Integrating Symbolic Regression and Photoplethysmography for Monitoring Blood Pressure Estimation

This paper advances non-invasive blood pressure (BP) monitoring by leveraging photoplethysmography signals, enhanced through the integration of symbolic regression (SR) and traditional machine learning techniques. Our novel methodology combines traditional SR-based and feature extraction methods, utilizing recursive feature elimination with cross-validation (RFECV) for optimal feature selection. Comparative analysis across extensive datasets shows that integrating SR with RFECV enhances model transparency and predictive accuracy, providing clinically interpretable mathematical expressions that improve our understanding of BP estimation dynamics, which is crucial for healthcare diagnostics.