ABSTRACT
This chapter discusses the challenge of detecting bilateral asymmetry in mammograms, along with the related difficulties encountered by radiologists. The fundamental notion of bilateral masking, based on the paradigm by Tabár, is recalled and framed in a protocol described to improve the detection of asymmetry between the two breasts of a patient. The chapter introduces the reader to bilateral asymmetry and breast cancer, and presents an overview of the work found in the literature on this topic. It describes the two publicly available databases used for validating the approach. Finally, the chapter presents the results of classification and validation procedures, and provides a discussion on a few challenging aspects. A strategy for the analysis of structural similarity is presented. The approach consists of sequential steps focused on landmarking of the two mammograms, on automatic bilateral masking of the overall breast regions, on the application of multidirectional Gabor filtering, and the extraction of spherical semivariogram descriptors and of structural similarity features.
