Accuracy of Smartphone-based Vital Monitoring Using Remote Photoplethysmography Technology Enabled WellFie application

Abstract

Background: Remote health monitoring technologies gained interest in the context of COVID-19 pandemic with potential for contactless monitoring of clinical patient status. Here, we examined whether vital parameters can be determined in a contactless manner using a novel smartphone-based technology called remote Photoplethysmography (rPPG) and compared with comparable certified medical devices. Methods: We enrolled a total of 150 normotensive adults in this comparative cross-sectional validation study. We used an advanced machine learning algorithm in the WellFie application to create computational models that predict reference systolic, diastolic blood pressure (BP), heart rate (HR), and respiratory rate (RR) from facial blood flow data. This study compared the predictive accuracy of smartphone-based, rPPG-enabled WellFie application with comparable certified medical devices. Results: When compared with reference standards, on average our models predicted systolic blood pressure (BP) with an accuracy of 93.94%, diastolic BP with an accuracy of 92.95%, HR with an accuracy of 97.34%, RR with accuracy of 84.44%. For the WellFie application, the relative mean absolute percentage error (RMAPE) for HR was 2.66%, for RR was 15.66%, for systolic BP was 6.06%, and for diastolic BP was 7.05%. Conclusion: Our results on normotensive adults demonstrates that rPPG technology-enabled Welfie application can determine BP, HR, RR in normotensive participants with an accuracy that is comparable to clinical standards. The Wellfie application with its integrated advanced machine learning algorithm outperformed other approved wearable devices (such as Apple watch, Fitbit charge 2, and Microsoft Band) for HR estimation. WellFie smartphone application based on rPPG technology offers a convenient contactless video-based remote solution that could be used in any modern smartphone.