We report the results of a comparative evaluation of the diagnostic capabilities of autofluorescence, diffuse reflectance,
and Raman spectroscopic approaches in differentiating the various types of breast tumors from normal breast tissues.
Optical spectra (n=293) were acquired ex-vivo from a total of 75 breast tissue samples belonging to six distinct
histopathologic categories: invasive ductal carcinoma, lobular carcinoma, ductal carcinoma in-situ, fibroadenoma, other
benign tumors, and normal breast tissue. Autofluorescence, diffuse reflectance, and Raman spectra were measured from
the same locations of a given tissue sample. A probability based multivariate statistical algorithm capable of direct multiclass
classification was developed to analyze the diagnostic content of the optical spectra measured from the same set of
breast tissue sites with these different techniques. The algorithm uses the theory of nonlinear Maximum Representation
and Discrimination Feature (MRDF) for feature extraction, and the theory of Sparse Multinomial Logistic Regression
(SMLR) for classification. The results of discrimination analyses reveal that the performance of Raman spectroscopy is
superior to that of all others in classifying the breast tissues into respective histopathologic categories. The best
classification accuracy was observed to be ~96%, 86%, 94%, 98%, 85%, and 100% for invasive ductal carcinoma,
lobular carcinoma, ductal carcinoma in-situ, fibroadenoma, benign tumors and normal breast tissues, respectively, on the
basis of leave-one-out cross validation, with the overall accuracy being ~97%.