Advisory Committee Chair
Donald B Twieg
Advisory Committee Members
Narasimha A Akella
Stanley J Reeves
Document Type
Thesis
Date of Award
2008
Degree Name by School
Master of Science in Biomedical Engineering (MSBME) School of Engineering
Abstract
A novel and recently developed fast magnetic resonance imaging (MRI) technique, Single-Shot Parameter Assessment by Retrieval from Signal Encoding (SSPARSE), promises significant improvements in robustness and accuracy of local signal parameter estimates compared to the conventional MRI methods. In using a more accurate signal model, the reconstruction for SS-PARSE differs from the traditional Fourier transform method. An iterative estimation algorithm, progressive length conjugate gradient (PLCG), is currently employed by SS-PARSE to reconstruct independent parameter maps of local magnetization, transverse relaxation rate and frequency. In practice, the large number of degrees of freedom and partial poor conditioning of the problem itself brings tremendous difficulties for PLCG convergence. In this thesis, investigations of the PLCG convergence were executed using simulations and experiments. Simulation studies were performed on a numerical clock shaped phantom to explore the convergence envelope, which contains the successfully “converged” parameter values but not the partially or “non-converged” parameter values. Also, various PLCG control parameter settings are tested, and the best setting in terms of the convergence envelope is a steadily increasing data length with a small regularization operator. Experimental four-tube phantom results confirmed the convergence characteristics predicted by simulation studies. Further, a signal-area based clustering iii method is presented which segments and eliminates background noise and unreliable estimates where local convergence failed.
Recommended Citation
Li, Ningzhi, "The Convergence Envelope For Iterative Estimation In SS-Parse" (2008). All ETDs from UAB. 3603.
https://digitalcommons.library.uab.edu/etd-collection/3603
