A computational framework to analyze time and location-dependent NIRS readouts for detecting cytochrome c oxidase

Sepsis is responsible for over 50% of hospital deaths. Changes in mitochondrial redox state can be indicative of cellular and organ function. Therefore, a method to evaluate mitochondrial utilization of oxygen continuously is critical. Cytochrome-c oxidase (CCO) is a mitochondrial enzyme that participates in oxidative phosphorylation and interacts with near-infrared light, potentially yielding an optical indicator of cellular metabolism. In this work, we use a computational framework to study the feasibility of utilizing photoplethysmography (PPG) signals for detecting CCO when HbO2 and Hb are present. We use a 3D Monte Carlo model of light absorption and transport in tissue to generate optical readouts in the form of temporal PPG signals corresponding to different physiological states. Furthermore, a machine learning model is trained to predict the CCO redox state from PPG signals containing different CCO and hemoglobin species concentrations.