardiac diagnostics and highlight the potential of the proposed system as a portable, energy-efficient, and scalable solution for continuous monitoring, personalized cardiovascular care, telemedicine, and early disease intervention.
Förderung
- Research Grant of Pukyong National University
Zitieren
APA
Mai, N., Ullah, K., & Chung, W. (2026). Wearable PCG-Based Edge Detection of Heart Valve Disorders Using Tiny Physics-Informed Neural Networks. IEEE Transactions on Instrumentation and Measurement, 75, 2515317-2515317. https://doi.org/10.1109/tim.2026.3716478
Vancouver
Mai N, Ullah K, Chung W. Wearable PCG-Based Edge Detection of Heart Valve Disorders Using Tiny Physics-Informed Neural Networks. IEEE Transactions on Instrumentation and Measurement. 2026;75:2515317-2515317. doi:10.1109/tim.2026.3716478
BibTeX
@article{Mai2026,
author = {Mai, Ngoc-Dau and Ullah, Kaleem and Chung, Wan-Young},
title = {Wearable PCG-Based Edge Detection of Heart Valve Disorders Using Tiny Physics-Informed Neural Networks},
journal = {IEEE Transactions on Instrumentation and Measurement},
year = {2026},
volume = {75},
pages = {2515317--2515317},
doi = {10.1109/tim.2026.3716478}
}
EUVIMED