Diffusion-weighted magnetic resonance imaging techniques
The aim of this project was to compare and evaluate other, non-EPI, diffusion-weighted MRI (DWI) sequences, through imaging experiments, on a phantom and in vivo, (using a 0.95 T system) and computer simulations, and to develop improved DWI methodology which could be implemented on standard hardware. Pulsed gradient spin echo (PGSE) and diffusion-weighted STEAM are slow multiple shot sequences, with measurement times of several minutes. Both sequences are highly sensitive to patient motion, but motion artifact was virtually eliminated using navigator echo phase correction and EGG triggering when diffusion-sensitisation was in the phase-encoding direction. It was demonstrated that both sequences can provide high quality images and allow accurate and straightforward diffusion-coefficient measurement when an imaging time period in the region of 20-30 minutes is available and when diffusion-sensitisation is required in one or two directions. A third direction of diffusion-sensitisation may be feasible if more sophisticated immobilisation or phase correction techniques are employed. A choice between PGSE or STEAM for a given application should take account of the Ti and T2 values of the imaged tissues, since a higher SNR might be provided by STEAM when the T1T2 ratio is high. A diffusion-weighted CE-FAST sequence was implemented with the novel modification of acquisition of a navigator gradient-echo, which was shown to reduce motion artifact when diffusion-sensitisation was in the phase-encoding direction. However, it has been demonstrated by other workers that unknown signal losses due to motion-induced phase incoherence between signal components may remain. The SNR (normalised with respect to the square root of the imaging time) in the phantom and in white matter was similar to that obtained using PGSE, but an advantage of CE- FAST is that it can be performed in a fraction of the measurement time of PGSE. Diffusion-sensitivity was much higher than in other sequences and the diffusion- attenuation was found to agree with an analysis presented in the literature. However, a major disadvantage of the technique, which precludes its use for many DWI applications, is the requirement of accurate knowledge of Ti, T2 and flip angle in order to calculate the diffusion coefficient or tensor. Prior to a study of diffusion-weighted snapshot FLASH, the effects of magnetisation evolution during snapshot FLASH acquisition on image quality and parameter measurement accuracy were first investigated, through phantom experiments and computer simulations, in the context of a r2-weighted snapshot FLASH sequence. It was demonstrated that magnetisation evolution effects can lead to significant error in parameter measurement, but that this error can be eliminated by using crusher gradients to prevent evolved magnetisation from contributing to the acquired signal. However, qualitative effects are not entirely eliminated, since a significant degree of edge blurring may remain, and there is a 50% loss of SNR inherent to the crusher gradient technique. It was then shown, theoretically and experimentally, that in diffusion-weighted snapshot FLASH, the crusher gradient technique not only addresses the problem of magnetisation evolution, but also eliminates the effect of phase shifts arising during the diffusion-preparation sequence.