A model-based framework for the detection of spiculated masses on mammography.

摘要: The detection of lesions on mammography is a repetitive and fatiguing task. Thus, computer-aided systems have been developed to aid radiologists. accuracy current much higher for clusters microcalcifications than spiculated masses. In this article, the authors present new model-based framework invented class linear filters, lesion converging lines or spiculations. These filters are highly specific narrowband which designed match expected structures As part algorithm, also novel technique enhance spicules mammograms. This entails filtering in radon domain. They models reduce false positives due normal structures. A key contribution work that parameters algorithm based measurements physical properties results presented form free-response receiver operating characteristic curves images from Mammographic Image Analysis Society Digital Database Screening Mammography databases.