Improvement of Benign and Malignant Probability Detection Based on Non-subsample Contourlet Transform and Super-resolution

Mammography is a standard method for early diagnosis of breast cancer. In this paper, a method has been provided for improving quality of mammographic images to help radiologists so that probability of benign or malign breast lesions can be detected faster and more accurate and false positive rate (FPR) can be reduced. The presented algorithm includes 3 main parts of preprocessing, feature extraction and classification. In the preprocessing stage, a region of interest (ROI) is determined and quality of images is improved by non sub-sample contour let transform (NSCT) and super resolution (SR) algorithm altogether. In feature extraction stage, some features of the image components are extracted and skewness of each feature is calculated. Finally, support vector machine (SVM) is used to classify and determine probability of benign and malign disease. The obtained results on MIAS database indicate efficiency of the proposed algorithm.