Detection of masses and the pectoral muscle region in mammography images considering the superposition effect

In this thesis, a new approach is proposed that determines masses in mammography images and the pectoral muscle region, taking into account the superposition effect. In order to increase the prominence of superpositioned regions, four different directions of Gabor filters are first used, and then topographic maps are formed. In this study, the Rule Based Classification of Isocontour Lines (RBCI) method is devoleped to determine regions of interest in the topographic maps. In this method, the nesting debth, morphological, intensity, and texture features calculated from the concentric isocontours within Optimum Lifetime (OLT) are used to detect the ROIs. The topographic map is divided into blocks to increase the success of the RBCI method on the distorted pectoral muscle regions under the effect of superposition. In this way, the effect of the distortion is localized in blocks and the sequence of contours with specific patterns of the undistorted pectoral muscle border parts are revealed. The suspicios regions with low prominence are detected with high accuracy using the developed the RBCI method. Studies in the literature that performed pectoral muscle region removal did not take into account the masses that were superpositioned with the pectoral muscle region. Within the scope of this thesis, the performance of the RBCI method in these situations examined in detail.