PPG-based Fluid Loading Prediction
Jan 1, 2024
·
1 min read
Overview
Developed a non-invasive AI system that predicts fluid loading responsiveness using photoplethysmography (PPG) signals.
Key Results
- Patient-level averaging: AUC 0.75
- 5-model ensemble: AUC ~0.85
- 7-model ensemble: AUC ~0.86
Clinical Collaboration
Conducted in partnership with Severance Hospital under IRB approval.
Authors
PhD Candidate
I develop AI-driven non-invasive biomarker prediction systems using
photoplethysmography (PPG) signals, bridging computational fluid dynamics
and deep learning for clinical diagnostics. My recent work on physics-integrated
blood viscosity assessment was published in Computer Methods and Programs
in Biomedicine (2025).