ISBN 978-3-8439-0333-2

978-3-8439-0333-2, Reihe Informatik

Tobias Böhler
Deformable Image Registration Methods for Clinical Applications of Magnetic Resonance Mammography

266 Seiten, Dissertation Universität Bremen (2011), Softcover A4

Zusammenfassung / Abstract

In clinical breast cancer diagnostics, magnetic resonance imaging complements the conventional x-ray based mammography examination, particularly for the clarification of suspicious lesion findings, periodic surveillance of high-risk patients, or navigation support during interventions and biopsies. In this thesis, three important applications of image registration methods in the context of magnetic resonance mammography are presented: the correction of motion artifacts in dynamic contrast-enhanced images, the matching of images from differently-weighted sequences, as well as the spatial synchronization of current and prior patient images in long-term follow-up imaging. For each of these targeted applications, specific image registration methods have been developed.

Supporting the implementation of the proposed methods, a flexible image registration software framework is introduced, focused on the unique requirements of a rapid development of application prototypes. Strategies for a qualitative and quantitative validation of registration results are discussed, and an objective accuracy assessment based on two different bio-mechanical breast models is described. In addition, a novel, specific measure for motion correction effectiveness is introduced, allowing an immediate assessment of registration performance in clinical workstation software. Using these techniques, the developed registration methods for the three targeted applications are illustrated and analyzed thoroughly, demonstrating their individual capabilities.