Functional dual-slope frequency-domain near-infrared spectroscopy data interpreted with two- and three-layer models

Functional near-infrared spectroscopy (fNIRS) is impacted by signal contamination from superficial hemodynamics. It is important to develop methods that account for such contamination and provide accurate measurements of cerebral hemodynamics.

We aim to investigate whether simulated data with two- or three-layer tissue models are able to reproduce in vivo data collected with dual-slope (DS) frequency-domain (FD) near-infrared spectroscopy (NIRS) on human subjects during brain activation.

We performed Monte Carlo simulations to generate DS FD-NIRS data from two- and three-layer media with a range of layer thicknesses and optical properties. We collected in vivo data with DS FD-NIRS (source-detector distances: 25 and 37 mm, wavelengths: 690 and 830 nm, and modulation frequency: 140 MHz) over the occipital lobe of human subjects during visual stimulation. Simulated and in vivo data were analyzed with diffusion theory for a homogeneous medium, and results were compared for each DS FD-NIRS data type.

We found that the main qualitative features of in vivo data could be reproduced by simulated data from a three-layer medium, with a second layer (representing the cerebrospinal fluid in the subarachnoid space) that is less absorbing and less scattering than the other two layers, and with a top layer thickness that represents the combined scalp and skull thickness.

A three-layer model is a viable improvement over a homogeneous model to analyze DS FD-NIRS data (or any other fNIRS data) to generate more accurate measurements of cerebral hemodynamic changes without a need for large datasets for tomographic reconstructions.