Research On Image Segmentation And Classification Based On Mammogram

Breast cancer is one of the main causes of death among women.Practice has provedthat early detection and treatment is the most effective method for the breast cancer, andreduce the patient’s mortality. Mammography is one of the most widely used methods inthe early diagnose of breast cancer. But there is no complete and practical treatmentprocesses and methods in practice and no complete breast contour extract method.Thisarticle focuses on the author’s research on image segmentation and classification in themammogram, and the establish of a complete set of operational processes.The first Step is the preprocessing, rotated and reflected the mammogram and set thepectoral muscles at the upper left corner, and then remove the noise and impurities in theimage, and next is the enhancement operation to increase the contrast, then use theproposed region growing method and set the growing point at the upper left corner of theimage to get the initial pectoral muscle segmentation points,then the discrete pointspolynomial curve fitting algorithm is applied to segment pectoral muscle, and thenthrough the proposed histogram-based segmentation algorithm combined with the dilateand erosion operation to extract the breast boundary contour. Through histogram andthreshold segmentation algorithm based on fuzzy theory be manipulated to obtain the ROI.Finally, the final ROI extraction using morphological’s filling and erosion operation.Secondly, extracting the ROI’s features for gray, geometric and texture, based onthose features we use the support vector machine to learn and classify the samples. We candiagnose diseases by classification results.Finally, the experiment has verified that the proposed mammogram segmentation andclassification algorithm.