Advances in magnetic resonance imaging reconstruction methods incorporating prior knowledge

Student thesis: Doctoral ThesisDoctor of Philosophy


In magnetic resonance (MR) imaging it is usually assumed that the acquired data
represent samples of the Fourier transform of the object. However, there are
many cases where this assumption is violated due to system imperfections, subject motion and deliberate undersampling in an effort to reduce scan times. This can lead to artefacts in the reconstructed images.
In this thesis two emerging clinical applications of MRI are considered: hybrid
PET-MR imaging and coronary MR angiography (CMRA). Factors giving rise to
substantial deviations from the basic Fourier model in these cases are described
and methods to reduce artefacts by incorporating additional information into the
reconstruction are presented. This information is either in the form of additional
measurements, or as sparsity priors.
Two major limitations of current techniques for PET attenuation correction using
ultrashort echo-time MRI are addressed: Artefacts due to eddy currents and prohibitively long scan times. To account for eddy currents the use of a magnetic
field camera to measure the true k-space trajectories is proposed. The method
is demonstrated in numerical and tissue phantoms and in vivo cranial imaging of
healthy volunteers. Parallel imaging and compressed sensing are then explored
to accelerate the acquisition.
A method to improve motion correction for CMRA is also proposed based on
a novel image navigation scheme. This method uses a golden radial trajectory,
which provides both high-temporal-resolution translational and low-temporalresolution affine motion estimates from the same navigator data. The approach is demonstrated in healthy volunteers, leading to improved depiction of the coronary arteries compared to when correcting only for translational motion. Furthermore, the proposed method gives rise to a predictable and reduced scan time compared to a gated diaphragmatic navigator scan, while maintaining a high image quality.
The advances in the reconstruction that are proposed in this thesis help to tackle
some of the major problems with UTE-based attenuation correction for PET and
with CMRA. The proposed methods help to bring these emerging applications of
MRI towards routine clinical practice.
Date of Award2015
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorClaudia Prieto Vasquez (Supervisor) & Tobias Schaeffter (Supervisor)

Cite this