Deep-learning-assisted thermogalvanic hydrogel fiber sensor for self-powered in-nostril respiratory monitoring.

Direct and consistent monitoring of respiratory patterns is crucial for disease prognostication. Although the wired clinical respiratory monitoring apparatus can operate accurately, the existing defects are evident, such as the indispensability of an external power supply, low mobility, poor comfort, and limited monitoring timeframes. Here, we present a self-powered in-nostril hydrogel sensor for long-term non-irritant anti-interference respiratory monitoring, which is developed from a dual-network binary-solvent thermogalvanic polyvinyl alcohol hydrogel fiber (d = 500 μm, L=30 mm) with Fe/Fe ions serving as a redox couple, which can generate a thermoelectrical signal in the nasal cavity based on the temperature difference between the exhaled gas and skin as well as avoid interference from the external environment. Due to strong hydrogen bonding between solvent molecules, the sensor retains over 90 % of its moisture after 14 days, exhibiting great potential in wearable respiratory surveillance. With the assistance of deep learning, the hydrogel fiber-based respiration monitoring strategy can actively recognize seven typical breathing patterns with an accuracy of 97.1 % by extracting the time sequence and dynamic parameters of the thermoelectric signals generated by respiration, providing an alert for high-risk respiratory symptoms. This work demonstrates the significant potential of thermogalvanic gels for next-generation wearable bioelectronics for early screening of respiratory diseases.