PulmoAI: A Robust Deep Learning Framework for Contactless Respiratory and Cardiovascular Health Assessment Using rPPG

This paper presents PulmoAI, a state-of-the-art implementation of Remote Photoplethysmography (rPPG) designed to democratize health monitoring. By transforming standard consumer webcams into medical diagnostic devices, PulmoAI extracts physiological signals—specifically Heart Rate (HR), Oxygen Saturation (SpO2), Respiratory Rate (RR), and Heart Rate Variability (HRV)—from facial video without physical contact. The system leverages the rPPG-Toolbox, utilizing 3D Convolutional Neural Networks (PhysNet) to achieve clinical-grade accuracy (MAE ~0.8 BPM). Furthermore, PulmoAI integrates a novel Medical Expert System based on the National Early Warning Score (N.E.W.S.) to detect early signs of COPD, Asthma, and Pneumonia. This paper details the multi-tiered "Full-Proof" architecture that ensures reliability across varying computational environments, ranging from high-end GPU servers to standard CPUs.