City Research Online

Abstract

Photoplethysmography is an optical technique that produces a wealth of information about cardiovascular health. Therefore, the technology has become an integral part of personal health monitoring devices. Given the importance of blood pressure measurement and control in physical and mental health, in recent years, the estimation of blood pressure from photoplethysmography has been an active area of research with promising results. Most studies on the subject rely on the morphological features of the photoplethysmogram. These features are highly prone to noise, changes in sensor placement, and skin properties; including skin colour. To address these limitations, we investigated the feasibility of using pulse rate variability features which are known to be less prone to the aforementioned limitations. To this end, we collected high quality photoplethysmograms using a bespoke, research-grade device from 18 healthy subjects. Approximately 15 min of photoplethysmograms and continuous blood pressure waveforms were collected from each subject. We trained machine learning models based on different feature sets and compared their performances. The model with morphological features alone outperformed the model with pulse rate variability features, root mean squared error (RMSE) of 6.32 vs 7.23 mmHg. However, the best performance was obtained using the combined set of features (RMSE: 5.71 mmHg). Combined, the evidence shows that the estimation of BP from PRV, alone or in conjunction with morphological features, is feasible. In light of the limitations of morphological features in estimation of blood pressure, our findings lend support to further research on the use of pulse rate variability features.

Publication Type:	Article
Additional Information:	© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Publisher Keywords:	Photoplethysmography, Pulse rate variability, Blood pressure, Features, Machine learning
Subjects:	R Medicine > RC Internal medicine T Technology > TK Electrical engineering. Electronics Nuclear engineering
Departments:	School of Science & Technology > Engineering
SWORD Depositor:	Symplectic Administrator

Export

Downloads