A method to reduce curvelet coefficients for mammogram classification

This paper presents a method for classification of normal and abnormal tissues in mammograms using curvelet transform. The curvelet coefficients are represented into certain groups of coefficients, independently. Some statistical features are calculated for each group of coefficients. These statistical features are combined with features extracted from the mammogram image itself. To improve the classification rate, feature ranking method is applied to select the most significant features. The classification results of support vector machine (SVM) using 10-fold cross validation are presented. The classification results show that the ranked features improved the classification rate up to 85.48% with group of 200 coefficients.