A tool for breast MR images processing and classification

One of the major problems for the women's quality of life in the industrialized countries is the early diagnosis of breast cancer. Contrast-enhanced magnetic resonance of the breast is the most attractive alternative to standard mammography but the manual inspection of images is a long, subjective and error-prone process. To face these problems, we propose a method whose steps are: segmentation, correction of movements and a dynamic study to search and classify the enhancing regions in the images. The proposed method removes noisy background and tissues not interesting for diagnosis, extracts the relevant information and classifies the resulting images by discriminant curves. The developed algorithms has been tuned and tested on 149 exams, supplied by the Imaging Dept. of the Medical Faculty of the University of Pisa. The paper will show the implemented system is innovative and requires minimal user interaction, providing results comparable with clinical direct diagnosis.