Akademik Veri Yönetim Sistemi
Medical Technologies National Congress (TIPTEKNO), Trabzon, Türkiye, 12 - 14 Ekim 2017
A pectoral muscle usually resembles a salient dense region, which is roughly triangular shape, in mediolateral oblique (MLO) position mammography. Similarly, masses usually appear to be salient dense in mammograms. Therefore, first of all, most computer-aided detection (CAD) systems remove the pectoral muscle region in order to reduce the number of false positives. In this study, a method is proposed to determine the pectoral muscle region by the optimal contour selection (OCS) approach. After the preprocessing stage, an isocontour map for the MLO mammogram image is formed. The proposed method examines the variation of the nesting contours of the pectoral muscle region. In the next stage, some discriminating features based upon shape and intensity for nesting contours is determined. The pectoral muscle border is determined by the defined OCS rules. In the final stage, roughness of the pectoral muscle border is removed by least-squares fitting method. This method, which was tested on 84 mammogram images taken from the MIAS database, performed the segmentation with %5.2 false negative (FN) and 3.5 % false positive (FP) ratios. For small size pectoral muscle regions, proposed method outperforms some state of the art studies in this field.