The CALMA project: a CAD tool in breast radiography

In this paper we present the Computer-Aided Library for MAmmography (CALMA Project), i.e. an automated search for the mammograms' texture, the massive lesions and microcalcifications clusters. CALMA's main purpose is to colle ct a database of mammographic images, developing CAD tools to be used as a second radiologist in the classification of breast cancer diseases. In this moment, 2200 images are already in our database, which is, to our knowledg e, the largest in Europe. One-third of our digitized images are pathologica l, and they are fully characterized by a consistent description and diagnos is. We try to perform automatically the classification of mammographic imag es on the bases of tissues' textures. Such a classification should be done in an unbiased way with respect to radiologists and should support their in terpretation of the mammographic image. Results obtained with non supervise d neural networks are shown, as well as results coming from a mixed approac h (features extraction combined with FF-ANN). Massive lesions are rather la rge objects to be detected, but they show up with a faint contrast slowly i ncreasing with time. The need for tools able to recognize such a lesion at an early stage is therefore apparent. Our tools are based on a ROI hunter p rocedure for spiculated lesions showing a number of false positives of the order of 1.4 per image and keeping a 85% sensitivity on our sample. A micro calcication is a rather small (0.1-1.0 mm in diameter) but very brilliant. Some of them, either grouped in cluster or isolated may indicate the presen ce of a tumor. Up to now only 40 images with microcalcications from our dat abase have been analyzed, and a CAD tool has been designed to detect cluste rs, reaching a correct classification of 90%. (C) 2001 Elsevier Science B.V . All rights reserved.