Diffuse reflectance spectroscopy of tissue allows quantification of underlying physiological and morphological changes associated with cancer, provided that the absorption and scattering properties of the tissue can be effectively decoupled. A particular application of interest for tissue reflectance spectroscopy in the UV-VIS is intraoperative detection of residual cancer at the margins of excised breast tumors, which could prevent costly and unnecessary repeat surgeries. Our multi-disciplinary group has developed an optical imaging device, which employs a model-based algorithm for quantification of tissue optical properties, and is capable of surveying the entire specimen surface down to a depth of 1-2 mm, all within a short time as required for intraoperative use. In an ongoing IRB-approved study, reflectance spectral images were acquired from 55 margins in 48 patients. Conversion of the spectral images to quantitative tissue parameter maps was facilitated by a fast scalable inverse Monte-Carlo model. Data from margin parameter images were reduced to image-descriptive scalar values and compared to gold-standard margin pathology. Use of a decision-tree based classification algorithm on the two most significant optical parameters resulted in a sensitivity of 79% and specificity of 67% for detection of residual tumor of all pathologic variants, with an 89% sensitivity for ductal carcinoma in situ alone. Preliminary data from this ongoing clinical study suggest that this technology could significantly reduce the number of unnecessary repeat breast conserving surgeries annually.