Simulation-driven probe design and dual-wavelength optimization for wearable cerebral oxygenation monitoring

We present a simulation-driven framework that jointly optimizes probe geometry, source-detector separation, and dual-wavelength configuration for wearable continuous-wave functional near-infrared spectroscopy. Five candidate layouts built around a non-collinear 2 × 3 matrix probe were evaluated on a Monte Carlo eXtreme platform using cortical-shell contribution ratio and detector-fluence metrics, followed by Cramér-Rao lower bound (CRLB) analysis of all ten wavelength pairs drawn from {660, 760, 810, 850, 904} nm, with tissue absorption including an explicit water term and pairs ranked directly by total estimation uncertainty (σtotal) rather than a weighted composite score. The two-dimensional cross-coverage geometry with the 660/904 nm pair yielded the lowest total estimation uncertainty (σtotal = 0.0490 μM, cond(A) = 1.50) and remained stable under pathlength, noise-model, and fluence-cutoff perturbations. A geometry-defined motor-cortex region of interest on a public finger-tapping dataset provided pipeline validation, a Colin27 anatomical-atlas simulation supported the directional robustness of the geometry result, and a four-channel prototype demonstrated hardware feasibility. The principal contribution is a reproducible simulation-statistics-validation methodology; the 660/904 nm configuration is the framework's model-based prediction under the present tissue model, to be confirmed by future in-vivo wavelength comparison.