New intensity based features for classification of mammograms
Pratham Arora, Mandeep Singh · 2014
Breast tissue density is a pivotal signpost for breast cancer risk. Many sundry methods have been proposed to classify the breast tissue density. In this paper, three new features are proposed, which can be used to classify breast tissue density into fatty and dense tissue type. The new proposed features are used with gray level co-occurrence matrix features to classify the mammograms through optimal feature selection process. The new features are based on the intensity of the grey level of the image. To corroborate the significance of new features, various standard classifiers are used. The results are able to perceive the feasibility of the proposed method to classify the breast density tissue into fatty and dense. The new proposed method gives 94.5% accuracy. We also juxtaposed the accuracy of proposed features with Haralick's texture features and the combination of both. All the classifiers used in this model were combined in the end and a classifier combination was used to calculate the accuracy on the basis of probability estimates. The results were more convincing than individual classifiers.