In the UK, the NHS radiology workforce is chronically understaffed, with a shortfall of around 33 per cent – and the issue is getting worse.
In fact, by 2025, the Royal College of Radiologists estimates that the shortfall will reach 44 per cent – with 3,600 new radiologists needed.
Put simply, while demand for diagnostics has risen, the number of radiologists hasn’t – and last year’s pause on non-urgent clinical has exacerbated the problem.
But in addressing the gap, AI has emerged as a powerful tool, as we discovered in conversation with Stine Molgaard Sorensen, co-founder and chief operating officer of Radiobotics.
The firm has carved a niche in the field by focusing on musculoskeletal scans, which, unlike breast screening for example, have been largely neglected in terms radiology AI.
With FDA clearance already granted for its RBknee product, which can identify the presence of osteophytes, subchondral sclerosis, joint space narrowing and diagnose osteoarthritis, Stine hopes the firm’s range will start to be adopted in healthcare settings within the next two years.
She said: “In two years, more than 4,000 radiologists will be needed in the UK. And the problem is, it takes around ten years to train to become a radiologist.
“I remember the very first conversation we had with a Danish professor in musculoskeletal radiology, who said that we can’t educate ourselves out of this problem, we will have to introduce new technology to solve problems instead.
“So that’s why I hope, in the next two years, we will have a penetration to such a level that we have gained a foothold in the UK and the US, which will then function as inspiration to others.”
The inspiration for Radiobotics came from co-founder Mads Jarner Brevadt, who previously worked in the orthopaedic department at Imperial College, London.
Stine said: “When Mads was working, he could see how the department was working with technology – and how they were not working with technology.
“And while he was doing the preparation for the orthopaedic surgeons, he began to think that it would be much better if machines could do this. Could you actually automate some of the images? – you had the pixels, and you had the data, which are both well suited towards machine learning.
“I was working in the tech space in Copenhagen at the time, and he reached out to see if I thought the idea could work. We spoke to some clinicians and, four months later, we incorporated Radiobotics.”
Stine says that each of the four co-founders – Pavel Lisouski and Martin Axelsen being the other two –bring a different skillset to the firm, enhancing its credibility.
She said: “We complement each other in terms of the competencies that you need to build a business, because, for health tech especially, you need to have someone rooted in the clinical side.
“Mads was a biomedical engineer, Martin, who is our chief scientific officer, has a PhD in medical imaging and machine learning, so he is running the scientific side, doing all the clinical validation.
“I take a more strategic approach, with a background in tech startup, and then we have Pavel, who is our chief technical officer.”
Using artificial intelligence, machine learning and computer vision, Radiobotics’ devices can assess images, detect major findings, analyse them and produce a text report and conclusion along with a visual overlay to support the radiologist’s decision.
The firm is currently working with the East Midlands Radiology Consortium to deploy its technology in an NHS setting.
And Stine notes similarities, in terms of the challenges of adoption, between the NHS and the Danish healthcare system.
“In some ways it can be quite similar in terms of, how will this benefit the patient? What is the value that this brings to the entire system? And that is something we’re working on proving, because we will need to get into the systems before you can claim these things.
“However, what I see in the NHS is a willingness to invest, which I would go so far as to say is larger than in Scandinavia.
“Look at the NHSX initiative, where they are investing a lot in AI technology and trying to create a framework for how we’re going to prove value within a smaller market to begin with, and how can that than be applied to the NHS in general.”
While Covid has contributed to the radiology backlog, however, Stine believes the pandemic has also paved the way for more innovation and trust in AI.
She added: “Four years ago, people would look at what we were doing and say, ‘Yeah, it looks very nice – come back when you actually have something.’
“Now we’re finding more people are open to AI and we’re also getting better at finding those people who are a) open to it and b) have experience of it.
“We need that investment from the clinical side too. That’s something I hope starts to happen, that hospitals have their eyes opened and don’t expect us to have figured it all out – because it’s innovation.
“It’s new technology that we have to validate together, not an enterprise IT system that has been available for the last 20 years.”
So, what’s coming up for Radiobotics? The firm has knees and hips covered – is the rest of the body next?
Stine says the reason for starting with those two body parts was volume – they’re the most diagnosed with musculoskeletal issues – but it’s not stopping there.
As she explains: “You can have a chronic disease with your joints, or you can break your bones. So that’s the two areas where we fit in, and in that space, are radiologists, orthopaedic surgeons and emergency care doctors.
“Through these three targets, we are looking at covering 70 per cent of the whole skeleton – that’s the goal.”