Image reconstruction in dynamic contrast enhanced magnetic resonance imaging

Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) is a powerful tool to detect cardiac diseases and tumors, and both spatial resolution and temporal resolution are important for disease detection. Sampling less in each time frame and applying sophisticated reconstruction methods to overcome image degradations is a common strategy in the literature.;In this thesis, temporal TV constrained reconstruction that was successfully applied to DCE myocardial perfusion imaging by our group was extended to three-dimensional (3D) DCE breast and 3D myocardial perfusion imaging, and the extension includes different forms of constraint terms and various sampling patterns. We also explored some other popular reconstruction algorithms from a theoretical level and showed that they can be included in a unified framework.;Current 3D Cartesian DCE breast tumor imaging is limited in spatiotemporal resolution as high temporal resolution is desired to track the contrast enhancement curves, and high spatial resolution is desired to discern tumor morphology. Here temporal TV constrained reconstruction was extended and different forms of temporal TV constraints were compared on 3D Cartesian DCE breast tumor data with simulated undersampling. Kinetic parameters analysis was used to validate the methods.;2D imaging with serial acquisition of different slices is regularly used for myocardial perfusion imaging. 3D imaging has potential advantages including robustness to through plane motion, and accuracy of sizing ischemia. Here 3D stack-of-stars sampling with spatiotemporal TV constrained reconstruction is developed and is shown to be a promising alternative for myocardial perfusion imaging.;Other groups proposed a number of reconstruction algorithms for undersampled MRI recently, including HYPR-LR, PR-FOCUSS, k-t BLAST/k-t SENSE, k-t FOCUSS and regularized iterative SENSE. The work here reveals the relationships among these methods by incorporating these algorithms into a generalized reference image framework. Reconstruction of simulated data, as well as undersampled myocardial cine datasets and perfusion datasets, showed that the superiority of x-t and x-f reference image is sensitive to the data characteristics and baseline images.;All of the above efforts will lead to improvements in the diagnosis of diseases like myocardial ischemia and breast tumors, through improving image quality and better quantifying kinetic parameters.