Higher-Dimensional Extensions of Nonlinear Inverse Reconstruction for Magnetic Resonance Imaging

by H. Christian M. Holme
Doctoral thesis
Date of Examination:2019-09-19
Date of issue:2020-02-05
Advisor:Prof. Dr. Martin Uecker
Referee:Prof. Dr. Martin Uecker
Referee:Prof. Dr. Stefan Luther
Referee:Prof. Dr. Herbert Köstler
Persistent Address: http://dx.doi.org/10.53846/goediss-7750

 

 

Files in this item

Name:2019_hcmh_PhD_Thesis_webview_pub.pdf
Size:72.1Mb
Format:PDF
Description:Dissertation

The following license files are associated with this item:

Abstract

English

Even though magnetic resonance imaging (MRI) has become progressively faster in recent years, acquisition speed is still a problem in current clinical settings. Physiological constraints such as gradient-induced peripheral nerve stimulation complicate further speed-up of the acquisition process. Therefore, techniques for image reconstruction from undersampled data have been a research focus, among them parallel imaging and compressed sensing. This thesis investigates how multi-dimensional extensions to regularized non-linear inverse reconstruction (NLINV), an established parallel imaging technique, can be used to increase reliability, flexibility, and robustness in image reconstruction. In contrast to model-based approaches, this is done without using specific assumptions about the underlying physical laws. The idea is explored in different experimental settings such as imaging with phase constraints, phase singularities, non-Cartesian acquisitions and real-time MRI, and cardiac self-gating. In particular, the extension of NLINV to compressed sensing is explored.
Keywords: magnetic resonance imaging; parallel imaging; real-time MRI; compressed sensing; NLINV; ENLIVE
 
Statistik