AI-enhanced flexible ECG patch for accurate heart disease diagnosis, optimal wear positioning, and interactive medical consultation
Xiaojiang Huang, Youlin Yuan, Jiming Liu, Jun He, Yunxiang Shi, Shuai Gao, Jun Wu, Xingjie Xu, Huiqing Zhang, Peng Li, Yao Yao, Wei Huang
Abstract
Continuous and reliable electrocardiogram (ECG) monitoring is crucial for the early diagnosis and intervention of heart diseases, which remain a leading threat to global health and mortality. Traditional ECG devices are often bulky, complex, and require hospital visits, limiting their practicality for daily use. To overcome these challenges, we have developed a wireless, flexible, and user-friendly ECG monitoring system integrated with advanced artificial intelligence (AI) capabilities. Our innovative ECG patch features an island-and-bridge serpentine structure, offering strain insensitivity of up to 100%, robust adhesion (7.6 kPa), and a high signal-to-noise ratio (28 dB). The accompanying mobile application leverages the interpretable attention transformer (IAT) model for heart disease diagnosis with up to 98% accuracy, a generative adversarial network (GAN) combined with convolutional neural networks (CNNs) and gated recurrent units (GRUs) for wear positioning correction with 85% accuracy, and GPT-based consultations with sub-second response times. This system enables real-time diagnosis, accurate wear positioning, and personalized medical advice, effectively bridging the gap between hospital care and at-home monitoring. Our work enhances accessibility to cardiac care, promotes early detection, and reduces the burden on healthcare systems.