Abstract
Masses are among the early signs of breast cancer, which is the second major cause of death in women. For mass detection, a mammogram is segmented into regions of interest (ROIs) that contain masses as well as suspicious normal tissues, which lead to false positives. The problem is to reduce the false positives by classifying ROIs as masses and normal tissues. Further, the detected masses are needed to be discriminated as benign and malignant. We investigate the performance of six different Gabor feature extraction approaches for these mass classification problems. These techniques employ Gabor filter banks for extracting multiscale and multiorientation texture features which represent structural properties of masses and normal dense tissues in mammograms. The feature extraction approaches are evaluated over the ROIs extracted from MIAS database. Successive Enhancement Learning based weighted Support Vector Machine (SELwSVM) is used to efficiently classify the generated unbalanced datasets. The best performance in terms of area under ROC curve (Az = 1.0) is obtained by the Gabor features extracted using first order statistics of the Gabor responses and LDA.