Abstract
Millimeter-depth sensitivity with frequency domain near-infrared spectroscopy has been challenging due to the breakdown of the diffusion equation for source-detection separations < 1cm. To overcome this challenge, we employ a Monte-Carlo lookup table-based inverse algorithm to fit small separation (3-6 mm) frequency-domain near-infrared spectroscopy (FDNIRS) data for absorption and reduced scattering coefficients. We verify this small separation FDNIRS method through a series of in vitro and in vivo studies. In vitro, we observed a root mean squared percent error (RMSE) in estimation of the reduced scattering coefficient and absorption coefficient of 2.8% and 7.6%, respectively, in liquid phantoms consisting of Intralipid and Indian ink, and a RMSE in estimation of oxygen saturation and total hemoglobin concentrations of 7.8 and 11.2%, respectively, in blood-mixed liquid phantoms. Next, we demonstrate one particularly valuable in vivo application of this technique wherein we non-invasively measure the optical properties of the mouse brain (n = 4). We find that the measured resting state cerebral oxygen saturation and hemoglobin concentration are consistent with literature reported values, and we observe expected trends during a hyper-/hypoxia challenge that qualitatively mimic changes in partial pressure of oxygen (pO2) measured simultaneously with an invasive pO2 sensor. Further, through simulations of the mouse head geometry, we demonstrate that the skull and scalp exert minimal influence on the estimate oxygen saturation, while leading to small but systematic underestimation of total hemoglobin concentration. In total, these results demonstrate the robustness of small separation FDNIRS to assess tissue optical properties at millimeter depth resolution.