Computer Aided Diagnostic System for Classification of Mammograms

Abstract

Breast cancer is the second cause of fatality among all cancers for women. Subsequently the reason of breast cancer remains anonymous, primary prevention becomes impossible. The most usual breast cancers types are mass (density) and micro-calcification. Early detection of these types of lesions is one of the important issues for breast cancer control. Currently, x-ray mammography is the particular most active, low-cost, and highly sensitive method for spotting small lesions. Thus, an adequate Computer Aided Diagnosis (CAD) classification system is designed for breast tumor for supporting immature radiologists in the diagnosis procedure. In this study, author proposed an efficient CAD system to categorize database into normal, benign and cancer breast tissue types. We have been extracted different type of squared shaped ROIs manually from the middle part of the breast. Statistical texture features are extracted from these ROIs. Finally, classification task is completed using Multilayer Perceptron (MLP) classifier, Support Vector machine (SVM) classifier and k-nearest neighbor (k-NN) classifier followed by feature selection technique. The outcomes of all the set of features are compared on the basis of their accuracy. Then, mammograms are categorized into normal, benign and cancer using the same procedure. The another proposed algorithm aims to developing an efficient image processing based Graphical User Interface (GUI) for detection of micro-calcification by wavelet transform in early stage. The result of first algorithm shows that accuracy reaches upto 90% when combination of optimal features used as input. We have successfully detected breast tumor using GUI model.