AI-Assisted Real-Time PPG Signal Analysis for Preliminary Cardiovascular Diagnosis with Explainable AI in Clinical Practice

Cardiovascular diseases are the reason of concern in todays world. As per the recent surveys it is due to the people's ignorance of their health, due to high time taking in the diagnosis using the existing invasive technologies. In this work, an early-stage heart problem detection using a SHAP framework based explainable AI and video PPG signals based system is presented. The proposed system uses the Mediapipe framework with clinical grade vPPG based region of interest (ROI) tracking where heart rate extraction is 94.3% accurate at 18-23ms latency. Additionally, the system applies XGBoost for disease classification while SHAP improves model interpretability and provides real time feature control and explanation. Output is a annotated real time video and explanation of report as a final result. The aim of this system is to propose modular systems which are light and handheld and enhance model interpretability and feature relevance, thereby improving trust and clinical usability of the model.