Classification of Clustered Microcalcifications in Mammograms using Topological and Shape Features

Abstract

Classification of benign or malignant microcalcification clusters in mammograms is a major diagnostic challenge for radiologists. Clinical studies have explained that malignant microcalcifications tend to be small and densely distributed while benign microcalcifications are generally larger, and more diffusely distributed. Most of the existing works are focused either on the shape, texture or the distribution of the clusters. Topological feature extraction method fail to discriminate malignant from benign when a single microcalcification is detected using segmentation approach. In this case shape feature of the individual microcalcification will discriminate malignant from benign. The size of the microcalcification is determined by the shape features and the distribution of the microcalcification is determined by the topological features of the cluster. Hence, topological features and shape features are extracted, which constitute the feature space for classifying microcalcification clusters. SVM is employed for classification. High classification accuracy and high true positive rate are obtained.