Show simple item record

On the Use of Temporal Information for the Reconstruction of Magnetic Resonance Image Series

dc.contributor.advisorFrahm, Jens Prof. Dr.
dc.contributor.authorKlosowski, Jakob
dc.titleOn the Use of Temporal Information for the Reconstruction of Magnetic Resonance Image Seriesde
dc.contributor.refereeFrahm, Jens Prof. Dr.
dc.description.abstractengStandard clinical magnetic resonance imaging uses data sampled at the Nyquist rate which limits it to frame rates of 1-4 frames per second. Over the past decade different methods for the reconstruction of images from sub-Nyquist sampled data have been proposed, raising frame rates beyond 30 frames per second and opening the door to real-time cardiac imaging, the study of speech and many more. Practically all such methods exploit spatio-temporal structure in the data as prior knowledge to an inverse problem. This work combines the method of non-linear inversion reconstruction and motion estimation by an optical flow, a model for in-plane motion that is frequently used in computer vision tasks. The model is quantitatively compared against simpler methods based on temporal finite difference schemes in simulations and real cardiac data. Adding an optical flow constraint to the reconstruction’s cost functional is shown to be advantageous in situations with purely in-plane motion while failure of the model is observed in the form of artifacts and blur during through-plane motion. A point that got little attention in related literature. The model’s limits are examined in detail and pointers to possible extensions to overcome them are given. Further, a simple centered finite temporal difference constraint for a batch of frames is shown to provide better image quality than a previously established scheme that considers only the backward temporal difference at a single frame. As in previous works, non-linear inversion reconstruction considers at most a small batch of frames for every time point. The method does not require availability of the entire image data series at any point during computation. It is thus conceptually capable of providing images in real-time simultaneously with the data acquisition, a prerequisite for interactive
dc.contributor.coRefereeEnderlein, Jörg Prof. Dr.
dc.contributor.thirdRefereeGrubmüller, Helmut Prof. Dr.
dc.subject.engReal-time Magnetic Resonance Imagingde
dc.subject.engCardiac Imagingde
dc.subject.engInverse Problemsde
dc.subject.engOptical Flowde
dc.affiliation.instituteGöttinger Graduiertenschule für Neurowissenschaften, Biophysik und molekulare Biowissenschaften (GGNB)de
dc.subject.gokfullBiologie (PPN619462639)de

Files in this item


This item appears in the following Collection(s)

Show simple item record