A Computerized System for Detection and Segmentation of Clustered Microcalcifications

Abstract

Over the past years, there have been extensive studies in developing effective CAD techniques for the deyection of microcalcifications. Umfortunately, most of them require manually locate supicious areas, called regions of interest (ROIs) prior to detection. In this paper, we present a three=stage computer aided diagnostic (CAD) method which not only automatically locates ROIs But also detects and segments clustered microcalcifications. The first stage is to extract the breast from the background followed by a second stage to detect and locate the suspicious areas, and finally the segmentation of clustered microcalcifications are carried out in the third stage. Performance of this method is evaluated by skillful radiologists based on a database developed by Nijmegen University Hospital and mammograms obtained from Taichung Veterans General Hospital. The initial assessment of the proposed system is very encouraging.