Image Analysis Methods in MRI Examinations of the Breast

Breast cancer represents a real emergency in many European and North American countries. In order to detect the diseases as soon as it starts developing, many countries have organized screening programs aimed to women monitoring. Traditional imaging modalities, like X-Ray mammography, do not provide certain and reliable results on young women or in women who underwent surgical interventions. Three-dimensional breast MRI has proven to be a valuable tool for disambiguate uncertain mammographic findings and for the pre-operative planning. However, in order for MRI of the breast to be effective, a contrast agent, highlighting areas with a high degree of vascularization, needs to be used. The full breast volume needs to be acquired once before and several times after the contrast agent injection. A contrast-enhance MR examination of the breast contains hundreds of images that must be inspected carefully. Automatic approaches to breast cancer detection can help radiologists in this hard tasks and speed up the inspection process. In this paper we present a survey of automatic approaches to breast cancer detection in MRI. We discuss both registration algorithms needed to align corresponding images acquired at different instants in time and classification algorithms able to label a suspicious area according to features automatically computed from the images.