Detection of Clustered Pleomorphic Micro-Calcifications in Digital Mammograms

In this paper, we present a novel multi-scale and multi-position classification (MSPC) method for detection of clustered pleomorphic micro-calcifications in digital mammograms. With this method, mammograms are divided into sub-images from which the image features are extracted and a cascaded Support Vector Machine (SVM) classifier is used to detect pleomorphic calcifications. Using the MSPC method, we robotically classify sub-images within a region of interest similar to other ROI methods used in CAD-based mammographic screening. Our experiments with this method using the Digital Database for Screening Mammography (DDSM) data show that the detection rate of clustered pleomorphic calcification (CPMC) can reach up to 97.26% with a 36.84% false positive rate.