Towards Reliable Diffusion MRI of Moving Organs

Diffusion Magnetic Resonance Imaging (DMRI) is a clinical imaging technique that has the unique potential to provide clinically-relevant information without the use of ionising radiation or invasive procedures. Although DMRI is often used in brain imaging, it is not currently widely used in other parts of the body because of a series of technical challenges that this grant will address. Our approach will be to describe the challenges in a unified mathematical framework and solve this using new acquisition and image reconstruction techniques. The intensities in diffusion weighted MR images originate from the distances that water molecules can diffuse. These diffusion distances are affected by the local cellular environment and changes in the environment are reflected in images. Diffusion MRI can reveal the directionality of structures, such as the orientation of cardiac muscle cells and changes in cell density and organisation due to cancerous tumours. Cardiac disease and cancer are very significant health issues worldwide for which DMRI may provide key diagnostic or therapeutic information. Cardiac disease is the main cause of death (ca.30% worldwide), and cancer the third (ca.12%) with lung and liver cancer among the most common.The correct functioning of the cardiac muscle is dependent on the complex orientations of the fibres within the heart wall. Being able to image these in-vivo could lead to enhanced diagnosis, treatment and surgery planning for conditions such as heart failure, congenital defects or remodelling following a heart attack. In cancer, DMRI has the potential to improve diagnosis, aid localisation and grading of tumours, support treatment selection, better identify residual and recurrent disease following treatment, and even predict treatment outcome and reduce diagnostic errors. However, DMRI currently has a limited role outside of the brain because of five main technological restrictions; motion of organs, magnetic field variations due to the different magnetic properties of tissues such as fat, bone and air, magnetic field inaccuracies due to inherent MR scanner imperfections, long scan durations and the question of how to interpret the data. Previously we have described the effect of complex non-rigid motion on MR images using a clear mathematical framework and used this to correct for motion. In this grant, we propose to extend these methods to include the challenges limiting DMRI. Taking this general view allows the challenges to be solved collectively rather than sequentially. The techniques will improve the reliability of DMRI and thus widen its clinical uptake. At the same time, the techniques permit more efficient use of scan time, either to encode more information or to shorten scan durations. A general formalism provides a new way of viewing the problem and lends itself to making use of the tools available from other branches of computational mathematics. To support this approach, we will need measurements of magnetic field imperfections caused by MRI scanner inaccuracies. These will be provided by field mapping hardware, which will be installed for the first time in the UK. In addition, novel techniques such as compressed sensing and new models of tissue diffusion will be explored to reduce overall scan times, improve accuracy and provide better interpretation of data.

The ultimate beneficiaries of this research will be patients who are suffering from some of the most common lethal diseases: cancer and cardiovascular disease. These diseases still have a significant morbidity despite recent advances. The better quality, more reliable and more widespread clinical applications resulting from this research are likely to have a significant impact on the management of these diseases. This impact is only likely to be realised longer term as there is a requirement for developed techniques to undergo comparative clinical trials and be adopted by scanner manufacturers and healthcare professionals. For patients to benefit, clinicians must have both the desire and ability to use the methods we develop. Creating clinical demand requires both publicity and more detailed comparative trials to prove clinical benefit. We will continue to publish in leading peer reviewed journals and present at major conferences. One of the RAs has 12 months of time allocated to validation, data handling and testing on small numbers of patients, as appropriate for EPSRC funding. UCL and KCL are also partners in a Comprehensive Cancer Imaging Centre. These structures provide an ideal environment for translation and thus impact through joint working, seminars and close clinical involvement. To make the methods clinically accessible, in the long term they need to be incorporated by scanner manufacturers into their products. We will work closely with Philips, one of the three principal MRI scanner manufacturers to ensure that pulse programming code we write is compatible with their software development plans. KCL has a close collaboration with Philips (via Dr A Wiethoff, Philips KCL based scientist). More general reconstruction code will be made open source and we have experience of this at UCL with a diffusion data analysis package (www.camino.org.uk ). The reconstruction code will also be advertised on the MATLAB File Exchange and on the new site provided by the main international society dedicated to providing an infrastructure to promote the rapid and robust development of advanced methods for data collection and image reconstruction in MRI. (www.ismrm.org/mri_unbound) The pharmaceutical industry needs to evaluate responses to candidate drugs. More reliable, higher resolution, faster or validated diffusion imaging could act as a biomarker for response. Increased reliability reduces variation and thus enables the use of smaller sample sizes at reduced cost. More accurate data could enable drug response to be observed sooner and thus speed-up treatment, drug development and time to market. Also, our work will enable related techniques that use diffusion type imaging to infer microstructure and these could provide the pharmaceutical industry with more information about how their drugs are causing changes at the cellular level. Clinical researchers at our institutions and their immediate collaborators will have access to new imaging methods and reconstructions to form a better understanding of disease, treatment and the types of imaging required to optimise clinical management. Training will be one of the most immediate beneficiaries of this research programme. Those employed on the grant will be able to access the Post Graduate training at UCL and KCL (existing students and staff already attend relevant modules from courses such as the Medical Image Computing MSc and Medical Physicist training). The UK economic competitiveness will benefit through attracting further research funding. The groups of all three investigators have previously attracted investment and research contracts from Philips Medical Systems. Reliable imaging increases the chances that pharmaceutical companies will choose our clinical sites for their trials and this is a lucrative source of income for the UK and exploits the good research base already in the UK. Public engagement activites will enthuse a wider audience.