Summary: | Magnetic resonance imaging requires a homogeneous B<sub>0</sub> magnetic field, however the field is often distorted by the presence of the subject. Techniques to homogenise the magnetic field, so called shimming, are commonly used. Conventional shimming methods homogenise the magnetic field once before scans. Dynamic shimming is an advanced method of updating shim fields during scans for homogenising the B<sub>0</sub> magnetic field in small volumes, which improves field homogeneity to a higher degree. However, hardware limitations and shim calculation errors pose challenges to dynamic shimming. Methods to optimise the performance of dynamic shimming are investigated in this thesis. Minimising eddy current effects induced by rapid shim switching is one of the most important topics for dynamic shimming. In this work, we use pre-emphasis to compensate for eddy current effects. Determination of pre-emphasis parameters requires the knowledge of eddy-current-induced fields. Chapter 2 compares the performance of two eddy current characterisation methods, an image-based method and a method based on a state-of-art field camera. Dynamic shimming is implemented based on the eddy current compensation results. Dynamic shimming requires robust shim setting calculation methods to deliver high-quality and stable shimming results. In Chapter 3, we propose a fast and fully automated regularised shim current determination algorithm to simultaneously improve the current efficiency and the robustness of the shim calculation against noise, while preserving good shimming quality. Following this, to improve the robustness of shim settings to residual eddy currents, a fast and simple regularised shim setting optimisation method is proposed. Additionally, the performance of dynamic shimming using different cost functions (e.g., signal loss, BOLD sensitivity) is evaluated. In typical MRI protocols, time is spent on acquiring a field map to calculate the shim settings for best image quality. In Chapter 4, we propose a fast template-based field map prediction method that yields near-optimal shims without measuring the field. This technique is initially implemented by using an averaged template created based on a field map database. Additionally, subject-specific information, like brain geometry, age and weight, is considered and used for improving field map prediction accuracy.
|