Medical Image Perception Research Group
Theme Leads
Dr Tim Donovan, Dr Peter Phillips, Sarah Davies
Our Team
Dr Tim Donovan, Dr Peter Phillips, Sarah Davies, and Dr Gareth Bolton
Brief description
What we do
Medical Image Perception research has been undertaken at the University since 1999 with the research group pioneering the application of eye tracking technologies in medical imaging to determine expertise related differences in searching, decision making and the type and frequency of errors. We have continued to develop novel methods of analysing and interpreting eye tracking data in response to the rapid increase in volumetric imaging. With the rapid implementation of AI in radiology we are now looking at the impact of this on decision making.
Our expertise
It is important to determine the empirical evidence for prevalent models of medical image perception, which is why the rigour of experimental psychology methods have been used to test assumptions about the way experts process complex visual information, and we use experimental eye tracking paradigms to gain an insight into expert performance.
Eye tracking studies carried out on static 2D images cannot be generalised to the 3D volumes of data from CT or MRI, which is why we have developed new ways of analysing, presenting and interpreting eye tracking data in complex visualisations where computer-aided detection can be used.
We are currently collaborating with researchers at Lancaster University, Edge Hill University and Nottingham University.
Our projects
We have an ongoing PhD research project evaluating human/AI interaction in medical image interpretation.
We know that expert observers are very efficient in the way they search through medical images, and it is thought that rapid scene recognition, or processing of the initial ‘gist’, underpins expert performance when searching for pathology in medical images. We are currently researching how experts exploit that initial glimpse and how it effects subsequent performance.
In collaboration with the University of Nottingham, we are exploring decision-making processes in breast cancer screening. Digital breast tomosynthesis creates layered imaging through breast tissue, reducing the visual overlap of tissue produced by projection mammography. The exploration of these layers extends the interpretation time of digital breast tomosynthesis compared to mammography. Using eye tracking, we can measure reading strategies and reader fatigue. We are also examining visual search patterns within these two breast imaging modalities and investigating whether the use of AI influences these search patterns.
Recent publications
- Donovan, T. et al. (2020) ‘Fetal eye movements in response to a visual stimulus’, Brain and Behavior, 10(8), p. e01676. Available at: https://doi.org/10.1002/brb3.1676.
- Litchfield, D. and Donovan, T. (2016) ‘Worth a quick look? Initial scene previews can guide eye movements as a function of domain-specific expertise but can also have unforeseen costs’, Journal of Experimental Psychology: Human Perception and Performance, 42(7), pp. 982–994. Available at: https://doi.org/10.1037/xhp0000202.
- Partridge, G.J.W., Taib, A.G., Phillips, P. et al. (2024) ‘Take a break: should breaks be enforced during digital breast tomosynthesis reading sessions?’, European Radiology, 34(2), pp. 1388–1398. Available at: https://doi.org/10.1007/s00330-023-10086-4.
Contact
Dr Tim Donovan – tim.donovan@cumbria.ac.uk
Dr Peter Phillips – peter.phillips@cumbria.ac.uk