A biophysical simulation framework for magnetic resonance microstructure imaging

This project develops a simulation system for the MR signal in biological tissue and its dependence on molecular dynamics as influenced by tissue microarchitecture and composition. The system is an essential tool in the development of next-generation non-invasive imaging techniques. Specifically, it underpins the development and translation of the emerging paradigm of microstructure imaging. The paradigm uses mathematical models, which relate the MR signal to underlying tissue properties, to estimate and map those properties by fitting the models voxel-by-voxel to combinations of appropriately sensitised image data. The approach provides much greater biological specificity than standard MRI, thus enhancing diagnosis and treatment planning.

The current generation of microstructure-imaging techniques is now starting to find widespread application in clinical studies. Prominent examples include NODDI for neuroimaging and VERDICT for cancer imaging, both developed by the investigators on this project. Those techniques are based entirely on diffusion MRI and their extension and refinement within that single contrast mechanism continues rapidly. However, a new generation of microstructure-imaging technique is just beginning to emerge that draws on multiple sources of MR contrast, for example combining diffusion MRI with relaxometry, susceptibility, etc. Such techniques offer great promise in the decades to come for the realisation of 'virtual histology' avoiding invasive procedures, such as biopsy, across a wide range of medical applications.

EPSRC grant EP/E064280/1, which finished in 2011, developed the current state-of-the-art simulation system within the Camino toolkit. That system underpinned the early development of the microstructure-imaging paradigm, which led to current techniques like NODDI and VERDICT. However, the current system is insufficient to evaluate even current microstructure imaging techniques, because it excludes key effects that influence the diffusion MR signal. Moreover, its implementation limits the simulation to molecular diffusion as the only source of MR contrast, which fundamentally prevents its extension for validation of next-generation techniques.

The new simulation system will use more sophisticated underlying models of tissue geometry and MR signal generation enabling it to support both modern diffusion-based microstructure-imaging applications and future multi-modal techniques. It provides a unique and invaluable validation tool allowing us to realise the full potential of quantitative non-invasive imaging in medicine and beyond. Within the project we demonstrate the new system by evaluating the performance of NODDI and VERDICT under a wide range of conditions. We also test two early examples of multi-modal microstructure imaging techniques paving the way for their robust development and eventual clinical translation.

Impact of the project arises through facilitation of the development of accurate and well-validated microstructure-imaging techniques.

Microstructure imaging promises advances in understanding and management of some of the biggest challenges facing 21st century healthcare; most directly: neurological diseases and cancer. The UK dementia platform estimates annual socioeconomic costs of dementia in Britain at around £17B. A treatment prolonging independent life of the average dementia patient by just one year would save around £1B per year in care costs as well as boosting the UK economy through revenue from the treatment if realised through its pharmaceutical industry and thriving community of related SMEs. The potential of microstructure imaging is to detect and classify disease earlier, enabling appropriate and early intervention, and to stage disease more accurately supporting effective treatment development and deployment. Annual costs of cancer have similar scale. Microstructure imaging promises rapid, non-invasive, and specific early diagnosis, supporting precision medicine and treatment delivery with similar socio-economic impact.

Current microstructure-imaging techniques are rapidly becoming part of the mainstream battery of imaging techniques used routinely in clinical studies and exams. NODDI is a key component in large-scale data-collection initiatives, such as the 1946-cohort imaging project (~1000 subjects), and most current large-scale brain imaging projects, such as the UK Biobank (100,000 subjects) and the Human Connectome Project (1000s of subjects), use protocols designed to support the technique. The more recent VERDICT technique is already a key component of large-scale prostate-cancer imaging initiatives at UCLH, such as the PROMIS (1000s of subjects) and INNOVATE (100s of subjects) projects, and other institutions are preparing to follow suit.

Despite their success, intense debate continues in the technical imaging-science community about the mathematical models and acquisition protocols at the heart of the techniques. The debate arises from incomplete understanding of both the biophysical sources of MR contrast and the effects of simplifying modelling assumptions that are essential to make stable front-line techniques. The simulation tool we propose here advances our understanding in these key questions providing confidence in the conclusions drawn from large-scale studies and clinical trials, as well as highlighting areas of weakness in current techniques to ameliorate for future generations.