Large-area stretchable dynamic 18-lead ECG monitoring patch integrated with deep learning for cardiovascular disease diagnosis

In the field of clinical cardiovascular diseases (CVDs), the 18-lead electrocardiogram (ECG) is seen as more comprehensive than that of the conventional 12-lead. However, the 18-lead acquisition system is bulky and involves extensive electrode use and intricate wiring, which limits its portability and widespread application, leading to a dearth of dynamic systems and datasets. Here, we develop a wireless, large-area, wearable ECG system designed to capture 18-lead ECGs. This system replaces rigid electrodes and hanging wires with soft, breathable patches that offer excellent adhesion and electrical stability, enabling high-fidelity ECG capture even in various interference scenarios. Compared to commercial devices with gel electrodes, the groundbreaking system matches their ECG recording, signal analysis, signal-to-noise ratio, and diverse physical profile applications. Leveraging deep learning, we designed the Deep Multi-Scale Attention Network (DMSANet), which accurately diagnoses 15 cardiac conditions (average F1 score: 0.896), excelling across 5 tasks on the PTB-XL dataset.