In spatial frequency domain imaging (SFDI), a spatially modulated intensity pattern is projected on to tissue, with the demodulated reflectance having more superficial sensitivity with increasing spatial modulation frequency. With sub-diffusive SFDI, very high (>0.5 mm-1) spatial modulation frequencies are projected yielding sensitivity to the directionality of light scattering with only few scattering events occurring and sub-millimeter penetration depth and spatial resolution. This technique has been validated in a series of phantom experiments, where fractal distributions of polystyrene spheres were imaged, and through a model based inversion, the size scale distribution versus overall density of these particles could be separated and visualized in spatially resolved maps. With sensitivity to localized light scattering over a wide field of view (11 cm x 14 cm), this technique is being translated for the application of intraoperative breast tumor margin assessment. To test sensitivity to changes in human breast tissue morphology, a cohort of over 30 freshly excised human breast tissue specimens, including adipose, fibroglandular, fibroadenoma, and invasive carcinoma, have been imaged and co-registered to whole specimen histology. Statistical analysis of the distributions of both textual raw reflectance parameters and model based optical properties for each type of tissue will be presented. Furthermore, classification algorithm development and analysis to predicted likelihood of cancer on the surface of the tissue will also be presented. Reflectance maps, optical property maps, and probability likelihood maps of spatially heterogeneous samples with multiple tissue types will also be shown.