Surface-based image reconstruction optimization for high-density functional near-infrared spectroscopy

Diffuse optical tomography (DOT) enables mapping of functional near-infrared spectroscopy channel-based optical density changes to spatial images of oxy- and deoxyhemoglobin. Accurate reconstruction requires optimization for specific probe geometries. Although prior work focused on volumetric voxel reconstructions with grid arrays, here we examine high-density hexagonal arrays for surface-based reconstructions of the brain and scalp.

We evaluate measurement and spatial regularization, spatial basis functions, and reconstruction strategies to reduce crosstalk and improve localization. Both single-wavelength (indirect) and dual-wavelength (direct) approaches are compared.

Simulations with a white-noise model guided parameter optimization using image quality metrics. Resting-state data were augmented with synthetic hemodynamic response functions (HRFs) to incorporate real measurement variance into the parameter optimization pipeline, and results were validated with a ball-squeezing motor task.

Gaussian spatial bases reduced brain-scalp crosstalk but lowered contrast-to-noise ratio and increased localization error. Indirect hemoglobin reconstruction decreased oxy-deoxy crosstalk. Validation data showed strong, lateralized motor cortex activation contralateral to the active hand.

High-density hexagonal arrays enable accurate surface DOT reconstructions when optimized. Resting-state data augmented with synthetic HRFs provide an effective strategy for parameter selection, yielding localized activation with a high contrast-to-noise ratio.