Research On Techniques Of Computer-aided Diagnosis Of Breast Cancer For Mammography

Breast cancer is the most frequently diagnosed cancer in women in worldwide. Mammography is the best available tool for screening for the early detection of breast cancer. Computer-aided detection and diagnosis (CAD) techniques could offer a cost-effective alternative to double reading as a means of reducing errors. There is no CAD product of China's own intellectual property rights. International study for CAD system, mass detection in particular, has made some good results, there exists great potential.The specific steps of mass detection are as follows: extraction of region of interest (ROI), segmentation of suspicious mass, feature extraction and classification.In the first step, a template matching approach is proposed to automatically extract ROI, which contains suspicious mass. The second step is to segment suspicious mass from normal tissue. To find the exact boundary of the suspicious mass, we first enhance the ROI before segmentation. And then we experiment two segmentation algorithms to the enhanced ROI, dynamic programming-based method and maximum entropy threshold based method. The result shows the first method yields better performance. In the third step, the features of suspicious mass are extracted. Based on these features, the final step is to classify the suspicious masses into two groups, positive and negative, by an improved K-Nearest-Neighbor classifier, and to cue the positives on the screen.The dataset comes from Digital Database for Screening Mammography (DDSM) of South Florida University. We choose 155 cases including 306 images from the database, each image contains a mass. The results of CAD scheme yield a sensitivity of 91.4%, with 2.04 false positives per image.