Continuous blood pressure prediction from pulse transit time using ECG and PPG signals

Abstract

High blood pressure (BP) is the most common cause of death and disability in the world, and is the largest contributor to heart and kidney disease. Current methods of measuring and monitoring blood pressure require either invasive procedures or intermittent inflation of a cuff to restrict blood flow. Thus a non-invasive method for continuous blood pressure monitoring is needed. Pulse transit time (PTT), has been reported to be highly correlated with blood pressure but data examining the effect of posture and activity on PTT based BP estimation are very limited. In this paper, PTT was computed using the windowed correlation between ECG and PPG signals. Continuous blood pressure was estimated using a previously published linear regression model. In fourteen healthy subjects, BP was estimated using PTT in 5 different positions (recumbent, seated, standing, walking, cycling) for each subject according to a preset protocol. Accuracy was increased when sparsified, preprocessed PPG signals were used. Furthermore, the observed errors of PTT measurement were within 1% of manual PTT measurement. The Root-Mean-Squared Errors (RMSE) in systolic and diastolic blood pressure between the reference standard oscillometric cuff-based device and the estimated BP from PTT were lowest when seated or standing and highest when walking or cycling. The mean difference ±standard deviation (SD) of the difference between the PTT-based estimated systolic BP and the reference standard was 0.07±5.8 mmHg in the seated position; however, this increased to 4.4±20.9 and 10.2±16.0 when walking and cycling respectively. Therefore, PTT-based BP estimation was reasonably accurate while stationary but not during motion and further improvements in estimation are required before its use for the estimation of ambulatory BP.