CONTOURLET BASED TEXTURE ANALYSIS AND CLASSIFICATION OF MAMMOGRAM IMAGES

摘要

In this paper we have proposed a fully automated Computer Aided Diagnostic (CADx) system that can aid the radiologists in reading vast number of mammograms generated during screening procedures. The aim of the proposed system is to minimize the number of false positives and the number of false negatives. The remarkable potential of contourlet transform in extracting texture features of images with smooth contours is exploited in the proposed method. Mammogram images are taken from MIAS (Mammographic Image Analysis Society) database and a Probabilistic Neural Network (PNN) is trained to classify the image as normal, benign or malignant. The Region of Interest (ROI) of size 256 x 256 pixels is segmented from the mammogram image using Otsus N segmentation method and the Contourlet Coefficient Co-occurrence Matrix (CCCM) features are extracted. The texture feature set comprising the dominant features is fed to a PNN for the purpose of classification. To evaluate the efficiency of the proposed system 3 fold cross validation is performed and a classification accuracy of 91.1% is obtained. The performance of the proposed method is compared against other methods in terms of sensitivity, specificity and classification accuracy. The results obtained prove that the texture analysis of mammogram images using CCCM features outperforms other methods in terms of classification accuracy and hence can be successfully applied for the classification of mammogram images.