Creating high fidelity digital tissue substrates for the development of non-invasive microstructural MRI

Cancer is one of the developed world's leading health care challenges for the 21st Century. Developing new imaging techniques that allow suspected cancers to be diagnosed without surgery, or determine early on whether a treatment is working, are important ways we may tackle this challenge.
Magnetic resonance imaging (MRI) is a technique that is already widely used in hospitals. MRI is often used in cancer to see if a tumour is growing or shrinking in response to treatments like chemotherapy or radiotherapy. New ways to use MRI are always being developed; one new development is a technique called microstructural MRI.
This technique allows researchers and clinicians to measure how the structure of tissues in the body are changing. For example, measuring the size of cells, or the orientation of blood vessels in a particular section of tissue. These measures of tissue structure are an important indication of healthy function or disease progression. In diseases like cancer, samples or biopsies of a tumour are taken by a surgeon. The biopsy is looked at under a microscope by a histologist who can use the shapes and sizes of the cells and blood vessels to diagnose or grade a cancer's severity. Being able to make these measurements with MRI would mean that tumour biopsies could be done easily without surgery. In addition, subtle changes in a tissues structure can indicate that a tumour is responding to a particular treatment long before a conventional MRI would detect tumour shrinkage. Using microstructural MRI for this purpose could help to identify treatments which aren't working, and allow them to be changed more quickly, hopefully leading to a better outcome for that patient.
Microstructural MRI has already been tried in the clinic with some success, however its use so far has highlighted that certain tissue structure measurements seem to have low reproducibility or vary in unexplained ways. I am aiming to investigate this variability. The way in which a tissue's structure can be found using MRI relies on a computer model to determine what tissue structures would have produced the MRI signal that is acquired. In order to investigate why particular outputs are varying, data are needed where the actual tissue structure or ground truth is known and can be compared to the MRI outputs. Ideally, this known tissue structure would be one that could be changed in a precise way to then see how the known change has affected the MRI output. However, acquiring these known tissue structures or ground truths is very difficult.
To get around this problem of missing ground truth data researchers have used virtual MRIs. A virtual MRI can take any digitally constructed shape and simulate the MRI you would acquire from it. These digitally constructed shapes can be made to look like tissues and can be easily manipulated by to investigate certain types of changes in tissue structure. Up till now researchers have used very simple digital shapes to represent tissues, for example a collection of perfect spheres to represent cells and cylinders to represent blood vessels. This does not represent the complexity of real tissue and many researchers believe this may be the reason for the difficulties in the clinical trial thus far.
My fellowship aims to create a library of these digital tissues which are based on the shapes I will extract from real tissue. To do this I will use a 3D imaging technique that I have developed. It uses a microsope to take 3D images of tissues at the scale of single cells for whole small tumours (~2cm3). I aim to extract the shapes of the cells and blood vessels from these images which I will then turn these into digital versions through a computer model. These digital tissues can then be used to conduct many virtual MRIs to understand the complex relationship between the MRI signal and the underlying tissue structure. This approach combines the flexibility of digital tissues with the complexity of real tissue structure.

Diffusion-weighted MRI (DW-MRI) is a technique used to aid in the clinical diagnosis of cancer and ischemic stroke. Recent application of microstructural tissue models to DW-MR signals has enabled quantitative assessment of tissue parameters e.g. cell size or vascular volume, and has been applied in clinical feasibility studies. Digital tissue substrates (DTS) have been central to the development of microstructural MRI. Synthetic DW-MR signals can be simulated from a DTSs and microstructural parameters extracted. The DTSs then provide a ground truth to evaluate the extracted parameters. As DTSs can be easily manipulated, they provide an ideal tool for systematically investigating how DW contrast is dependent on tissue geometry; and for developing new methods of DW-MR signal quantification, such as machine learning. Current DTSs consist of collections of spheres and cylinders to represent cells and vessels respectively. There is growing realisation in the field that these simple geometries may not capture key changes in tissue microstructure associated with certain pathologies; and cannot provide adequate validation. Recent advances in three-dimensional optical imaging provide a unique opportunity to create empirically-derived DTSs with much greater scale, complexity and realism than has previously been achieved. These optical imaging techniques (optical projection tomography (OPT) and high-resolution episcopic microscopy (HREM)) can generate image data from large tissue volumes (~2cm3) at sub cellular resolution (typically ~1um). In this fellowship, I will create a library of DTSs using 3D optical imaging data and a 3-dimensional cellular Potts modelling platform. To create such a library, I will extract the morphometric parameters from 3D imaging data using machine learning segmentation. These parameter spaces will then be used to parameterise the computational model, which can then generate DTSs within a particular parameter space.

This research will develop a digital library of tissue geometries which in turn will be used aid in the translation of microstructural MRI for clinical use in cancer. This work addresses the current barriers to that translation by providing a computational resource that can be used to validate and develop existing microstructural MRI. Aside from the academic beneficiaries there are several groups that could benefit or be impacted by this project.
Health Benefit to Patients: The development of microstructural MRI aims to enable the assessment of tumour aggressiveness or response to treatment without the need for invasive biopsies or ionising radiation. Such a technique could improve the outcome for patients by identifying ineffective treatment regimens sooner. It could also reduce the need for surgery, with the associated risks and disruption to normal life. This impact would ultimately be over the long term, of the next 20-30 years, however, given that the technique is already in early-phase clinical trials, there is potential impact in the next 5 years for patients in further clinical trials.
Benefit to the UK economy through industry: Extension of the use of MRI within clinical practice or preclinical research will directly benefit the MRI industry over the mid and long time-scales. The MRI industry was used as a case study by Oxford Economics in 2012 and was conservatively estimated to contribute between £587 and £685 million to UK GDP in total from 2010-2015. Development of MRI techniques will also aid in ensuring the competitiveness of the UK in the global MRI market which was estimated to grow to around £6.2 billion by 2015. In addition to the MRI industry, there may be long term impacts for the drug development industry which is also an important contributor to the UK economy. Providing drug companies with imaging technologies that allow for non-invasive monitoring of treatment responses is applicable to a huge range of potential new therapies. Microstructural MRI may provide new insights into drug response time-courses, enable the earlier identification of non-effective drugs and reduce the use of animal testing through enabling longitudinal studies.
As a key component of the UK's economy the benefit to the NHS is also important to highlight. Recent studies into the economic cost-benefit of multiparameteric MRI in prostate cancer have shown an economic benefit if the correct patient cohort is selected. The development of microstructural-MRI may be expected to have a similar impact on the NHS in the long term.
Impact on Clinicans/Healthcare professionals: This research will ultimately impact clinical practice through providing additional aids to clinical decision making. Clinicians will be both impacted and critical to the successful translation of this research. This group of stakeholders should be engaged with throughout the project itself, with the understanding that there will be short term impact around awareness of the research, mid-term impact in carrying out clinical trials, and educating other clinicians; and long term impact through potential change to routine clinical practice.
Benefit to young people: This research is an example of multi-disciplinary work and could impact the choices of the next generation of research scientist, clinician's and research funders. This could build capacity in multi-disciplinary science, and increase the awareness of the value of multi-disciplinary research for medical research. The impact may occur over long term timescales if applied to junior school age children (20-30 years), short or mid time-scales when considered in its application to university undergraduates (3-5years).