Inferring Optical Tissue Properties from Photoplethysmography using Hybrid Amortized Inference

2025-10-02

AI Summary

This study addresses the inherent tension among prediction accuracy, clinical interpretability, and sensor design guidance in deep learning–based photoplethysmography (PPG) analysis. Specifically, it aims to robustly and physiologically interpretably invert tissue optical parameters—such as blood volume fraction and absorption coefficient—from PPG signals. To this end, we propose PPGen, a biophysically grounded generative model that embeds light transport physics into a deep generative architecture, ensuring physiological plausibility of inferred parameters; and Hybrid Amortized Inference (HAI), a variational inference framework integrating neural amortization with explicit uncertainty quantification and model-misspecification diagnostics. Experiments demonstrate that our approach significantly improves estimation accuracy and generalizability of key physiological parameters under diverse noise conditions and multi-sensor configurations. The method establishes a new paradigm for PPG-based noninvasive monitoring—rigorously grounded in optical physics while retaining clinical utility and sensor design insight.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Probabilistic Inference
• Computer Vision: Low Level & Physics-based Vision

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

Smart wearables enable continuous tracking of established biomarkers such as heart rate, heart rate variability, and blood oxygen saturation via photoplethysmography (PPG). Beyond these metrics, PPG waveforms contain richer physiological information, as recent deep learning (DL) studies demonstrate. However, DL models often rely on features with unclear physiological meaning, creating a tension between predictive power, clinical interpretability, and sensor design. We address this gap by introducing PPGen, a biophysical model that relates PPG signals to interpretable physiological and optical parameters. Building on PPGen, we propose hybrid amortized inference (HAI), enabling fast, robust, and scalable estimation of relevant physiological parameters from PPG signals while correcting for model misspecification. In extensive in-silico experiments, we show that HAI can accurately infer physiological parameters under diverse noise and sensor conditions. Our results illustrate a path toward PPG models that retain the fidelity needed for DL-based features while supporting clinical interpretation and informed hardware design.

Problem

• Inferring optical tissue properties from PPG signals
• Bridging predictive power and clinical interpretability gap
• Correcting model misspecification in physiological parameter estimation

Innovation

• Hybrid amortized inference for physiological parameter estimation
• Biophysical model linking PPG signals to optical properties
• Robust parameter inference under diverse noise conditions