Bias-invariant, detectability-aware dual-distance optical reflectance sensing for layered biological media

Optical reflectance sensors are widely used for non-invasive tissue assessment but may be intrinsically biased by superficial absorbers, limiting reliability in layered media with variable surface composition. In skin-interfaced sensing, epidermal melanin is a dominant confounder that can systematically modulates detected reflectance and undermines equitable performance. Here we introduce a bias-invariant optical sensing strategy based on a detectability-aware dual-distance reflectance design that aims to suppress superficial absorption while preserving subsurface sensitivity. Using Monte Carlo photon transport simulations of layered skin, we show that conventional single-distance reflectance varies strongly with melanin, but a logarithmic near-far distance combination reduces melanin-absorption-induced variability by more than an order of magnitude. Then, by jointly analyzing bias suppression, subsurface sensitivity, and signal feasibility, we derive and propose design rules to identify buildable detector geometries near practical detectability limits. Although demonstrated here using epidermal melanin, the proposed framework can be generalized to apply to layered biological media to facilitate bias-invariant, depth-selective optical sensing under practical detectability constraints, with subcutaneous injection-site assessment in diabetes care presented as a motivating application.