Based On Morphological Characteristics Of Mammary Tumors Of The Mammography Image Analysis

Breast cancer is a severe disease in women causing a series of family and social problems. Early diagnosis and treatment of it is the key step to reduce the mortality rate. Mammography is considered to be the most effective and reliable diagnostic tool. However, the similarity of the densities in each part of the breast leads to the difficulty in diagnosis. Computer-aided diagnosis (CAD) systems can play a supportive role in the early detection and diagnosis of mammary cancer. Experimental studies have shown that a reliable CAD system can make the images more objective and standardized and increase the diagnosis sensitivity.The study is based on the concept of modularization in CAD system, and completes detection and analysis of the breast mass automatically in the MATLAB environment.First, more than 20 mammograms were initially refined by the means of the breast extraction algorithm. Almost 2/3 pixels without meaningful information were deleted in average. The sizes of the refined images are smaller compared with previous ones. Therefore, the algorithm can reduce the computing burden of segmentation in the following steps.Second, three different methods were applied for mass segmentation after the elimination of the background of the ROI. The mass boundaries obtained by these methods (Maximum Entrophy, Otsu Method, Region Growing Algorithm)were compared with those obtained artificially which are regarded as the gold standard to calculate the overlap rates for the detection of their sensitivity and accuracy. And the results showed that all the three methods can draw the region of the mass effectively. The overlap rates of these methods are 0.812±0.031,0.747±0.065,0.837±0.067respectively. The overlap rates among them showed statistical significance (P<0.05) using student t-test. Third, 159 breast tumor images confirmed by pathology were divided into two groups: the malignant group (97 images) and the benign group (62 images). Each of the mass images was processed with the Modified Region Growing Algorithm to extract the characteristic of its density, contrast, radial distance, roughness, moments, compactness, spicular index(SI), and so on. The values of the following characteristic between the two groups: contrast, roughness, fourth-order moment,compactness and SI were compared by means of the student t-test. And, there were statistical significance between the two groups in terms of all the four above parameters.At last, common image-processing tools, such as the adjustment of image size, location, rotation, contrast, brightness, and smoothness, the display of fake-color or inversed image, and the measurement of distance,were added on the users'application interface to meet the clinical demand.The study can detect and analyze the breast mass in the mammogram automatically and effectively, which is helpful for the study of CAD system in the clinical diagnosis of breast tumors.