Functional near infrared spectroscopy using spatially resolved data to account for tissue scattering: A numerical study and arm-cuff experiment

Abstract

Functional Near-Infrared Spectroscopy (fNIRS) aims to recover changes in tissue optical parameters relating to tissue hemodynamics, to infer functional information in biological tissue. A widely-used application of fNIRS relies on continuous wave (CW) methodology that utilizes multiple distance measurements on human head for study of brain health. The typical method used is spatially resolved spectroscopy (SRS), which is shown to recover tissue oxygenation index (TOI) based on gradient of light intensity measured between two detectors. However, this methodology does not account for tissue scattering which is often assumed. A new parameter recovery algorithm is developed, which directly recovers both the scattering parameter and scaled chromophore concentrations and hence TOI from the measured gradient of light-attenuation at multiple wavelengths. It is shown through simulations that in comparison to conventional SRS which estimates cerebral TOI values with an error of ±12.3%, the proposed method provides more accurate estimate of TOI exhibiting an error of ±5.7% without any prior assumptions of tissue scatter, and can be easily implemented within CW fNIRS systems. Using an arm-cuff experiment, the obtained TOI using the proposed method is shown to provide a higher and more realistic value as compared to utilizing any prior assumptions of tissue scatter.

Keywords: near-infrared spectroscopy; tissue optics; tissue scattering.

Figure 1

A qualitative schematic showing the sensitivity of light‐attenuation in a three‐layered medium. The two detectors are modeled at 37 and 43 mm from the source

Figure 2

Sensitivity of attenuation with respect to depth for individual detectors as well as the gradient between them

Figure 3

Scatter plot of condition numbers calculated for E and J with (A) exponential and (B) linear model for scatter

Figure 4

Normalized histogram of optimal wavelengths that satisfy the low condition number criterion for (A) exponential and (B) linear model of scatter

Figure 5

Probe locations shown on a 3‐layered model of human head

Figure 6

Recovered tissue oxygenation index values of the brain with standard‐deviation of recovery across 10 head models, (A) homogeneous head model, (B) three‐layered head model, using both exponential (exp.) and linear (lin.) scattering models

Figure 7

Box‐plot of absolute error in the estimation of tissue oxygenation index (TOI), showing the distribution of errors across different ground‐truth TOI and scattering parameters for, homogeneous and three‐layered head model

Figure 8

Recovered b‐value (corresponding to linear scattering model) with Spectrally Constrained Spatially Resolved Spectroscopy for (A) homogeneous head models, and (B) 3‐layered head models, of 10 different subjects with different scattering parameters. The boxplot shows the distribution of recovered “b” for different cerebral tissue oxygenation index (50% to 80%). The solid line represents the ground‐truth values of the homogenous medium in figure (A) and of the cerebral region in figure (B). The dotted line marks the value assumed in spatially resolved spectroscopy method

Figure 9

Recovered tissue oxygenation index values with standard‐deviation of recovery across 30 different cases of randomly varying scattering properties, for a skin + skull thickness of (A) 0 mm, (B) 4 mm, (C) 8 mm and (D) 14 mm

Figure 10

Boxplot of absolute error in the estimation of tissue oxygenation index (TOI) for different skin + skull thicknesses, showing its distribution across different ground‐truth TOI and scattering parameters

Figure 11

Tissue oxygenation index measurement on human forearm: Rest arm recovered b‐value from SCSRS is 0.79 ± 0.007 μm⁻¹, and Occlusion arm recovered b‐value from Spectrally Constrained Spatially Resolved Spectroscopy is 0.82 ± 0.003 μm⁻¹