iPPG 2 cPPG: Reconstructing contact from imaging photoplethysmographic signals using U-Net architectures.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Medical Imaging

Background:

• Remote vital sign monitoring using cameras, specifically imaging photoplethysmography (iPPG), is advancing.
• Current research aims to estimate blood pressure (BP) remotely, but limited data hinders progress.

Purpose of the Study:

• To develop a method for converting iPPG signals to contact photoplethysmography (cPPG) signals.
• To enable subsequent blood pressure estimation from iPPG data by leveraging large existing databases.

Main Methods:

• A two-stage approach: first, converting iPPG to cPPG using video datasets.
• Second, estimating BP from the converted cPPG signals using deep learning models (U-shaped architectures) on continuous wavelet transform (CWT) representations.

Main Results:

• Neural architectures accurately reconstructed cPPG signals from iPPG signals using CWT.
• The method demonstrated good agreement with conventional metrics and waveform estimators.
• This validates the feasibility of BP estimation from iPPG via cPPG conversion.

Conclusions:

• This work presents the first successful method for accurate cPPG reconstruction from iPPG signals, meeting pulse waveform criteria.
• The developed technique paves the way for reliable remote blood pressure estimation using video data.
• Future research will integrate BP estimation models trained on large datasets like MIMIC.