Mammography Taxonomy for the Improvement of Lesion Detection Rates

This work in this paper aims to assist CAD by classifying the context in which lesions can occur. This is achieved by a mammogram taxonomy system which automatically classifies breast parenchyma. The classification engine is an artificial neural network which successfully forms narrow classes to capture the subtleties in parenchyma variation. This paper presents the result of a series of experiments that digitized 628 mammograms at 50 µm from the mammographic archive at Falun Central Hospital.