Cambridge's Raj Jena becomes UK's first Professor of AI in Radiotherapy

Cambridge oncologist Raj Jena has been appointed the UK’s first Clinical Professor of AI in Radiotherapy.

The creation of the new Cambridge University Clinical Professorship signals the importance of AI in the fight against cancer and builds on the University’s work to explore and apply the very latest technology to the world’s major challenges.

The technology is already being used in the diagnosis and treatment of some cancers, but Professor Jena – an academic radiation oncologist at Cambridge University Hospitals NHS Foundation Trust, and a researcher in the University’s Department of Oncology – says AI has the potential to transform the way patients experience cancer care through more innovative and personalised therapies.

However, developing increasingly powerful tools requires ever larger amounts of information to train AI models on, and so building the research collaborations needed to help gather information on this scale is one of the areas where Prof Jena hopes his new role can help.

“There are amazing AI tools being developed, but they need more data to feed on than one hospital or one healthcare system can generate. The Professorship will make a concrete difference in terms of building networks and leading projects.

“We need to bring together the right teams and expertise to research these things, to understand the technology and of course develop guidance and good practice, because trust is key, as is making sure you bring others with you. It’s one of the things I’m most excited about with this new role. You can design an AI but if you don’t explain what you’re doing, and how it will help, you’re limiting the good it can do.”

Prof Jena’s research into AI in radiotherapy has already seen the OSAIRIS tool – which helps prepare scans, and cuts the time patients have to wait between referral and starting treatment – become the first cloud-based AI technology to be developed and deployed within the NHS. OSAIRIS automatically segments radiotherapy images, identifying and protecting healthy tissue by drawing around it, in just a few minutes, under the control of a clinician.

Now he is designing a new smart radiotherapy machine that goes one step further, and helps identify tumours before suggesting the best possible radiotherapy treatment. The research aims to support cancer services in developing countries, where radiotherapy machines and the staff to run them are in short supply. The new tool will feed scans into an AI foundation model – trained on an international dataset of medical images – and mark out the tumour target and the best radiotherapy plan to treat it.

“It takes over from OSAIRIS, but of course marking out tumours is much more difficult than marking out healthy tissues – and is usually where the human expertise comes in,” says Prof Jena. “However, it looks as if the programme can make a reasonable first attempt, and we’ve already had some initial success.”

Currently, the technology is not at a stage where a patient could be treated based on the diagnosis, but Prof Jena says because it can identify similar cancers from a huge database of cases from around the world, it can help offer insights by detailing how those other patients were treated.

“The beauty of using AI in this way is that we start out with the kind of simple tools that we've already been using, to process images and add content, but then we’re adding the AI of a large knowledge system. And finally we’re adding the power of a foundation model to sift through this massive amount of information. Individually we can do all those things already, the key is putting them all together.”

And Prof Jena says the technology will only become more sophisticated.

“We’ll start to see AI impacting in each stage of the patient’s treatment,” he said. “There’s a lot of AI that’s focusing on being able to pick up tiny cancer signals among all the ‘noise’. So we’ll have more early detection programmes, and instead of having to wait for a cancer to be big enough to see on a scan, we’ll be asking can we instead analyse breath, or a finger prick of blood?”

AI is also likely to play a bigger role in helping clinicians decide on treatment paths for patients after they have been diagnosed, integrating huge amounts of separate data, including genomics, and suggesting the most effective way forward.

Indeed, another AI research study at Cambridge will look at the patient in their entirety and suggest a personalised treatment plan accordingly. So if, for example, a patient has lost a significant amount of muscle mass, it may suggest a particular drug would be unhelpful, and if it identifies something like a central obesity pattern – consistent with diabetes or metabolic syndrome – it might suggest that this condition should be treated before a particular type of medicine is used on the cancer.

Prof Jena says the power of AI to mine vast amounts of genomic data to inform treatment has been a ‘game changer’, slashing the time it takes to map the full genomic readout of a tumour.

“It’s coming down and down, and for some cancers it's as fast as four days now. And that's amazing when you think about it. In the old days, somebody would put a needle into a tumour, and then a pathologist would put some dye on it and look at it under a microscope, and then they would decide the best way forward with a very limited amount of information from the tumour itself. Now, suddenly, we’re able to digest several gigabytes of information about a tumour’s genetic profile and reduce all of that into actionable data within days.”

Genomic analysis on this scale means some patients – for example brain tumour patients – have been able to access new treatments that would not otherwise have been thought of. The tools look for patterns in the tumour genome regardless of the type of cancer, and these can be correlated to new therapies available in the clinic.

“It’s wonderful to find a match,” says Prof Jena, “because there’s absolutely no way you would have been able to read the entire genomic profile of a tumour and pick up the signature that unlocks a potential treatment.”

In the future, Prof Jena believes patients will be able to access their own ‘disease avatar’ to help monitor their condition. So, a patient who notices a slight change within themselves might be offered a ‘first pass’ by an AI-supported system – which could offer some reassuring context based on the information they supply – instead of that information going straight to their doctor or specialist nurse.

“What might be a worrying change for one patient could be completely expected in another,” says Prof Jena. “The system might escalate unexpected changes, and put the patient straight through to their doctor, or it might instead suggest monitoring the situation, and get back in touch with the patient the next day to follow-up. The key aspect of an avatar based model is that it would individualise its action based on what it knows about the patient.”

And in terms of radiotherapy specifically, Prof Jena says AI will not only continue to simplify and accelerate the complicated and time-consuming process of preparing scans, but it will also play a bigger role in informing treatment plans.

He says: “We have known for a long time that tumours shrink and change in the weeks that pass during radiotherapy treatment. We have simply lacked the means to change the treatment each day in line with the changing shape of the tumour. AI can transform this process, so that each day we could take a new scan, build a new picture of the tumour, and recalculate the treatment plan while the patient is lying on the treatment couch. It means we’re adaptive and protecting everything outside the tumour. It will give us much more information to track how tumours are changing, almost in real time, and allow our radiotherapy to be as precise as possible.”

Published: 17th October 2024
Interview and words: Stephen Bevan
Images: Cambridge University Hospitals

_{The text in this work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License}