The Method Of Mammogram Image Processing

In recent years, breast cancer is becoming one of the most common malignanttumors, for the incidence of female breast cancer is higher and higher. According tothe latest statistical data, the probability of the incidence rate in American womenwith breast cancer is the highest among all cancers. Based on a survey of metropolitanwomen, breast cancer is also the most common malignancy of women, and there is anincreasing trend in the incidence.Early diagnosis and treatment are the keys to reduce breast cancer mortality.Currently the most common method to inspect the breast cancer is mammographyimaging, but because the huge amount of data and the imaging characteristics of earlybreast cancer is not obvious the earlier diagnosis is very difficult. With thedevelopment of computers and medical science, computer-aided diagnosis of breastimages has become an important research topic, which mainly includes the studies ofmasses, calcifications and breast density. This paper mainly study the masses andbreast density.Mass detection on mammography is an effective method for breast cancerdiagnoses. Two automated mass detection method are proposed, that all include: first,mammograms are preprocessed to remove background, tags, pectoral muscles andnoises, and segment of breast; second, Kmean method is used to segment the RegionOf Interest(ROI); Next, features of mass such as shape and texture are extracted;finally, according to the extracted features to separate the masses and normal tissues.The difference between the two methods is: the first method is using a threshold valueaccording to the feature in the final classification, while the second method is usingSVM (Support Vector Machine, SVM) in ROI for classification.Breast density is another manifestation in the X-ray image. This paper presents aglobal-based breast density estimation method, which compared with the traditionalmethod, the feature extraction and the selection of classifier are its innovation. Theselected feature is l-level histogram in the uniform blocks which tile in themammogram, the classifier is using the ELM.