Estimating Blood Pressure with a Camera: An Exploratory Study of Ambulatory Patients with Cardiovascular Disease

TL;DR

It is reported that facial rPPG yields a signal that is comparable to finger PPG, and the accuracy of BP prediction in subjects with atrial fibrillation was not inferior to subjects with normal sinus rhythm, highlighting the potential of rPPG for hypertension monitoring.

Abstract

Hypertension is a leading cause of morbidity and mortality worldwide. The ability to diagnose and treat hypertension in the ambulatory population is hindered by limited access and poor adherence to current methods of monitoring blood pressure (BP), specifically, cuff-based devices. Remote photoplethysmography (rPPG) evaluates an individual's pulse waveform through a standard camera without physical contact. Cameras are readily available to the majority of the global population via embedded technologies such as smartphones, thus rPPG is a scalable and promising non-invasive method of BP monitoring. The few studies investigating rPPG for BP measurement have excluded high-risk populations, including those with cardiovascular disease (CVD) or its risk factors, as well as subjects in active cardiac arrhythmia. The impact of arrhythmia, like atrial fibrillation, on the prediction of BP using rPPG is currently uncertain. We performed a study to better understand the relationship between rPPG and BP in a real-world sample of ambulatory patients from a cardiology clinic with established CVD or risk factors for CVD. We collected simultaneous rPPG, PPG, BP, ECG, and other vital signs data from 143 subjects while at rest, and used this data plus demographics to train a deep learning model to predict BP. We report that facial rPPG yields a signal that is comparable to finger PPG. Pulse wave analysis (PWA)-based BP estimates on this cohort performed comparably to studies on healthier subjects, and notably, the accuracy of BP prediction in subjects with atrial fibrillation was not inferior to subjects with normal sinus rhythm. In a binary classification task, the rPPG model identified subjects with systolic BP $\geq$ 130 mm Hg with a positive predictive value of 71% (baseline prevalence 48.3%), highlighting the potential of rPPG for hypertension monitoring.

Figures (5)

• Figure 1: rPPG Blood Pressure Estimation Pipeline. The rPPG signal is preprocessed frame-by-frame from the video. After preprocessing, five consecutive beats with good quality are selected by template matching. The selected beat sequence is encoded with a conformer and features embeddings are appended with embeddings from demographic and clinical history features. Finally, a multi-layer perceptron outputs the blood pressure estimate. Blood pressure measurements from the oscillometric cuff are used as ground truth data for model training, not for individual calibration. MLP = multi-layer perceptron, BMI = body mass index.
• Figure 2: Comparative Segments of PPG, Remote PPG, and ECG Across Different Heart Rhythms.
• Figure 3: Binary Classification of SBP. Classifying hypertensive vs non-hypertensive subjects defined as SBP $\geq$ or $<$ 130 mm Hg, respectively (based on the 2017 ACC/AHA BP guidelines for the general population.) The prevalence (or pre-test probability) of SBP $\geq$ 130 mm Hg in this cohort was 48.3%. The hybrid model produces the most accurate model with the highest sensitivity.
• Figure 4: Correlation and Bland-Altman Plots. Systolic blood pressure estimation results from face rPPG.
• Figure 5: Box Plots. A comparison of the PPG signal quality from our dataset to the UCI dataset from Kachuee et al. kachuee2015cuff.