| dc.description.abstract |
<p>Diffusion-weighted MRI (DWI) is an essential tool in clinical applications such
as detecting ischemic stroke, and in research applications studying neuronal connectivity
in the brain. Diffusion-weighted imaging with multi-shot echo-planar acquisition (DWEPI)
offers several advantages over single-shot EPI, including improved spatial resolution
and reduced off-resonance and susceptibility artifacts. However a major limitation
of multi-shot DWEPI is its sensitivity to patient motion during the application of
diffusion gradients; the motion generates phase errors which vary from shot to shot,
resulting in artifacts in the reconstructed image. </p><p>Most current methods for
correcting motion-induced artifacts involve the use of navigator echoes to estimate
the shot-to-shot phase errors. Accurate navigator information comes at the expense
of increased scan times however, which is generally undesirable. The aim of this study
is to therefore develop and demonstrate the use of an alternative phase estimation
technique, iterative phase-cycling, as a new method to correct motion artifacts in
DWI without the use of navigators. The developed method involves an iterative column
by column estimation of phase errors in the aliased image and reconstruction of an
artifact free image using the estimated errors.</p><p>In this study the technique
was applied to correct artifacts in simulated images, hybrid-simulated images, and
true four-shot DWEPI images. Accuracy of the phase-cycling method was evaluated by
computing residual image errors and ghost-to-noise ratios after correction. The efficiency
of phase-cycling was evaluated by recording computation times of the correction process.
Multiple optimization techniques were developed and used in the experiments, and the
accuracy/efficiency of these techniques were also assessed.</p><p>Results of the experiments
demonstrated the ability of phase-cycling to greatly reduce motion-induced artifacts
in multi-shot DWEPI at reasonable computation times. The phase-cycling model used
in this study accurately estimated linear and nonlinear errors along the frequency
encoding direction and linear errors along the phase encoding direction. An additional
mathematical framework is presented illustrating the potential of phase-cycling to
correct nonlinear errors along the phase encoding direction in future work.</p><p>The
study establishes the developed technique as a unique and effective method for correcting
motion artifacts in DWEPI without the cost of increased scan times.</p>
|
|
