A Novel Mammography Image Representation Framework with Application to Image Registration

X-ray mammography is a fundamental tool for breast cancer detection and diagnosis. A difficult problem arises when analyzing, integrating and comparing the information from different mammograms due to intensity changes and distortions induced by breast deformations. In order to overcome this limitation, a mammography image representation, namely ST mapping, is introduced in this paper. The proposed method consists of mapping the image intensities according to a curvilinear coordinate system that adapts to the breast geometry in order to yield a deformation-robust representation of the image features. In a practical application, the ST mapping is exploited for performing image registration. To our knowledge, this approach is completely novel since it does not require neither computing global or local geometric transformations nor finding point correspondences between images. In contrast, the registration is performed only based on the breast contour. Experiments with synthetic image deformations of real mammography images are provided in order to show the robustness of the proposed method to general deformations.