Finding an optimum immuno-histochemical feature set to distinguish benign phyllodes from fibroadenoma

PurposeBenign phyllodes and fibroadenoma are two well-known breast tumors with remarkable diagnostic ambiguity. The present study is aimed at determining an optimum set of immuno-histochemical features to distinguish them by analyzing important observations on expressions of important genes in fibro-glandular tissue.MethodsImmuno-histochemically, the expressions of p63 and α-SMA in myoepithelial cells and collagen I, III and CD105 in stroma of tumors and their normal counterpart were studied. Semi-quantified features were analyzed primarily by ANOVA and ranked through F-scores for understanding relative importance of group of features in discriminating three classes followed by reduction in F-score arranged feature space dimension and application of inter-class Bhattacharyya distances to distinguish tumors with an optimum set of features.ResultsAmong thirteen studied features except one all differed significantly in three study classes. F-Ranking of features revealed highest discriminative potential of collagen III (initial region). F-Score arranged feature space dimension and application of Bhattacharyya distance gave rise to a feature set of lower dimension which can discriminate benign phyllodes and fibroadenoma effectively.ConclusionsThe work definitely separated normal breast, fibroadenoma and benign phyllodes, through an optimal set of immuno-histochemical features which are not only useful to address diagnostic ambiguity of the tumors but also to spell about malignant potentiality.

► Study addressed diagnostic ambiguity and differential malignant potentiality of benign phyllodes and fibroadenoma.
► Immunohistochemical molecular pathology features were semi-quantified.
► F-Score ranked features were analyzed by statistical distance functions to classify tumor based on optimum set of features.
► A set of seven immunohistochemical features on collagen and micro vessel density came out as optimum features for tumor classification.