Subtle Directional Mammographic Findings in Multiscale Domain

摘要

The aim of our research is to extract local subtle directional texture orientation on mammographic images using a set of differentiating features calculated in various transformation domains. The main goal in this paper was to establish the usefulness of multiscale transformations, in particular the complex wavelet transform in automatic recognition of indefinite directional pathologies on mammograms. Our initial test was conducted on ROIs of mammograms containing one of typical breast cancer signs - architectural distortions (33 ROIs out of all analyzed 289 ROIs). The promising results have been achieved. It seems that the complex wavelet transform is effective domain to extract well-differentiating features.