Microcalcifications Detection in Mammographic Images Using Texture Coding

Breast cancer represents the most frequently diagnosed cancer in women. In order to reduce mortality, early detection of breast cancer is important, because diagnosis is more likely to be successful in the early stages of the disease. This paper presents a new method for automatic detection of clustered microcalcifications in digitized mammograms. Compared to previous works, the innovation here is that the processing is performed in the coded images instead of the original ones. This new method uses the coding of textures of the mammographic images on the basis of which Haralick features are computed for SVM classification purpose. By comparing our results with those found in the literature, we proved that the method of coding developed does not degrade the quality of the contained information in the mammograms and enormously reduces the computing time of the haralick vector parameters released from the cooccurrence matrix. Furthermore, the rates of classification found by using the coded images are much improved compared to those obtained on the basis of original images. Classification rates enhancements were also revealed by testing our method compared to rank coding method.