Thursday, January 15, 2026

Inferring Optical Tissue Properties from Photoplethysmography utilizing Hybrid Amortized Inference


Sensible wearables allow steady monitoring of established biomarkers resembling coronary heart price, coronary heart price variability, and blood oxygen saturation through photoplethysmography (PPG). Past these metrics, PPG waveforms include richer physiological data, as latest deep studying (DL) research reveal. Nonetheless, DL fashions usually depend on options with unclear physiological which means, making a rigidity between predictive energy, medical interpretability, and sensor design. We tackle this hole by introducing PPGen, a biophysical mannequin that relates PPG alerts to interpretable physiological and optical parameters. Constructing on PPGen, we suggest hybrid amortized inference (HAI), enabling quick, sturdy, and scalable estimation of related physiological parameters from PPG alerts whereas correcting for mannequin misspecification. In in depth in-silico experiments, we present that HAI can precisely infer physiological parameters beneath various noise and sensor situations. Our outcomes illustrate a path towards PPG fashions that retain the constancy wanted for DL-based options whereas supporting medical interpretation and knowledgeable {hardware} design.

Related Articles

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Latest Articles