Detecting Architectural Distortion in Mammograms Using a Gabor Filtered Probability Map Algorithm

Breast Cancer is a disease that is prevalent in many countries. Computer-Aided detection (CAD) systems have been developed to assist radiologists in detecting breast cancer. This paper discusses an algorithm for architectural distortion (AD) detection with a better sensitivity than the current CAD systems. 19 images containing ADs were preprocessed with a median filter and Gabor filters to extract texture information. AD probability maps were generated using a maximum amplitude map and histogram analysis on the orientation map of the Gabor filter response. AD maps were analyzed to select ROIs as potential AD sites. AD map analysis yielded a sensitivity of 79% (15 out of 19 cases of AD were detected) with a false positive per image (FPI) of 18. Future work involves the development of a second stage in the algorithm to reduce the FPI value and application of the algorithm to a different set of database images.