Computational validation of optical adaptive depth steering for continuous-wave fNIRS.

Abstract

Continuous-wave functional near-infrared spectroscopy (CW-fNIRS) traditionally utilizes fixed-geometry optode configurations, which introduce significant anatomical bias due to inter-subject variations in skull thickness and phenotypic barriers such as dense hair. To address these limitations, we propose an Optical Adaptive Depth Steering (O-ADS) framework that mathematically synthesizes a "Virtual Optode" using a high-density multi-distance array (8-32 mm) and a linearly constrained minimum variance (LCMV) spatial beamformer. We computationally validated this framework using high-fidelity Monte Carlo simulations on layered cranial slabs and a 3D anatomical atlas (Colin27). Results demonstrate that O-ADS exhibits high computational resilience to anatomical bias, achieving up to a 39-fold improvement in relative brain sensitivity for thick-skull morphologies compared to traditional 32 mm static sensors. Furthermore, under realistic spatiotemporal noise conditions involving heterogeneous scalp hemodynamics and simulated phenotypic barriers (8 mm reference channel failure), O-ADS maintained a significantly higher mean signal recovery purity (38.3% ± 22.1%) than standard Multi-Distance Regression (MDR) (21.8% ± 15.4%). This software-defined approach provides a scalable solution for optical neuroimaging, reducing the dependence of sensor performance on individual anatomical and phenotypic variance to facilitate more inclusive cerebral monitoring under challenging physical constraints.