STFC received an allocation of 10 three year National Productivity Investment Innovation Fellowships as part of Government funding for the Industrial Strategy Challenge Fund (ISCF). Proposals were invited from departments, or consortia of departments, to host one or more Fellowships which started on or before April 2018.
The National Productivity Investment Innovation Fellowships support excellent researchers to work on research projects which build programmes of innovation and translation from research within the STFC core programme remit. Projects may cross disciplines, particularly across the UKRI remit, and are broadly aligned with the core themes, challenges and opportunities highlighted in the Industrial Strategy Green Paper.
Research is focused on providing interdisciplinary answers to complex problems as well as addressing focused industry and sectoral needs through, for example, proof of concept and market research to the social and economic benefit of the UK.
STFC Innovation Fellow, University of Hertfordshire
We are fast approaching a new era of Astronomy, the Peta-byte era. As new surveys, telescopes and simulations come online, we are generating more and more data that cannot possibly be analysed and categorised by humans. To tackle this, it is necessary to develop Machine Learning algorithms that can sort through the masses of data being generated. Machine learning is the practice of developing models using large data sets and rules that the computer uses to learn about the data. By repeating the same task, and modifying the results, computers can learn to reproduce known classification systems, or come up with their own. This is in contrast to more traditional classification methods based on mathematical equations or those that rely on humans, such as citizen science and expert classification.
There are two broad families of ML algorithms, supervised and unsupervised (sometimes referred to as trained and untrained). In supervised learning, the algorithm is given a large data set it can learn patterns from, which it then applies to new unseen data. In the unsupervised case, the data is analysed without a training set, so the computer is coming up with its own labels, clusters and classifications. The advantage of unsupervised learning is it does not require a large input data set, and can be readily deployed on new data without lengthy classification and training from old data. The downside is the computer cannot qualitatively describe its data clusters, it is just capable of telling us which groups of data are most closely related to the others. It requires human intervention to label to classification groups.
At Hertfordshire, as an Innovation Fellow, I am developing an unsupervised machine learning algorithm which can be deployed on a variety of image classification problems. The principle among these is that of galaxy identification and classification in large imaging surveys, such as those from the Hubble Frontier Fields, or from new data sets from the upcoming Euclid and LSST programs. Finding and assessing the colours, morphologies and other properties of galaxies in new data sets can be done in many ways using machine learning, but they mostly rely on input from previous survey data to lay out the groups the galaxies will be sorted into. For example, one could develop a neural network which classifies the shapes of galaxies by using the Galaxy Zoo 2 dataset as a training example. The algorithms developed using unsupervised techniques will be free from this requirement, and won’t carry any of the biases inherent in these old datasets.
The real benefit of developing unsupervised image classification systems, however, is their applicability outside of their intended design. As the algorithm is not trained on a specific set of astronomical images, it could be used in a variety of other situations. For example, it could analyse molecular clouds and hydro dynamical simulations. Or, it could be used outside of astronomy, in remote sensing applications such as agricultural satellite imaging, RADAR imaging of the Earth or other planetary surfaces, SONAR imaging of the oceans or even in autonomous vehicles. My work as Innovation Fellow will develop the algorithm designed for astronomical purposes into a system that can be applied to a wide range of imaging problems.
Dr Charlie Jeynes
STFC Innovation Research Fellow
Title of Research: Applying an astrophysics modelling tool to improve the diagnosis and treatment of cancers using theranostic nanoparticles
Theranostics is an exciting new avenue in medicine that seeks to combine diagnostic and therapeutic techniques as a single treatment.
These are generally based around nanoparticles that both act as imaging agents as well as in a therapeutic capacity. An ideal theranostic agent is non-toxic, accumulates preferentially in tumours and gives excellent contrast with medical imaging techniques like MRI or CT. Further, an external energy source directed at the tumour will ‘activate’ the accumulated theranostic, for instance releasing a drug or inducing hyperthermia.
Although this external source can take a few forms such as magnetic, ultra-sound or X-rays, in this fellowship, I am focusing on using Near Infra-Red light as the implementation is cheap, convenient and used widely in photodynamic therapy.
Many hurdles remain before theranostic technology is seen in the clinic. Critical to its success is the concentration of the theranostic within the tumour, and subsequent ‘killing dose’ that it receives once an external light source is applied.
Computional modelling is a crucial tool here to quickly and cheaply simulate theranostic concentrations and irradiation dose scenarios in which good imaging contrast can be achieved while subsequent treatment regimes give acceptable tumour control.
The University of Exeter is home to the TORUS Monte Carlo code that has been used to model a wide variety of astrophysical radiative transfer phenomena. Prof Time Harries and his team have recently extended this code to include the transfer of light through tissue, and I am applying and modifying (ARC)-TORUS for theranostic applications.
To help me achieve this, I am working with experts in clinical Photodynamic Therapy (Alison Curnow, Exeter Medical School and Royal Cornwall Hospital), and theranostic experimental design Prof Nick Stone (University of Exeter).
In addition, I am investigating the commercial appetite within medical physics for a tool that can accurately and swiftly model the transport of non-ionising radiation through heterogeneous tissue that predicts dose and imaging quality in a broad range of medical scenarios.
This work complements my expertise in quantitative imaging of single nanoparticles in human tissues using super resolution microscopy and X-ray microanalysis.
Dr David Mahon
STFC Innovation Fellow, University of Glasgow
Title of Research: Development of a portable muography system for civil engineering and nuclear applications
Muography is one of the fastest-growing fields in applied particle and nuclear physics. This passive imaging technique uses naturally-occurring background radiation in the form of cosmic-ray muons, to inspect the contents of complex, shielded structures that cannot be investigated using conventional imaging techniques such as X-rays. Using information from the Coulomb scattering of muons within an unknown volume, and by inferring its most probable vertex, a 3D image based on material density can be reconstructed. This innovative technique is now being used to help address some of the most complex challenges within society.
Since 2009, the University of Glasgow has pioneered the use of muography for the characterisation of nuclear waste containers. With funding from the Nuclear Decommissioning Authority and Sellafield, and in collaboration with the UK National Nuclear Laboratory, muon imaging technology and imaging software were developed and tested with a series of nuclear industry samples. During these successful tests, small fragments of fuel material were identified within shielded, concrete-filled containers. In 2016, this research was spun-out with the formation of Lynkeos Technology Ltd., which has subsequently commercialised this Muon Imaging System (MIS) and deployed its first machine on the Sellafield site.
David is a former Royal Society of Edinburgh Enterprise Fellow and is a Director of Lynkeos Technology. During this STFC RCUK Innovation Fellowship, David will develop a next-generation, portable MIS to address challenges within both the civil engineering and civil nuclear industries, including the structural health monitoring of ageing infrastructure such as bridges.
Dr Sarah Bugby
University of Leicester
The University of Leicester, with support from the STFC, has a unique heritage in space instrumentation and is well known for designing and building detectors of high energy photons for X-ray astronomy and planetary space missions. The detection of X or gamma radiation is also important for aspects of medicine, nuclear power and security, and for characterising materials in sectors like manufacturing, mining, and archaeology.
Specialist detectors are needed to be able to record and visualise high energy photons. Rather than a using a silicon CCD, high energy detectors may incorporate a scintillator stage, or use new semiconductor materials. Novel compound semiconductors are varied and many are 'tunable', able to exhibit different properties depending on their chemical makeup. They can be far more sensitive to high energy photons than silicon, or may be able to operate at far higher temperatures. My research investigates the performance of new detector types for a range of applications, but I have a particular interest in working with clinicians to bring new technologies into hospitals and benefit patients.
Since 2012, I have been working on the development of a portable gamma camera for medical imaging – originally based on technology developed at Leicester for the XMM-Newton EPIC X-ray astronomy camera. Medical gamma cameras, important for diagnosis in a range of conditions including cancers, can take up a whole room in a hospital. Our portable camera is handheld and so it can be brought to the patient wherever they may be - even in the operating theatre. The camera visualises the size and shape of radioactive sources and combines this with an optical image for localisation. This could make a big difference to surgeons, who currently perform radioguided surgery with a non-imaging detector.
As an STFC RCUK Innovation Fellow I will be working with clinicians and industrial collaborators to move the portable gamma camera from small scale clinical pilots to surgical use. As the detection of high energy photons has a wide array of applications, I am also adapting the technology for use in different sectors such as small animal imaging and nuclear decommissioning. My fellowship will also allow me to investigate novel materials and detector designs, and the integration of these into medical technologies.
Tel: +(44) (0)1793 413195
Email: Fellowships Office
Last updated: 04 March 2019