A deep learning-assisted skin-integrated pulse sensing system for reliable pulse monitoring and cardiac function assessment

Long-term pulse monitoring plays a vital role in the prevention and diagnosis of cardiovascular diseases since pulse signals can provide rich characteristics of cardiac conditions. Flexible sensors mounting on the wrist for pulse monitoring have drawn great attention due to the advantages of soft, skin-interfaced and non-invasive features. However, slight wrist movement and transposition of the sensor on skin would dramatically impact the signal quality because of motion artifacts and interfacial properties. Here, we report an intelligent skin-integrated system by combining a sensitive pressure sensor array and a signal enhancement algorithm to provide reliable pulse signal monitoring. The sensors have the advantages of low cost, simple fabrication, reusability and high sensitivity. The multiple sensor channels offer rich information of vital signals, which can facilitate deep learning-based algorithms to denoise and reconstruct pulses more accurately and thus enable precise heart rate monitoring and heart variability assessments. The reported technology suggests a novel approach for pulse waveform detection in static and dynamic conditions, allowing users to more easily achieve reliable pulse monitoring.