Dr Jane Sosabowski was awarded funding to improve non-invasive imaging of mouse models of pancreatic cancer.
Principal Investigator: Dr Jane Sosabowski
Organisation: Queen Mary University of London
Award type: PhD studentship
Start date: 2015
Duration: 3 years
There are almost 10,000 new cases of pancreatic cancer in the UK each year. While survival rates for most cancers have increased over the last 40 years, the prognosis for patients with pancreatic cancer is poor, with few surviving more than a year after diagnosis. Research on pancreatic cancer and the development of treatments predominantly uses mice. This includes genetically altered animals and xenograft models either with cell lines or patient-derived tissue.
The KPC mouse model is commonly used. This has a conditional point mutation in both the p53 and KRAS genes. It develops pancreatic ductal adenocarcinoma with similarities to human tumours as well as associated comorbidities of cachexia, jaundice and ascites. For xenograft models, cells or patient samples are either transplanted heterotopically or directly into the pancreas. Orthotopic tumours have the advantage of providing the opportunity to study tumour development in the appropriate pancreatic microenvironment.
The mouse pancreas lies in the upper abdomen behind the stomach and is soft and diffuse in comparison with the human pancreas which makes it difficult to define and distinguish from the surrounding tissues. These factors mean that sizing orthotopic tumours and monitoring their growth over time, for example to assess treatment efficacy or for the purposes of humane endpoints, is challenging. Palpation is difficult and does not provide accurate quantitative information. Ultrasound can be used but animals have to be shaved, image acquisition is highly operator dependent, and analysis of volumetric images is time consuming. Longitudinal studies typically require mice to be culled and the pancreas removed. Consequently, studies require large numbers of mice for each time point and individual animals cannot be tracked.
Magnetic resonance imaging (MRI) is already used in mouse studies on pancreatic cancer. It is readily applicable, even for inexperienced operators, with the fast-growing localised tumours that occur with orthotopic transplants. However, MRI image analysis is challenging with genetically altered models, such as the KPC mouse, where tumour growth is typically slower (three to six months) and where there is a gradual transformation, sometimes across the whole pancreas. Although there are commercially available tools to aid with image interpretation, these largely only work for welldefined organs such as the liver and stomach and not the pancreas.
3Rs benefits (actual and potential)
Algernon Bloom, the PhD student, has used a low field, small animal MRI instrument to build a library of images of the healthy pancreas and pancreatic tumours in the KPC mouse. By calculating various features for each 3D pixel in the MRI images (e.g. intensity, gradient, spatial and probability features), Algernon has used machine learning to develop an autosegmentation tool that can automatically identify a tumour-positive pancreas with 95% accuracy compared to veterinary radiologists and image analysis experts. Where the model and the image analysis experts disagreed, in the vast majority of the cases, this was because the model had correctly identified a tumour earlier than the experts. This early identification can inform the use of humane endpoints and the monitoring of mice.
Algernon and Jane are working with the Boston-based company Invicro to incorporate the pancreatic segmentation tool into their existing 3D mouse atlas software. It is also being applied to other projects at the Barts Cancer Institute, Queen Mary University of London (QMUL), where there are more than 90 researchers working on pancreatic cancer, and through new collaborations with researchers in Glasgow, Cambridge and London. Research at the Institute has already shown that the use of MRI for orthotopic tumours reduces the number of mice used per study from 12 to eight animals. Wider uptake of the autosegmentation tool will ensure the advantages of MRI imaging, such as the ability to do longitudinal imaging on the same animals, can be applied more readily to genetically altered models of pancreatic cancer.
Scientific and technological benefits
The low field MRI instrument used by Algernon was funded through an NC3Rs Infrastructure for Impact grant in 2013 to Professor John Marshall at QMUL, with Jane as a co-investigator. The instrument has been applied to a wide range of animal studies at QMUL including cancer (e.g. pancreatic, brain and lung) and trauma injury models, demonstrating the utility of low field MRI. Impacts include developing new methods to assess lung tumours in mice to improve the humane endpoints used, and demonstrating that MRI can be applied to monitor tumours of the omentum (a fold of the peritoneum connecting the stomach and the abdominal viscera), which is adjacent to the pancreas.
John and Jane have tracked the reduction in animal use from the instrument with their ‘Mouse Lives Saved’ Counter, with the number currently standing at 2,600 mice.
Algernon has presented his machine learning tool at a number of international conferences, including the World Molecular Imaging Congress in Seattle in 2018 where he won a poster prize. He has participated in more than 50 hours of public engagement activities, including as a STEM Ambassador, to champion the 3Rs. In 2017, Jane gave a presentation on the 3Rs and imaging at an NC3Rs-hosted event in London as part of the ‘Pint of Science’ festival. Working with veterinary colleagues at QMUL, Jane has published papers on anaesthesia and monitoring animals during MRI and other welfare considerations.
This case study was published in our 2019 Research Review.