Research On Breast Mass Detection Based On Dual-view Mammograms

Breast cancer is one of the most common cancers for women. Most breast cancers can be detected from breast masses. Mammograms are an effective method of early detection of breast cancer. Radiologists often compare the two views of the same breast when reading mammograms. Dual-view- based CAD was developed from that model.The first step of our dual-view-based CAD was detecting candidate masses separately from the two views of the same breast. A region restricted segmentation method was used to segment the breast region. Then, suspicious masses were located by a multi-scale template matching method. Logistic classifier was used to calculate the detection score of each suspicious mass.According to the fact that the same mass in different views has identical distance from the nipple, suspicious masses detected form dual-views were matched as matched pairs. Features were extracted from each matched pair, and then classified by Logistic classifier. False positives were reduced by the detection scores of suspicious mass and matched pairs.184 cases were chosen from Digital Database for Screening Mammography of South Florida University, when the sensitivity of single-view-based CAD reached 90%, average 2.9 false positive regions were detected in each image. For the dual-view-based CAD, the sensitivity and false positive were 90%, and 1.98 respectively. Average one false-positive is reduced in dual-view-based CAD as in single-view-based CAD.