Michael Hofer1, Thomas Pock2, Karin Kapp3, Thomas Bauernhofer4, Franz Ebner5, Rudolf Stollberger1
1Institute of Medical Engineering, Graz University of Technology, Graz, Austria; 2Institute for Computer Graphics and Vision, Graz University of Technology, Graz, Austria; 3Department of Radiotherapy-Radiooncology, Medical University Graz, Graz, Austria; 4Department of Oncology, Medical University Graz, Medical University Graz, Austria; 5Department of Radiology, Medical University Graz, Graz, Austria
Non-rigid image registration, independent from contrast changes in dynamic time series, is implemented using a TV-L1 optical flow algorithm to compensate physiological and unintentional motion to get a better quantification of DCE-MRI data. Contrast invariant structure images of the whole sequence are created and registered to a template image which is defined by the structure image of the first component of the principal component analyses. The performance of the algorithm was evaluated using real patient DCD-MRI tumor sequences where movement artifacts of tumor and surrounding tissue could be markedly reduced which subsequently improved the quantification of DCE-MRI data.