Which innovations will have the greatest impact in radiotherapy by 2030? That was the question posed in the closing session of last week’s ESTRO 2022 congress; and five experts stepped up to respond.
As often seen in debate-style ESTRO sessions, competition was intense and gimmicks were plentiful, with all talk titles based on movies and a definite sci-fi twist. Before battle commenced, the audience voted for their preferred innovation based on the presentation titles. This opening vote put personalized inter-fraction adaptation as the winner. But could the speakers change the audience’s mind?
I, Robot
First up was Yatman Tsang from Mount Vernon Cancer Centre in the UK, who was tasked with arguing that by 2030, automation will have replaced humans in most aspects of the radiotherapy pathway.
“When I was first given the topic, I put ‘I, Robot’ into Google. I decided that instead of doing my slides, I’d rewatch the movie, to look for insight for my talk,” he admitted. “Then I thought, actually my job is quite easy, I’ll just present the facts and they will vote for me.”
So Tsang began by highlighting the prevalence of automation in our everyday routines. A wide range of technologies are available to reduce human intervention and increase efficiency in various tasks. Examples include sensors that control lights or escalators to save energy, or contactless payment cards. “All of this automation is already embedded in daily life and we are very familiar with it,” he said.
Moving onto automation in radiotherapy, Tsang explained that the radiotherapy pathway – a series of processes performed to deliver treatment – is extensive, complicated and involves a lot of people in different roles. He suggested that automation and robotics could take on some of the time-consuming tasks in this pathway, freeing up people’s time to invest elsewhere.
Machine learning makes its mark on medical imaging and therapy
Tsang noted that automation in radiotherapy is a popular topic, with the number of published studies on this theme increasing from 30 in 2011 to 381 in 2021, and many talks in this area at the ESTRO congress. He conceded that some colleagues are less keen on the idea of automation, thinking that machines may not perform tasks as well as humans. “But we are the ones that decide what, when and how we want to use automation,” he emphasized. “We should let the machines do the time-consuming tasks that we design for them.”
“Automation is everywhere,” he concluded. “And as the other panel members give their talks, I want to point out that automation will be the greatest innovation to help all of these four achieve the results that they want to achieve.”
Inter(fraction)stellar
The second speaker, Stine Korreman from Aarhus University in Denmark, proposed that in the future, every target, every plan and every fraction will be adapted to individual risk and response models.
Korreman began by sharing an image taken with a new imaging modality, containing a lot of new, hard-to-interpret information. It was, in fact, the first image sent back to Earth from the James Webb telescope. She explained that the same type of situation can arise in the medical field when encountering new modalities or other new information that’s difficult to interpret and unclear how to use.
“We can take two paths,” she said. “The Interstellar path, where we look at these as opportunities: so much information, so much to explore. Or the Don’t Look Up path: ‘this is a lot of information, we don’t understand it, let’s stick to what we know’. Of course, I’d like to propose the Interstellar path.”
Patients want to know how the tests performed on them are used to personalize their treatment, whether the scans taken every day are used to improve their plan, and whether the measured tumour response is accounted for. “At the moment, we are not really doing this,” said Korreman. “Our objective should be fully personalized radiotherapy in which we do risk profiling for every patient’s initial prescription, risk-based target definition, including microscopic spread, and dose painting to target every part of the patient with exactly the correct amount of dose.”
AI framework uses medical images to individualize radiotherapy dose
Citing the ESTRO 2022 programme, she noted that many researchers are already developing adaptation and personalization for every part of the radiation treatment chain. Ultimately, combining all of this research will enable the full chain of personalized radiotherapy, with every target, plan and fraction adapted to each patient.
“We have the choice of following the Interstellar path, exploring and putting it to the test, or to not look up and just stick to what we know,” Korreman concluded. “I say let’s not have all the information disappear into a black hole. Instead, even though it may be hard to use and difficult to interpret, use this information to personalize radiation therapy at every level for every patient.”
Rogue One (to five)
Next, Alison Tree from the UK’s Royal Marsden/Institute of Cancer Research explained why all radiotherapy should be delivered in a maximum of five fractions.
Aiming to sway the audience vote, Tree let the infamous Star Wars opening crawl argue the case for her, introducing the idea of a war against unnecessarily long radiation treatment. Instead, the crawl explained, we should use the force in just three to five days, to provide a new hope: a world where cancer can be cured in less than a week.
Tree explained that the idea of fractionation originated almost 100 years ago, when Claudius Regaud studied whether irradiation could cause sterility in rams. He observed that after delivering one large radiation dose, sperm were still produced, but that lots of small doses of radiation effectively stopped spermatogenesis. “That would be fine if our objective was to prevent rams having babies, but actually we’re trying to cure cancer,” she pointed out.
What’s more, today’s radiotherapy technologies can deliver dose so precisely that the α/β ratios defining the dose sensitivity of the tumour and healthy tissues don’t really matter. So why do patients still receive radiotherapy over more than five days, travelling to hospitals every weekday for weeks on end?
Tree cited the large body of evidence showing that hypofractionation is effective and feasible in most common tumour types. Breast radiotherapy can be performed safely in five fractions, for example, and prostate treatments using fewer fractions are just as effective.
Fewer, higher radiation doses safe for early breast cancer
And how low can we go? Teams at the Royal Marsden and elsewhere are studying two-fraction prostate stereotactic body radiotherapy, with MR-guided adaptation for all fractions. “It’s early days, but so far so good,” said Tree. Studies are also starting on single-dose ablative radiotherapy in oligometastases and primary lung cancer. “That would be a real step change, to be able to see, diagnose and treat the patient all in one day,” she noted.
Tree concluded by urging the audience to think of the polar bears. “We modelled that if you dropped from 20 to five fractions, just in the UK and just for prostate cancer, you would save 3.5 million kilogrammes of CO2 over one year. Hypofractionation will make a difference to patients by 2030 and save the planet.”
FLASH (Gordon)
Speaker number four, Pierre Montay-Gruel, from the University of Antwerp and the Iridium Network in Belgium, presented a talk entitled FLASH: He’ll save every one of us. “Since I was given this title, I’ve had this song in my head,” he said.
For ESTRO, however, FLASH is defined as radiotherapy delivered extremely fast at ultrahigh dose rate using electrons, photons or particles. “What is amazing about FLASH is that it does not induce classical radiation-induced toxicity patterns on normal tissues, but has a high tumour efficacy,” Montay-Gruel explained. “That may make it possible to increase the therapeutic window in radiation therapy, and that’s what everybody in the room here has been trying to do for years.”
It was 2014 when Vincent Favaudon and colleagues first demonstrated that increasing the dose rate can protect normal tissue without impairing anti-tumour efficacy. “Now FLASH is everywhere and everyone is talking about it,” said Montay-Gruel, pointing out the large number of FLASH talks at ESTRO this year, with attendees spilling out of the auditoriums. “FLASH makes people talk, discuss, brainstorm, be curious, and that’s what we need in the field.”
Importantly, FLASH is already in clinical trials, with the first patient treated for skin lymphoma in 2019, at Lausanne University Hospital. Current trials are examining proton FLASH for bone metastases, electron-beam FLASH for skin metastases, and other trials are planned, such as breast cancer treatment using intraoperative radiotherapy.
FLASH radiotherapy: from preclinical promise to the first human treatment
But many questions remain. Radiobiologists, for example, need to investigate normal tissue toxicity and tumour kill mechanisms, and assess which models to use. Physicists must focus on dosimetry, designing new irradiation systems and optimizing treatment planning. Clinicians, meanwhile, must decide which treatment modality to use and which tumour types to treat. “We’ve got a lot of work to do, but we have a very powerful research tool,” said Montay-Gruel. “FLASH radiation therapy is a new tool for clinicians, but also for researchers, and that’s amazing”
Circling back to FLASH “saving every one of us”, Montay-Gruel defined “us” as radiation oncology professionals, whom FLASH will save by renewing the research field and triggering creativity, and also the patients, who may perhaps be saved from cancer. “Let’s be honest, FLASH will not replace 100 years of radiotherapy techniques,” he said. “But I hope it will help us renew those techniques and one day help us treat cancer.”
The Matrix
The session’s final speaker was Jean-Emmanuel Bibault, from Université de Paris in France, who argued that in the future, treatment decisions and optimal steering of radiotherapy will be based on big data and cloud computing. “I’m going to be talking about the matrix – basically that’s just cloud computing for optimal treatment decision,” he explained.
Today, there are numerous new therapies and new biomarkers available and its essential to choose the right ones to treat each patient optimally. We are also entering the era of big data, said Bibault, and here we are only scraping the surface. “We have lots of treatments and are going to have a have huge amount of data and no idea what we should be doing with it,” he said.
With the human brain only able to make optimal decisions when considering up to five variables, our brains are already saturated, Bibault pointed out. Thankfully, we have cloud computing. “We are currently building systems that are able to take much better decisions than us. Our responsibility is we have to use them for patients,” he explained. “We can’t simply reject AI or cloud computing because we are afraid to lose our jobs.”
The transition to cloud computing will no doubt face obstacles along the way. Bibault noted that some physicians are afraid that they will end up with no choices to make, just functioning as some kind of robots. But he emphasized that radiotherapy already works within sets of guidelines, which are essential for the patients, and predicted that the situation will not be so different from today.
By 2030, Bibault admitted, it’s likely that all radiotherapy will be automated, personalized and hypofractionated, and that FLASH will save us. “But the innovation piloting all that is going to be treatment decision and cloud computing,” he said. “Just like Neo learnt Kung Fu with a single button, I want to learn how to treat my patients with a single click of a button.”
Following the five presentations, the ESTRO attendees were invited to reconsider their earlier selections. The results revealed that Alison Tree may have used the force to change the audience’s minds, with hypofractionation clearly winning the final vote. Tree was rewarded with a prize of a Lego Millennium Falcon Microfighter. I look forward to seeing what ESTRO comes up with next year in Vienna.
