DrSVision: A Machine Learning Tool for Cortical Region-Specific fNIRS Calibration Based on Cadaveric Head MRI.

Functional Near-Infrared Spectroscopy is (fNIRS) a non-invasive neuroimaging technique that monitors cerebral hemodynamic responses by measuring near-infrared (NIR) light absorption caused by changes in oxygenated and deoxygenated hemoglobin concentrations. While fNIRS has been widely used in cognitive and clinical neuroscience, a key challenge persists: the lack of practical tools required for calibrating source-detector separation (SDS) to maximize sensitivity at depth (SAD) for monitoring specific cortical regions of interest to neuroscience and neuroimaging studies. This study presents DrSVision version 1.0, a standalone software developed to address this limitation. Monte Carlo (MC) simulations were performed using segmented magnetic resonance imaging (MRI) data from eight cadaveric heads to realistically model light attenuation across anatomical layers. SAD of 10-20 mm with SDS of 19-39 mm was computed. The dataset was used to train a Gaussian Process Regression (GPR)-based machine learning (ML) model that recommends optimal SDS for achieving maximal sensitivity at targeted depths. The software operates independently of any third-party platforms and provides users with region-specific calibration outputs tailored for experimental goals, supporting more precise application of fNIRS. Future developments aim to incorporate subject-specific calibration using anatomical data and broaden support for diverse and personalized experimental setups. DrSVision represents a step forward in fNIRS experimentation.