Accurate and efficient reconstruction of hemodynamic changes is an important step towards the implementation of NIRS as an enhanced clinical tool for understanding oxygenation changes at various depths within the brain. Depth information could provide insight on how oxygen transported to the tissue. For this work, we ran Monte Carlo simulations to develop sensitivity profiles for various source-detector separations. The source-detector separations were based on our custom built 108 channel NIRS probe and consisted of separations of 15 mm, 30 mm, 36 mm, and 45 mm. We used the mesh-based Monte Carlo program MMCLAB (Fang et al. 2010) to acquire the sensitivity profiles. The sensitivity profiles consisted of a tetrahedral mesh which was converted to a regular grid in three-dimensional space. Then, the structural tensor was calculated for each voxel and the Hamilton-Jacobi equation was solved anisotropically for the tensor volume. As the result, the distance map was in same space as the calculated tensor volume. Using this distance map, we modeled the probabilistic path of photons. We then weighted the hemodynamic changes acquired by our NIRS probe according to the probabilistic path to reconstruct hemodynamic changes in the prefrontal area of the brain.