Monte Carlo-based correction of geometric and instrumental biases in single-distance time-domain near-infrared spectroscopy on skeletal muscle.

Time-Domain Near-Infrared Spectroscopy (TD NIRS) data from skeletal muscle are commonly analyzed using homogeneous models for photon diffusion, introducing substantial bias when the skeletal muscle is overlain by a thick superficial adipose tissue. We used Monte Carlo simulations of photon propagation in layered media to quantify these effects and to develop correction strategies for both homogeneous and two-layer analytical models. A total of 1500 single-source-detector distance TD NIRS curves with varying optical and geometrical properties were simulated and fitted with a homogeneous or layered model for photon diffusion. By comparison with the ground-truth absorption coefficients of the lower layer, correction curves were obtained. The role of the instrument response function was investigated, highlighting its impact on homogeneous and layered analyses. A simulated vascular occlusion test showed that corrections reduced the median absolute percentage error of recovered hemodynamic parameters from ≃ 30-40% to below 10% (homogeneous) and to ≃ 2-3% (two-layer). We tested the applicability of the proposed approach in-vivo via measurements on a volunteer during cuff occlusion on the leg.