TitleMultimodal image registration using multivariate information theoretic similarity measures
NameChappelow, Jonathan (author), Madabhushi, Anant (chair), Bhanot, Gyan (internal member), Boustany, Nada (internal member), Bloch, B Nicolas (outside member), Tomaszewski, John E (outside member), Rosen, Mark (outside member), Rutgers University, Graduate School - New Brunswick,
Imaging systems in medicine,
DescriptionMultimodal and multiprotocol image registration refers to the process of alignment of two or more images obtained from different imaging modalities (e.g. digitized histology and MRI) and protocols (e.g. T2-w and PD-w MRI). Registration is a critical component in medical applications including image guided surgery, image fusion for cancer diagnosis and treatment planning, and automated tissue annotation. However, registration is often complicated on account of differences in both the image intensities and the shape of the underlying anatomy. For example, non-linear differences in the overall shape of the prostate between in vivo MRI and ex vivo whole mount histology (WMH) often exist as a result of the presence of an endorectal coil during pre-operative MR imaging and deformations to the specimen during slide preparation. To overcome these challenges, we present new registration techniques termed Combined Feature Ensemble Mutual Information (COFEMI) and Collection of Image-derived Non-linear Attributes for Registration Using Splines (COLLINARUS). The goal COFEMI is to provide a similarity measure that is driven by unique low level textural features, for registration that is more robust to intensity artifacts and modality differences than measures restricted to intensities alone. COLLINARUS offers the robustness of COFEMI to artifacts and modality differences, while allowing fully automated non-linear image warping at multiple scales via a hierarchical B-spline mesh grid. In addition, since routine clinical imaging procedures often involve the acquisition of multiple imaging protocols, we present a technique termed Multi-attribute Combined Mutual Information (MACAMI) to leverage the availability of multiple image sets to improve registration. We apply our registration techniques to a unique clinical dataset comprising 150 sets of in vivo MRI and post-operative WMH images from 25 patient studies in order to retrospectively establish the spatial extent of prostate cancer (CaP) on structural (T2-w) and functional (DCE) in vivo MRI. Accurate mapping of CaP on MRI is used to facilitate the development and evaluation of a system for computer-assisted detection (CAD) of CaP on multiprotocol MRI. We also demonstrate our registration and CAD algorithms in developing radiation therapy treatment plans that provide dose escalation to CaP by elastically registering diagnostic MRI with planning CT.
NoteIncludes bibliographical references
Noteby Jonathan Chappelow
CollectionGraduate School - New Brunswick Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.