|
Title
|
|
|
|
Strategies for efficient acquisition and reconstruction of structural and quantitative MRI
| |
|
Author
|
|
|
|
| |
|
Abstract
|
|
|
|
| Magnetic resonance imaging (MRI) is a valuable tool for investigating the brain, both in the clinic and in research. Its main drawback is its long acquisition time, which can be shortened by minimising the amount of data which is acquired and by using it efficiently. Three objectives were set for this thesis, each of which makes efficient use of the acquired data. Two objectives relate to compressed sensing (CS). CS algorithms allow the reconstruction of signals or images from incomplete data, by acquiring this data in a pseudo-random way and finding the most compressible image still consistent with the acquired data. The final image quality depends on which data is sampled, and several acquisition strategies were tested on both simulated signals, test objects, and using actual acquisitions. Firstly, CS was implemented for a fast spin echo (FSE) sequence, and it was found that optimal results occur when the majority of the data is sampled in the centre of k-space, with progressively less data being sampled closer to the edge of k-space. Secondly, a method to accelerate the acquisition of isotropic in-vivo high-angular radial diffusion imaging (HARDI) data using an FSE sequence was developed. While slow, the FSE sequence results in few artefacts, and CS can be used to acquire the large number (>50) of volumes required for HARDI faster. In this case, CS can also be used to subsample the diffusion signal, in the q-space, to acquire fewer volumes. Using knowledge from the first objective, strategies for subsampling the data, either only in q-space or both in k-space and q-space, were tested. We found that subsampling the q-space only is most efficient, and 15 to 20 volumes proved sufficient to yield high quality reconstructions without major differences in quantitative diffusion measures or tractography results. Finally, a method was developed to quantify micron-sized iron-oxide particles (MPIO) in MRI-images. These particles are used to label cells and cause a negative contrast, i.e. hypo-intensities. However, positive contrast images can be constructed to improve the visibility and localisation of MPIO, which was done efficiently from a single acquisition. To quantify MPIO robustly and reliably the normalized average range (nAR) is introduced. The nAR compares the average value of regions of interest (ROI’s) to that of a control ROI in range filtered images, and showed a higher sensitivity in optimized positive contrast images and is not sensitive to the bias field of the receiver coil. |
| |
|
Language
|
|
|
|
English
| |
|
Publication
|
|
|
|
Antwerp
:
University of Antwerp, Faculty of Pharmaceutical, Biomedical and Veterinary Sciences
,
2021
| |
|
Volume/pages
|
|
|
|
270 p.
| |
|
Note
|
|
|
|
:
Verhoye, Marleen [Supervisor]
:
Sijbers, Jan [Supervisor]
| |
|
Full text (open access)
|
|
|
|
| |
|