Research feature: Scanning: the horizon

Fiona Gilbert in the Radiology Department

After seeing ever more precise imaging and computer-aided technology improve the detection of disease, Consultant Radiologist and Professorial Fellow Fiona Gilbert is excited about the potential for Artificial Intelligence to transform medicine, but aware too of some of the risks.

What can AI do well in medicine?

There is a lot of concern about AI, but in relation to medicine it has huge potential. I think it’s going to reduce our workload dramatically. Imaging is terrific because it’s digital data and so you can train algorithms to read it.

I’ve been massively fortunate to be in the speciality of radiology, because over the last 30 years or so the technological improvements and advances have been phenomenal. Every year there’s something new to look at or learn about, or we apply a new technique and follow up a patient’s outcome. That’s been a thread all through my career, taking imaging technology and saying: does it make a difference?

Before I came to Cambridge, I did a big study on computer-aided detection, where an algorithm will mark a suspicious area that might be a cancer on a mammogram (breast X-ray), to see whether this improves the human readers’ performance. We found it could be as good as the standard two readers, although we had a slightly higher recall rate. The team think this next generation of algorithms are probably a bit better, so they won’t end up recalling too many women, worrying patients unnecessarily and causing extra work.

One of my PhD students did a retrospective study of mammograms, comparing the performance  of commercial AI tools. All were good at early detection, which is key to patient outcomes. We held a meeting at Newnham with AI companies and experts and agreed a statement about what we should do to test AI tools, The Newnham Report, which is an important intervention in this area.*

So what are the some of the challenges and risks?

The greater use of AI calls for different skill sets, with more maths required in medical education, because more and more we will be reliant on and should be analysing big data. The existing data we have, if used effectively, can help shape health policy and how we deliver care, by analysing demographic data – incorporating shopping and eating habits, for example. We have a joint grant now with departments in maths, computer science, imaging and healthcare. Bringing together these multiple disciplines is where we’ve made the biggest advances in medicine.

The challenge is that the algorithms are not perfect; they miss things and so the radiologist can lose confidence in them, and they can introduce bias into reading. For example, if an algorithm has been created in one particular ethnic population it may not work so well in a different country where the population is completely different or much more diverse.

The implications of all this go much further; with medical records being digitised, we need to be very careful about how that data is used, with proper audit mechanisms in place. The algorithms are only as good as the quality of the data on which they were created.

What other areas are you looking at?

With a grant from Cancer Research UK I’m looking at improving the breast screening programme for people with very dense breast tissue, where X-ray is not particularly effective. Could supplemental imaging techniques (whole breast ultrasound, a contrast mammogram or MRI examination) detect breast cancer earlier? We’re also doing a collaborative European study, MyPeBS, where we’re stratifying women according to risk and if they’re high risk we image them more frequently. There is a lot going

What first drew you to radiology?

I was an undergraduate in medicine in Glasgow and following my house jobs I did a year in oncology and was fascinated by imaging. I took a job in Aberdeen, which was one of the earliest places to do magnetic resonance imaging (MRI), trained in radiology and became a consultant after five years, just when the UK breast cancer screening programme was starting.

I really loved these exquisite pictures that were being produced. I’m a very visual person. I understand a disease better because I can see the extent of it in the imaging. During my career, the technology has got more and more sophisticated. MRI and CT scans are now phenomenal, we can see more and more detail.

How did you move into medical research?

I hadn’t done formal research training but fortunately various academics in Aberdeen took me under their wing and I learned the basics. Then I learned on the job working on the Scottish Back Trial, thanks again to great supervision and mentoring. The trial was a multi-centre study looking at lower back pain and the influence of
CT or MRI scans on patient outcomes. We found that after two years, the outcomes were the same whether the patient had had a scan or not. It shocked a radiology conference in Chicago, but in reality, back pain often gets better by itself. In some areas scanning is essential, but when resources are limited, you need to know
what makes the biggest impact on patient care.

Through that study I worked with statisticians and health service researchers and learned research methodology: how to collect and organise data, write up results and do presentations. It was like doing a PhD.

After I’d worked on a few other major studies, I was appointed to the Chair of Radiology in Aberdeen – which was surprising to me and to my father, who was also a radiologist. They decided I had potential, so they invested in me. I sought expertise from top researchers in the medical school, and they helped me build an academic department.

‘It’s so important to invest time and energy connecting with people, and now as a senior woman I really want to make sure the younger ones coming through get the opportunities.’ Fiona Gilbert

You’ve touched on mentoring and support – how important has that been for you?

When I started in research, I recognised that I knew nothing. As an undergraduate I had written a dissertation, with some literature searching and learning how to write a scientific article, but that was about the extent of my experience. I was hugely lucky that I found mentors: I recognised my shortcomings and my lack of skills, so I sought out people who would be willing to spend some time teaching me or working with me on collaborative projects.

I was lucky, too, that I met lovely people in different organisations doing research and that often helps you progress, you know, if you make these connections with different people. I think it’s so important to invest time and energy connecting with people, and of course now as a senior woman I really want to make sure the younger ones coming through get the opportunities.

One of the main things that really helped was that my husband was supportive. Family is terribly important to me and I’ve been very lucky: I have a son and two daughters, and when the children were little my husband’s parents helped to look after them. I would sometimes have to take them into the hospital when I was on call, fortunately not often. Now that my career has taken me south, my family lives within ten minutes’ walk in London, which is fantastic.

What brought you to Cambridge?

For many years I was building up the department as I brought my children up in Aberdeen. My husband was working several days a week in London and, as the children went away to university, eventually it was me and the dog in Aberdeen and everyone else down south. So when I was invited to apply for the Cambridge Chair, my husband Martin said, ‘Why would you not?’ I was very fortunate to be appointed 12 years ago.

Then I joined Newnham and that means having a much greater connection to the University. My department is embedded in the hospital, so at Newnham I am able to enjoy the full opportunities that the University is able to offer, meeting other academics. You have this amazing expertise everywhere you turn. Newnham is so welcoming, irrespective of your background or the area in which you’re working. It’s a very positive, nurturing environment. I’m not going anywhere – although I am stepping down as Chair of Radiology, I will be working part-time as a Director of Research and Fellow Emerita of Newnham.

* Published as T. J. A. van Nijnatten et al., ‘Overview of trials on artificial intelligence algorithms in breast cancer screening – A roadmap for international evaluation and implementation’, European Journal of Radiology 167: 111087 (October 2023).

This article was first published in the Newnham College Roll Letter 2022-23