Near-infrared optical imaging aims to reconstruct the absorption µ a and scattering µ s coefficients in order to detect tumors at early stage. However, the reconstructions have only been limited to µ a and µ s due to theoretical and computational limitations. The authors propose an efficient method of the reconstruction, in 3D geometries, of the anisotropy factor g of the Henyey-Greenstein phase function as a new optical imaging biomarker. Methods: The light propagation in biological tissues is accurately modeled by the Radiative Transfer Equation (RTE) in the frequency domain. The reconstruction algorithm is based on a gradient-based updating scheme. The adjoint method is used to efficiently compute the gradient of the objective function which represents the discrepancy between simulated and measured boundary data. A parallel implementation is carried out to reduce the computational time. Results: We show that by illuminating only one surface of a tissue-like phantom, the algorithm is able to accurately reconstruct optical values and different shapes (spherical and cylindrical) that characterize small tumors-like inclusions. Numerical simulations show the robustness of the algorithm to reconstruct the anisotropy factor with different contrast levels, inclusion depths, initial guesses, heterogeneous background, noise levels and two-layered medium. The crosstalk problem when reconstructing simultaneously µ s and g has been reported and achieved with a reasonable quality. Conclusions: The proposed RTE-based reconstruction algorithm is robust to spatially retrieve and localize small tumoral inclusions. Heterogeneities in g-factor have been accurately reconstructed which makes the new algorithm a candidate of choice to image this factor as new intrinsic contrast biomarker for optical imaging.