Funded Projects

PhD Awards

A methodology for breast density measurement using the HEXITEC pixellated spectroscopic technology

Academic supervisor: Dr Silvia Pani, University of Surrey

Academic supervisor: Prof Philip Evans, University of Surrey

Partner supervisor: Dr Emma Harris, Institute of Cancer Research

Partner supervisor: Mr Matthew Wilson, STFC UKRI Rutherford Appleton Laboratory

Breast density describes the proportion of glandular and fatty tissues in a breast and is known to be an indicator of breast cancer risk. Women with dense breasts can be up to 6 times more likely to develop breast cancer in their lifetime. A reliable method to measure breast density would allow each woman to be allocated a personalised screening schedule, with highrisk women being screened more often than low-risk women. This would optimise breast screening, its associated cost, and ensure that women are not unnecessarily exposed to potentially harmful X-ray doses.
The project will develop a safer, more precise breast density measurement technique than mammography. We will use novel X-ray detector technology allowing simultaneous acquisitions of multiple images at different X-ray energies, giving better discrimination of tissues at a minimal dose. We will determine how this detector can be best used to obtain precise measures of breast density, and develop computer algorithms to analyse the information provided by multiple X-ray energies to provide a map of the thicknesses of glandular and fatty tissue across the breast. The algorithms will be developed using computer simulations, and tested in the laboratory on custom-developed test objects.

Portable hybrid gamma-optical camera for quantitative 3D precision imaging in cancer diagnosis

Academic supervisor: Dr Sarah Bugby, Loughborough University

Academic supervisor: Dr Georgina Cosma, Loughborough University

Partner supervisor: Dr Paul Cload, Serac Imaging Systems Ltd. (SIS)

Medical gamma cameras are used routinely, and the resulting images are of great value to the clinician in the staging and treatment of a variety of conditions. Some aspects of cancer diagnosis, such as staging to investigate whether cancer has spread to other areas of the body, require surgical intervention.
Prior to surgery, SPECT or PET cameras are used for surgical planning. However, these images can’t take into account the changes in patient position or surrounding tissue during surgery, and the imaging systems are far too large to be used during surgery. Instead, radioguidance is provided by non-imaging gamma probes via an audible signal. However, these probes can’t provide precise localisation of sources, cover the entire surgical field of view, or provide information on the depth of a source within tissue.
This project will build on STFC-developed technology – a portable high resolution hybrid gamma-optical camera - now being developed by SIS. We will extend this concept using stereoscopic imaging techniques and new image analysis processes, so that the system can provide depth information in real time.
This technique for intraoperative gamma imaging would be applicable to a range of cancer diagnosis and staging or other procedures, decreasing time in surgery and improving patient outcomes.

Scoping Study Awards

Development of radiation detectors for medical imaging using opaque scintillators

PI: Prof Jeffrey Hartnell, University of Sussex

Co-I: Patrick Begley, Royal Sussex County Hospital

Positron Emission Tomography (PET) is crucial to diagnosis and staging of many cancers, but availability is currently limited. In 2019 the first total-body PET scanner was built in California, offering 40x greater sensitivity for scans of the whole body although with a price tag of about £10 million. A large fraction of the cost of these scanners is the transparent scintillator crystals that detect the radiation.
We have recently developed a new concept for a scintillator detector. Many radiation detectors use scintillators, which are materials that give off light when a neutrino hits them. Traditional scintillator detectors have required transparent scintillators to allow detection of the light, while our new concept requires an opaque scintillator. The opacity causes the light to bounce around close to where it is produced and then a dense lattice of fibre optic cables is used to extract the light. Such a configuration enables fast and high-resolution imaging capabilities. This novel detector technology is also particularly well suited to covering large areas such as those needed in a total-body PET scanner.
In this project we will quantify the spatial and timing resolution of a simulated prototype opaque scintillator detector element of a PET scanner. Simulation and reconstruction software will be developed to achieve this, based on algorithms developed for neutrino detectors such as MINOS and NOvA. In particular, we will investigate how the performance depends on variables such as fibre pitch, scattering length and scintillator/fibre decay times. This will inform future designs of prototypes that we aim to construct.

Proof of Concept Awards

Modelling of pulse pile-up and deadtime effects on quantitative imaging potential of x-CSI systems at medically relevant fluxes

PI: Dr Dimitra G. Darambara, The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust (ICR/RMH)

Co-I: Prof Val O’Shea, University of Glasgow

Traditional x-ray systems produce black and white images which are good for differentiating dense tissues like bone from softer tissues, but not very good at distinguishing between these softer tissues (e.g. fat vs muscle). Tumours tend to be slightly denser than healthy tissue, however so are a range of other, less dangerous growths such as cysts and polyps. Currently, to classify suspicious x-ray masses as cancerous or not patients have to undergo follow up testing such as a biopsy, which can be more distressing for the patient. X-ray photon Counting Spectral Imaging (x-CSI) is a new technique, which can produce colour x-ray images that provide a great deal more information to the radiologist at a fraction of the radiation dose to the patient.
Early work has already shown that x-CSI systems can readily distinguish between muscle and fat, as well as water and bone, tempting many to speculate that cysts, polyps and tumours could be differentiated without further analysis. This would greatly reduce the number of follow up procedures patients would need, allowing them to receive reassurance or treatment quicker than currently possible. A range of questions remain about how best to use the data x-CSI produces however, and in order to answer these questions soon accurate computer models that can simulate patient scans are required. Unfortunately, current computer models are unable to account for the electronics complications that are introduced by the high speeds used in medical imaging applications. The aim of the current work is to develop a complete computer model of the electronics inside an x-CSI system which can be incorporated into existing computer models of the rest of the system. It is hoped that doing so will allow existing questions about x-CSI system design and data use to be rapidly answered, giving patients access to this cutting-edge technique in their clinic.

Machine Learning System for Decision Support and Computational Automation of Early Cancer Detection and Categorisation in Colonoscopy

PI: Prof Bogdan Matuszewski, University Of Central Lancashire

Co-I: Prof Victor Debattista, University of Central Lancashire

Co-I: Mr Adnan A. Sheikh, East Lancashire Hospitals NHS Trust (ELHT)

Colorectal cancer (CRC) is one of the leading causes of cancer deaths worldwide, e.g. in the United States, it is the third largest cause of cancer deaths. In Europe, it is the second largest cause of cancer deaths, with 243,000 deaths reported in 2018. Colon cancer survival rate depends strongly on an early detection; decreasing from 95%, when detected early, to only 35% when detected in the later stages; hence the importance of colon screening. It is commonly accepted that most colorectal cancers evolve from polyps. Typically, a colonoscopy screening is used to detect polyps before any malignant transformation or at an early cancer stage. Optical colonoscopy is the gold standard for colon screening; however, colonoscopy has some significant limitations. Various recent studies have reported that between 17%-28% of colon polyps are missed during routine colonoscopy screening procedures, with about 39% of patients having at least one polyp missed. It has also been estimated that improvement of polyp detection rate by 1% reduces the risk of CRC by 3%.
With application of new advanced machine learning methodologies (the so-called deep learning) and advanced visualisation, it is conceivable to increase significantly robustness and effectiveness of colorectal cancer screening, improving lesion detectability and the accuracy of their histological characterisation, as well as reducing cost, risk and discomfort to patients. The proposed research will investigate possible ways to leverage this novel technology to develop better early CRC detection within colonoscopy procedures.The proposed methodology is very computationally intensive hence the need for collaboration between engineers, clinicians and experts in high-performance computing who are typically focused on high-end computationally demanding physical simulations. This project brings all these expertise together to assist clinicians in improving the outcome of the colonoscopy screening procedure.