A New Feature Extraction Approach Using Contourlet Transform and T-Test Statistics for Mammogram Classification

Authors : Nebi GEDİK

Pages : 16-20

Doi:10.17694/bajece.557693

View : 17 | Download : 8

Publication Date : 2020-01-31

Article Type : Research Paper

Abstract :In this study, a CAD system is recommended for the classification of mammography images as normal-abnormal and benign malignant. The proposed system consists of the feature extraction, determination of the distinguishing capabilities of the features and selection of the features using by dynamic thresholding according to the determined distinguishing capabilities. It uses the contourlet transform to extract features. The distinguishing capabilities of the features are determined by using t-test statistics, and the thresholds are applied to those values to select effective ones. Classification is performed using support vector machine algorithm for every iteration with each thresholding step. Among the results of the iteration performed, the optimum data that have the best performance, which is they have maximum accuracy result with the minimum number of features, is selected as the optimum value. To evaluate the optimal feature set, classification carries out using the feature set applying 5-fold cross-validation. According to the results, the proposed method can be accepted as a successful CAD system. 
Keywords : Classification, CAD, Mammogram, Contourlet transform