Design And Implementation Of Breast Lumps Segmentation Algorithm

Breast cancer remains a leading cause of cancer deaths among women as one of the common malignant tumor. Early diagnosis and treatment is an efficient way of reducing the morality of breast cancer. Breast mammography X-ray has the characteristics of simple operation, relatively small risk and low cost, it is currently considered the most reliable method, and playing an important role in improving the patients’ survival rate.Because there are a large number of images, the doctors have the large workload so that cause missed diagnosis or misdiagnosis. In America, the delayed treatment of breast cancer accounts for 6%~10% because of missed diagnosis or misdiagnosis. Computer-aided and diagnosis(CAD) system may improve the rate of diagnosis. However, traditional CAD is similar to the way of “black box”, which only hints the lesion area but don’t explain the reason why these areas are marked, yet has the problem of low detection sensitivity so that the doctor lose confidence in CAD. However, computer-aided diagnose of mammographic masses using content-based image retrieval(CBIR-CAD) not only hints lesion area but also returns K images with the most similar image to the interested area.This paper study around the key technology in CBIR CAD system to make the lump realize effective segmentation, detection and classification, further improve the efficiency and accuracy of diagnosis. A number of suspicious lumps segmentation methods are summarized in this thesis study. Further compared and analyzed their performance and advantages. A new way which called CV-RE is proposed to segment the suspicious lumps from area-of-interest(ROI). It combines rectangle rough region extracting(RE-ROI) with C-V model. Extracting rectangle rough region by these way of template matching、image binary conversion and mass region dilating. Suspicious lumps segmentation using C-V model in this area. Effectively solve the limitation of the traditional C-V model in dealing with the image of uneven background. Improve the accuracy of breast lumps segmentation significantly.