Accelerated Radial Magnetic Resonance Imaging: New Applications and Methods
AuthorBerman, Benjamin Paul
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractMagnetic resonance imaging is a widely used medical imaging technique, and accelerated data acquisition is critical for clinical utility. In this thesis, new techniques that incorporate radial acquisition, compressed sensing and sparse regularization for improved rapid imaging are presented. Sufficiently accelerated imaging methods can lead to new applications. Here we demonstrate a solution to lung imaging during forced expiration using accelerated MRI. A technique for dynamic 3D imaging of the lungs from highly undersampled data is developed and tested on six subjects. This method takes advantage of image sparsity, both spatially and temporally, including the use of reference frames called bookends. Sparsity, with respect to total variation, and residual from the bookends, enables reconstruction from an extremely limited amount of data. Dynamic 3D images can be captured at an unprecedented sub-150 ms temporal resolution, using only three (or less) acquired radial lines per slice per time point. Lung volume calculations based on image segmentation are compared to those from simultaneously acquired spirometer measurements. Additionally, accelerated imaging methods can be used to improve upon widely used applications; we also present a technique for improved T₂-mapping. A novel model-based compressed sensing method is extended to include a sparse regularization that is learned from the principal component coefficients. The principal components are determined by a range of T₂ decay curves, and the coefficients of the principal components are reconstructed. These coefficient maps share coherent spatial structures, and a spatial patch--based dictionary is a learned for a sparse constraint. This transformation is learned from the coefficients themselves. The proposed reconstruction is suited for non-Cartesian, multi-channel data. The dictionary constraint leads to parameter maps with less noise and less aliasing for high amounts of acceleration.
Degree ProgramGraduate College