Ring-BP: Using a Wearable Smart Ring to Cufflessly Estimate Blood Pressure with Mobile and Efficient Net

Blood pressure (BP) is an important vital indicator of human health. Continuous monitoring of BP levels is an urgent demand worldwide, as continuous BP readings can provide richer information for the evaluation of cardiovascular health conditions. As the traditional cuff-based monitoring method is inconvenient and can cause pain to users, the photoplethes-mogoraphy(PPG) based blood pressure monitoring methods have attracted great attention because they can offer continuous blood pressure cufflessly. In this paper, we use the wearable smart ring PPG signal and efficient mobile network to monitor blood pressure cufflessly. We first train the model on the open-source VitalDB dataset, and then transfer the model to a self-collected RingConn-StaticBP dataset. Extensive experiments show that transfer learning technology, personalization calibration, demographic features, and BP_LEVEL feature can improve the accuracy of blood pressure estimation. Among VitalDB 185 test subjects, the best mean error (ME)±standard deviation error(SDE) for SBP is 1.432±8.091 mmHg, and ME±SDE for DBP are 0.731±4.565 mmHg, which almost meets the AAMI standard. After personalizing the model pre-trained with VitalDB, the ME±SDE is 0.978±4.773 mmHg for SBP, and 0.737±3.396 mmHg for DBP in the RingConn-StaticBP dataset. The results show the effectiveness and potential of continuously measuring blood pressure without cuffs using a wearable smart ring.