Slice-selective parallel transmit magnetic resonance imaging at 7T

Abstract

Ultra-high field (UHF) Magnetic resonance imaging (MRI) at 7 Tesla and beyond provides gains in signal-to-noise ratio (SNR), enabling faster and higher-resolution imaging. However, UHF MRI comes with significant technical challenges, especially radiofrequency (RF, or B+1 ) field inhomogeneity which produces signal dropouts and contrast losses. Parallel transmission (pTx) technology, employing multiple transmit channels to dynamically control RF pulses, is a promising solution to overcome these limitations. While pTx has shown maturity in 3D neuroimaging with universal pulses, its application to 2D imaging suffers from optimisation complexities, scanner integration hurdles, and specific absorption rate (SAR) constraints, hindering translation into routine clinical use. The overarching aim of my PhD is to develop fast and robust pulse design methods for 2D pTx imaging, with the long-term aim of bringing pTx scans into the routine workflow of neuroscience studies and clinical applications. My thesis investigates three main topics: pulse optimisation efficiency and robustness, sequence and scanner integration, and SAR management. Central to this work is the development of Bayesian Optimisation of GrAdient Trajectory (BOGAT), a fast, robust algorithm for 2D spokes pTx pulse design, in Chapter 4. By decoupling gradient trajectory and RF optimisations, BOGAT reduces computation time to under 10 seconds while achieving near-optimal solutions (<1.1x global minimum cost) in 93% of cases. BOGAT pulses improved flip angle homogeneity and EPI image quality in volunteers. Complementing the BOGAT pulse designs, in Chapter 6, a scanner software Sequence Building Block (SBB) was developed to streamline spokes pTx pulse implementation on Siemens Terra scanners. The effect of hardware timing imperfection on spokes pulses was explored in detail and the correction methods were integrated in the SBB. The software module enables improved integration with common imaging sequences and accelerates the sequence preparation with spokes pulses. Chapter 5 applies existing pTx methodology to diffusion imaging, showing that 2-spoke pTx pulses reduced diffusion fitting uncertainty by 6.2% and enhanced crossing-fibre resolution compared to conventional cirucularly polarised (CP) mode scans. However, it highlights the trade-off between image homogeneity improvements and SAR constraints, and presents evidence of how novel pulse designs like BOGAT and coils with full virtual observation point (VOP) supervision can alleviate these challenges. Additionally, it demonstrates the practical challenges of implementing pTx in clinical settings. The new BOGAT and spokes SBB developments are exemplified in 2D turbo spin-echo (TSE) imaging in Chapter 7. Specifically, I apply a novel scalable spokes pTx pulse form derived from Average Hamiltonian Theory. The scalable pulses preserved phase patterns, so when used as both excitation and refocusing pulses, they better satisfy the CPMG conditions. This improves hippocampal T2-weighted image homogeneity compared to CP mode and existing dynamic RF shim optimisations. SAR constraints remain a practical limitation, particularly for our Nova 8Tx32Rx head coil lacking full VOP supervision on Terra scanners. Chapter 8 reviews current SAR supervision approaches and presents empirical power cutoff comparisons with legacy systems to justify an initial increase in the per channel power limit from 1 W to 1.5 W. Finally, Chapter 9 investigates the diversity of ‘good-enough’ RF shim solutions arising from magnitude least squares (MLS) optimisation. This may facilitate future pulse design research. In summary, the methods that I have developed enable simpler, quicker, more robust applications of pTx methodology for key 2D neuroimaging sequences (e.g. diffusion EPI and TSE). My hope is that these will stimulate wider availability of these high-fidelty whole-brain scans that are urgently needed to advance clinical and neuroscience research.