Artificial intelligence (AI) has revolutionized oncology care as a whole1, and it is starting to show great potential to alter the roles of radiotherapy oncologists. According to Matthew A. Manning, MD, the future of AI in radiation oncology could be in diagnostics and clinical decision support.
“AI can ingest large amounts of data and process more data than a human being could in a reasonable timeframe. We expect that clinical decision support tools radiation oncology will come around assessing individual patient records and refining treatment recommendations based on that artificial intelligence can also look for patterns which may not be discernible to people looking to marginalize information and help guide us to discern patterns of which patients may be at higher risk for complications or higher risk for failure and unless the treatments improve,” said Manning, chief of Oncology at Cone Health.
The main use of AI in radiation oncology is in aiding the application of radiation. Manning explained that adaptive radiation therapy and motion management can be more precise with AI. Moreover, AI has been brought into important clinical tasks, like data sharing and response evaluation.
“In radiation oncology, also, there are different medical record platforms necessary for the practice that are often separate from the hospital medical record. Creating these interfaces that allows reductions in the redundancy of work for both clinicians and administrative staff is important. Tools using AI and business intelligence are accelerating our efforts in radiation oncology,” Manning stated.
In the interview, Manning discussed the many radiation oncology tasks that have been improved with AI and its prospective utility in cancer diagnosis and treatment decisions.
Targeted Oncology: Can you talk about the different ways AI is leveraged in radiation oncology?
Manning: AI is the backbone of some of our clinical decision support tools. It helps with driving automation. We can reach oncology or across oncology in general [with AI]. The amount of medical science information is exponentially expanding and doubling every 2 months or so. While it wouldn’t be possible with clinicians to update knowledge on the most up to date science, using knowledge-based AI and non-knowledge-based AI offers clinical decision support.
In radiation oncology, also, there are different medical record platforms necessary for the practice that are often separate from the hospital medical record. Creating these interfaces that allows reductions in the redundancy of work for both clinicians and administrative staff is important. Tools using AI and business intelligence are accelerating our efforts in radiation oncology.
By accelerating radiation oncology treatment planning and administrative burdens, we are able to accelerate the initiation of cancer treatment for patients with cancers that are growing. Delays in time to treatment will improve outcomes and cure rates.
What about the role AI plays in data sharing? What tools are available or being developed?
I think the role for AI for each technology is just beginning. I think we’re going to see a lot fill up over the years. One area where AI is helpful currently is with software platforms related to the billing and coding or charges for treatment. That data needs to be reconciled against what’s been documented, and through natural language processing through AI evaluation of the existing documentation, we have software platforms that can verify what we’re billing for weeks documented correctly. It flags cases and will give you updates when necessary.
There are newer advances like image-guided and adaptive radiotherapy. What you know about how AI is be utilized for those purposes?
Image-guided radiation therapy implies that while the patient is on the treatment table, the position of the target is verified. The patient is repositioned by millimeters to ensure that the target is in the crosshairs of radiation. There are a lot of way to do image-guided radiation, including prior to radiation delivery, which just ensures that patients are properly set up. Then, there’s even intrafraction monitoring, where image guidance occurs during the radiation treatment itself. It takes just a few minutes and allows you to adapt the radiation treatment position.
There’s also an adaptive radiation therapy where the patient arrives and their images are acquired, and their planning is updated to account for changes in their anatomy. Adaptive radiation therapy in its beginning stages around the United States. In the future, we expect that we will be customizing radiation treatments not just per patient, but we’ll be customizing radiation treatments per patient each day based on their anatomy. We will be able to see anatomy changes based on what patients eat, drink, their body weight. Those types of things can be adjusted to optimize radiation therapy.
Can you talk about motion management through AI? Why are these tools important?
Motion management is very important. People who are alive move. Their lungs move as they breathe their heart beats, and things like lung cancers, and even cancers of the liver, move up and down as the respiratory cycle proceeds. In the past, we had to take that motion into account and just treat the entire potential series of locations of the tumor.
But nowadays with motion management, we can have tools to reduce motion by compressing the abdomen, and tools to track motion like intrafraction imaging, or having patients hold their breath, and treat holding the deep inspiration breath hold. With motion management, we’re able to reduce the amount of mole tissue it’s being exposed to high-dose radiation and potentially reduce complications from radiation necrosis response.
What is role is AI playing in response evaluation?
Some tumors change as they’re being treated with radiation. Ideally, they’re shrinking in size, sometimes they increase initially. With response evaluation, we’re able to assess the target and revise the radiation treatment using adaptive radiation therapy to ensure that the radiation is being delivered to tumor.
One instance where that’s relatively common is with central cancers that are obstructing healthy lung tissue. When we initially meet the patient, their cancer and their obstructed collapsed lung may be difficult to discern from one another. As a course of radiation precedes the tumor regresses, the airways open at the lung areas. Now, we may be able to refine radiation targeting to cover the tumor only and not treat healthy lung with high-dose radiation.
How do you think AI will be helpful in the future?
AI is increasing in its strengths and in terms of its training, and we’re starting to see use cases for AI. In current state, AI is making its way into computer-assisted detection on diagnostic films. Looking at a chest CT, trying to find more knowledge, we see that AI is very powerful at finding spots that maybe the human eye may miss. In terms of radiation oncology, the use ultimately will be in the setting of clinical decision support.
AI can ingest large amounts of data and process more data than a human being could in a reasonable timeframe. We expect that clinical decision support tools radiation oncology will come around, assessing individual patient records and refining treatment recommendations based on that artificial intelligence that can also look for patterns which may not be discernible to people looking to marginalize information and help guide us to discern patterns of which patients may be at higher risk for complications or higher risk for failure unless the treatments improve.
I think in present state, we’re starting to see artificial intelligence get into the diagnostics of oncology, in terms of the therapeutics, it won’t be more people will be relying on artificial intelligence to assist in radiation oncology.
Luchini C, Pea A, and Scarpa A. Artificial intelligence in oncology: current applications and future perspectives. Br J Cancer. 2022 Jan;126(1):4-9. doi: 10.1038/s41416-021-01633-1