Thermo-responsive and phase-separated hydrogels for cardiac arrhythmia diagnosis with deep learning algorithms

• DOI: 10.1016/j.bios.2025.117262
• Publication Date: 2025-02-14T21:12:06Z
• Authors: Hui Chen; Jian Zhou; Huan Cao; Dongfang Liang; Lei Chen; Yuanfan Yang; Lu Wang; Jianfei Xie; Huigao Duan; Yongqing Fu
• Keywords: Cardiac arrhythmia diagnosis; ECG; Temperature; Hydrogels; Arrhythmias, Cardiac; Biosensing Techniques; Switchable adhesion; Hydrogel; Wearable Electronic Devices; Electrocardiography; Deep Learning; AI; Humans; Electrodes; Algorithms

Adhesive epidermal hydrogel electrodes are essential for achieving robust signal transduction and cardiac arrhythmia diagnosis, but detachment of conventional adhesive dressings easily causes secondary damage to delicate wound tissues due to lack of programmable capability of changed adhesion. Herein, we developed hydrogel-based skin-interfacing electrodes capable of on-demand programmable adhesion and detachment to capture electrocardiogram signals for diagnosing cardiac arrhythmia. This was achieved by integrating dynamic multiscale contact and coordinated regulation through temperature-mediated switchable hydrogen bond interactions in phase-separated smart hydrogels. Through micro-scale regulation of adhesive molecules and meso-scale modulation of the modulus, the hydrogel electrodes can be rapidly transited between a slippery state (adhesion ∼1.3 N/m) and a sticky one (adhesion ∼110 N/m) during its peeling from skin. This achieves an 84.5-fold increase of on/off adhesive energy (or reducing the adhesion at the skin interface by 98%) at low temperatures compared to normal skin temperature. A real-time cloud platform was developed by integrating hydrogel electrodes, enabling remote electrocardiogram (ECG) monitoring. For clinical applications, such developed skin sensing platform effectively captured cardiac activities in patients with eight common arrhythmias, achieving by the recorded high-fidelity and analyzable electrical signals. With the assistance of deep learning algorithms, we demonstrated a wearable cardiac arrhythmia intelligent diagnosis system which enables real-time conversion of the collected ECG data into diagnostic evaluations with a recognition accuracy of 98.5%.