Microcalcification Clusters Detection Algorithms Based On SVM In Mammograms

In this paper, the detection of micro-calcification cluster was in-depth study in mammograms, a algorithms is presented to detect the microcalcification clusters based on support vector machine and nonsubsampled contourlet transform in mammograms, it effectively improve the detection accuracy of microcalcifications. At present, the most effective detection method is to detect micro-calcifications in mammograms, and determine whether the deterioration of micro-calcifications. The microcalcifications are the earliest sign of breast carcinomas and their detection plays a particularly important role in the early diagnosis. The most important parts would be described as follows:(1) Due to the mammograms are low contrast images, a method is presented to enhance the mammograms based on Nonsubsampled Contourlet Transform, which can effectively suppress background and noise, the same time enhance the micro-calcification regions.(2) In order to reduce the false positive rate of detection, the Micro-calcification features were extracted from the Contourlet coefficients of the image, which are the input of SVM classifier. Compared with the wavelet transform, the features which were extracted from Nonsubsampled Contourlet Transform domain can better represent the image information.(3)The SVM classifies contains one input, two hidden and one output layers. The system classifies the mammogram images as normal or abnormal, and abnormal severity as benign or malignant.Compared with traditional methods, the proposed algorithm has good practicability, which can effectively detect the abnormal mammograms, and classify severity as benign or malignant.