High resolution imaging of tissue properties with optimized precision using Ultra High Field MRI

Student thesis: Doctoral ThesisDoctor of Philosophy


Quantitative MRI (qMRI) aims to directly estimate tissue parameters rather than obtain images weighted on them, enhancing reproducibility of the results. However, long acquisition times have hampered its translation to clinics. Many methods have been proposed to make acquisitions more efficient, with a multitude of qMRI methods available in the literature. On the other hand, ultra-high field (UHF) provides increased signal-to-noise ratio (SNR) that can allow higher resolution imaging and/or shorter acquisition times. However, at UHF there is also increased B+1 and B0 inhomogeneity and higher SAR deposition. Parallel transmit (PTx) combined with more advanced pulse design methods is a flexible solution to overcome B+1 inhomogeneity. During my PhD I investigated how to compare different qMRI methods and developed two RF pulse design methods for pTx systems to overcome B+1 inhomogeneity in Magnetization Transfer (MT) imaging.

The performance of a qMRI method depends on several factors such as SNR and sensitivity to the parameters. To facilitate the comparison of qMRI methods for their merits only and normalise for external factors, an efficiency metric was presented based on previous comparison studies. A comparison of different qMRI methods for T1 and T2 brain mapping was performed, focusing on two main classes: (1) methods using steady-state sequences (e.g., Joint System Relaxometry) and (2) methods using transient sequences (e.g., MR Fingerprinting). Transient methods were found to be up to 3.5 times more efficient than steady-state methods for both T1 and T2 mapping. However, transient methods like MR Fingerprinting commonly employ large undersampling factors that if used with suboptimal reconstructions were found to undermine the potential gains they offer in effciiency.

Whereas increased SNR at UHF can benefit qMRI, B+1 inhomogeneity can considerably degrade parameter estimation. Advanced pulse designs for pTx systems can help mitigating this problem. However, current pulse design methods only consider the magnetization rotation as described by the Bloch equations and are not suitable for systems with a semisolid component that have MT. The concept of flip angle does not apply for semisolid magnetization as it follows a different physics from the Bloch equation. Instead the semisolid magnetization saturates in a rate that depends on the root-mean-squared B+1 (Brms 1 ). We proposed a novel pTx pulse design called PUlse design for Saturation Homogeneity (PUSH) that employs saturation sub-pulses applied off-resonance and considers the Brms 1 distribution they create. PUSH pulses gave more uniform MT contrast in-vivo for a variety of scenarios and also higher contrast than using the circular polarized (CP) mode under the same operational constraints.

Although MT contrast is commonly obtained by employing off-resonance MT preparation pulses (like PUSH), 'excitation' pulses that aim to generate transverse magnetization also apply some power and create 'incidental' MT contrast. This can be particularly problematic for T1 mapping, where several studies have reported estimation bias. We proposed a new generalized radiofrequency pulse design that homogenizes both flip angle and Brms 1 spatial properties from an RF pulse. This design is a combination of current pulse designs that consider only flip angle and of PUSH that considers only Brms 1 . This is shown to control incidental MT effects from on-resonance pulses in T1 estimation, obtaining consistent T1 estimates.

Date of Award1 Dec 2022
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorJo Hajnal (Supervisor) & Shaihan Malik (Supervisor)

Cite this