Classification of mammographic microcalcification clusters using a combination of topological and location modelling

Abstract

We have investigated the classification of micro-calcification clusters in mammograms by combining two existing approaches. One of the approaches involves extracting and using topological information (connectivity) about micro-calcification clusters as feature vectors to classify them as being benign or malignant. The other approach involves extracting and using location details of micro-calcification clusters (where they appear in a breast and/or mammogram) as feature vectors to classify them as being benign or malignant. We have investigated various aspects of both methods and their combination. Our initial results, based on MIAS and DDSM indicate no significant improvement over the topological approach on its own.