U-FaceBP: Uncertainty-aware Bayesian Ensemble Deep Learning for Face Video-based Blood Pressure Estimation

Researchers Yusuke Akamatsu, Akinori F. Ebihara, and Terumi Umematsu have introduced U-FaceBP, an uncertainty-aware Bayesian ensemble deep learning approach for estimating blood pressure from face videos. Unlike traditional methods that rely on specialized medical devices like cuffs, this system uses remote photoplethysmography (rPPG) to extract pulse waves from standard camera footage. The key innovation is its explicit modeling of aleatoric uncertainty (noise in the data) and epistemic uncertainty (model limitations) using a Bayesian neural network (BNN) ensemble. U-FaceBP combines three modalities: rPPG signals, PPG signals derived from face videos, and raw face images, each processed by separate BNNs whose outputs are fused for final BP estimation.

Validated on two large-scale datasets totaling 1,197 subjects from diverse racial backgrounds, U-FaceBP outperforms state-of-the-art BP estimation methods in accuracy and reliability. The uncertainty estimates are shown to be informative for guiding modality fusion, assessing prediction reliability, and analyzing performance across racial groups. Accepted by IEEE Transactions on Instrumentation and Measurement, this work addresses a critical challenge in contactless health monitoring: the inherent uncertainties in rPPG-based BP estimation. By quantifying when predictions are trustworthy, U-FaceBP could enable practical, camera-based health screening in telemedicine, clinics, and consumer devices, though real-world deployment would require regulatory approval and further validation.