Detect abnormalities in mammograms by local contrast thresholding and rule-based classification

Abstract

Mammography, which uses X-ray technology to image the breast, is currently the most effective and reliable method for early cancer detection. There exists limitations of human observers: up 30% of breast lessions are missed during routine screening. It is believed that computer-aided detection (CAD) schemes could ultimately provide a useful "second option" for radiologists and potentially improve their diagnostic accuracy. The proposed detection process bases on local contrast thresholding and rule-based classification which is performed over the preprocessed and segmented mammograms. A relatively high detection rate of suspicious abnormal regions (mass and/or microcalcification) on the testing set of mammograms from Mini Mias Database implies that the proposed method can assist technologists in more efficiently and accurately locating the exact areas for subsequent exams. ©2010 IEEE.