Skip to Main Content
Computer-aided diagnosis (CAD) system can promote the detection accuracy by providing a “second opinion” to the radiologist, so high accuracy detection of mass in mammogram is critical for improving the performance and efficiency. In this paper, we designed a mass auto-diagnosis method in mammogram based on texture features. First, the mass was detected base on bilateral comparison, and the center of region of interest (ROI) was located. Second, fractal dimension and two-dimensional entropy were calculated as the texture features. Last, the kinds of ROI were diagnosed by Support Vector Machine (SVM), mass or normal region. A total of 106 prior mammograms were automatically detected, experimental results indicate that mass and suspected region have obvious difference in the fractal dimension and other texture features, and SVM is an effective classify method, and reduce the error rate in the mass detection, and the performance of the method is a sensitivity of 85.11% at 1.44 false positives per image.