Back to Projects

Image Acquisition and Reconstruction (pre-2019)

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

Project ID: 2017_110

Student: David Leitao

1st supervisor: Shaihan Malik, King’s College London
2nd supervisor: Jo Hajnal, King’s College London

To design a new quantitative MRI approach for Ultra-High Field (UHF) MRI. UHF-MRI brings with it the promise of enhanced signal-to-noise (SNR) leading to very high resolution imaging. However UHF-MRI is restricted by stringent constraints on radio-frequency (RF) power absorption and hardware limitations, as well as highly spatially non-uniform RF fields.

The project will investigate approaches for optimizing the efficiency of quantitative MRI (qMRI) sequences in the context of UHF-MRI, with the aim of producing robust T1/T2 parameter maps within the shortest possible amount of time.

The tissue signal in MRI is in general a complex function of many factors including water content, relaxation times (T1/T2), macromolecular composition, macro and microvasculature, fat content, diffusion properties and many more. Conventional MR imaging uses standard protocols whose tissue contrast is ‘weighted’ towards one or more parameter, and radiologists interpret these from experience. Quantitative MRI (qMRI) instead aims to directly measure many of these important parameters, to directly quantify tissue properties. This offers simpler quantitative comparison of images between subjects and longitudinally, and for modelling of underlying disease processes. QMRI typically requires the collection of more data than normal imaging, hence acquired spatial resolutions are limited by lengthy acquisition times.

UHF-MRI (at 7T and above) promises increased signal to noise (SNR) which can be translated into very high resolution images. A new pan-London 7T MRI facility is planned to be installed at St.Thomas’ from late 2017. The facility will be accessible to clinical researchers from all over London, and a developed qMRI method would be applicable across the wide range of planned clinical research.  This project proposes to explore existing qMRI approaches and their application to UHF-MRI. There are particular challenges in doing this, particularly that hardware and safety constraints at UHF are much more stringent and that the radio frequency magnetic fields (B1) are highly spatially non-uniform.

There are two main approaches for qMRI: ‘standard’ approaches such as DESPOT1 employ pulse sequences constructed such that the received signal has a simple functional dependence on the sequence parameters and the quantitative parameters of interest. Safety constraints for UHF MRI will constrain the choice of sequence parameters, particular the flip angles that can be achieved when using balanced SSFP sequences, as DESPOT does. B1 variability means that tissues at different spatial locations within the same patient are effectively subject to different pulse sequences making it difficult to find an overall optimal operating point.

An alternative qMRI approach is ‘MR Fingerprinting’ (MRF)2. Here a randomly varying sequence is used, which does not give a simple relationship between signal and quantitative parameters. Instead signal simulations (via Bloch equations) are used to generate predicted signal ‘fingerprints’ for different tissue types and a dictionary based reconstruction is used to estimate the parameters. This method has been extended to the highly spatially variable B1 fields seen at UHF using the plug-n-play method3, using parallel transmission (PTx) MRI to improve quality of parameter estimation. However the complexity of the ‘fingerprints’ makes optimization of the information content of the sequence – as can be done for more standard qMRI4 not straightforward. Further, contamination from other effects such as magnetization transfer and diffusion mean that more complex models may be needed to obtain accurate quantitative information from MRF.

This project will investigate both types of qMRI from the perspective of optimizing information content (or achievable precision in parameter estimates) per unit time, within strict hardware and safety constraints at 7T.

Back to Projects