Rule based elimination of regions of interest in pectoral muscle on mammograms
Hayati Türe, Temel Kayıkçıoğlu · 2016
In process of computer-aided detection of anomaly regions in mammography images, the pectoral muscle region may involve false positives. In this study, instead of extracting pectoral muscle region after detecting its boundary, elimination of false positives in this region is aimed by a ruled-based classification technique. A topographic representation that is called as iso-level contours map is used in the stage of image segmentation. The performance of the proposed method is tested on mammography images containing a mass 56 and 50 normal from MIAS database. By this method, totally 210 false positives in pectoral muscle is reduced to 10. Furthermore, this performance is achieved without causing any false negatives in all tested images.