VidBP: Detecting Blood Pressure from Facial Videos with Personalized Calibration

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Computer Vision

Background:

• Facial video analysis of blood volume pulse (BVP) shows correlation with blood pressure (BP).
• Individual variations in BVP-to-BP mapping limit generalizability of existing methods.
• A need exists for personalized, non-invasive blood pressure monitoring solutions.

Purpose of the Study:

• To develop and validate VidBP, a novel blood pressure detection system using facial BVP.
• To address inter-individual differences in BVP-BP relationships through personalized calibration.
• To offer a convenient and cost-effective approach for home blood pressure monitoring.

Main Methods:

• VidBP utilizes a pre-trained model to extract BP-related features from facial BVP.
• Personalized calibration involves creating a subject-specific dictionary of BVP features and BP labels.
• Blood pressure estimation is performed by comparing current BVP features to the personalized dictionary.

Main Results:

• VidBP demonstrated significantly lower blood pressure estimation errors compared to state-of-the-art methods.
• Performance was validated on 640 samples across 16 subjects.
• Personalized calibration enhanced the accuracy of capturing individual BVP-BP mappings.

Conclusions:

• VidBP offers a feasible and accurate method for estimating blood pressure from facial videos.
• The system's personalized calibration is a key advantage for individual BP monitoring.
• This technology holds potential for convenient, non-invasive, and cost-effective home blood pressure management.