Breast cancer detection and diagnosis is a critical and complex procedure that demands high degree of accuracy. This paper develops an algorithm for mass detection on mammograms. As masses in X-ray mammograms are subtle and difficult to diagnose by radiologists, such tools often serve either as second opinion or pre-reader and are, therefore, of critical importance. The first step of the algorithm is a wavelet-based enhancement which removes bright background due to dense breast tissues while preserving features and patterns related to the masses. Entropy maximization thresholding (EMT) is then applied to segment the enhanced image, followed by a median filter which discards noise. The method was tested on a database of 105 mammograms digitized at a pixel size of  and previously marked by radiologists. We found a detection rate of 98.571. The detection performance of the computer aided diagnosis (CAD) system was further assessed by receiver-operating characteristic (ROC) analysis. An area under the ROC-curve (AUC) of 0.938 and 0.926 were, respectively, obtained with and without enhancement step. The high performance indicated the usefulness of the EMT method.

breast mass, entropy maximization, image enhancement, thresholding, wavelet.