Voting Based CAD Model for Breast Cancer Classification

Abstract

AbstractFor the diagnosis of any disease, the design of Computer-Aided Diagnostic (CAD) model is a significant one that guides physicians to ensure their opinion on diagnosis and so it influences the decline in mortality of breast cancer. This paper intends to devise a CAD model for mammogram classification i.e. benign or malignant and the same is done in three steps. The pre-processing phase aims for the removal of noise, pectoral muscle, and other unwanted objects. The region of interest (ROI) is extracted from the pre-processing step. Then extraction of features is carried out by making use of two-dimensional discrete wavelet transform (DWT), and the significant features are selected by applying the discrimination power analysis (DPA) technique. Eventually, the last step involves the classification using the Logistic Regression Classifier (LRC), Nu-Support Vector Classifier (Nu-SVC), Complement Naïve Bayes classifier (CNBC). Finally, the above said classifiers are conceptually combined by making use of the average predicted probabilities to obtain robust performance. The CAD model is evaluated with the MIAS dataset that provides a classification performance of 97.39% accuracy with a soft voting technique.KeywordsMammogramMIASBreast cancerSVMWaveletVoting