Heuristic approach for computer-aided lesion detection in mammograms

Digital mammography is a common screening method for early detection of breast cancer. Its efficiency varies from 60 to 90 %, depending on various factors such as breast density, quality of the mammogram as well as experience and knowledge of the radiologist (Andolina and Lillé in Mammographic imaging: a practical guide, Lippincott Williams & Wilkins, Philadelphia, 2010). One of the most effective ways to increase the cancer detection rate is double reading (Fernandez-Lozano in Soft Comput 19(9):2469–2480, 2015). Regarding the fact that only the early-stage patients have high chances of survival, a computer-aided detection (CAD) system for mammography that provides a second, independent diagnosis should be considered a valuable and lifesaving tool. In this paper we present a new heuristic approach focused on analyzing characteristics of the mammogram that may indicate the presence of breast cancer. Described soft computing detection algorithm based on local features allows us to extract microcalcifications and possible tumor areas from the image. Because calcifications are associated with certain types of lesions, we believe that this idea would result in improving existing medical information systems. Future inclusion of fuzzy classifiers in the algorithm may also provide additional diagnostic value. Conducted research confirms that proposed procedure correctly identifies the regions of interest and could be used as a base of a CAD system in a double-reading procedures.