Skip to main content
Medical physics

Medical physics

Artificial intelligence smartens up cancer care

15 Jul 2019 Sponsored by Varian
Treatment-planning software gets smarter
Intelligent solutions: Varian is incorporating machine learning into treatment-planning software to enable clinicians to make faster and better informed decisions. (Courtesy: Varian)

More people around the world are being diagnosed with cancer every year. Thankfully, researchers and clinicians continue to devise ever more effective and sophisticated techniques to combat the disease, resulting in better patient outcomes and improved survival rates for most forms of cancer. But both of these important trends have created a data challenge in the clinic: how to collect, process and analyse increasing amounts of data to extract the useful information that will deliver the most effective treatment plans and the best outcomes for patients.

Artificial intelligence is emerging as a vital tool to tackle this data deluge. Using computer algorithms to search for important signals in the noise can reduce treatment times, improve the quality of care, and make the best use of valuable resources. “We have demonstrated that AI can make work more efficient without compromising on quality of care, and in many cases it can improve the care that patients receive,” says Corey Zankowski, senior vice-president of Oncology Software Solutions at Varian, which is focusing on developing and delivering intelligent cancer care solutions. “This application of AI to healthcare will continue to expand to new workflows, reaching across all disciplines of oncology, including multidisciplinary care delivery and patient symptoms management.”

Zankowski is confident that AI can make a real difference in all types of clinical scenarios. AI can allow cancer-care teams to make faster and better informed decisions, speeding up the treatment process and making the experience less stressful for patients. It can increase the scope for more personalized treatment planning, and ensure that all patients benefit from best practices as well as the collective experience of oncology care teams. AI could also transform the ability of clinical teams to use and learn from their patients’ data, creating holistic views to enhance both the treatment and the outcomes.

Zankowski says that Varian has been working with customers in all parts of the world to understand their specific challenges, and has partnered with some of them to create the huge datasets needed to build effective AI software. “With AI algorithms tailored to the specific needs of clinicians and patients, we hope to spread best practices globally and improve treatments for all cancer patients.”

Artificial intelligence reaches the clinic

Zankowski explains that existing oncology software has already brought machine intelligence into the clinic. As an example, Varian introduced RapidPlan knowledge-based treatment planning in 2014, which speeds up treatment planning and aims to deliver more consistent results by using machine learning to create preconfigured treatment plans based on data obtained from previous clinical experience. This software uses dose and anatomy information from existing plans to predict a more optimal dose distribution for new patients, based on their contoured anatomy.

RapidPlan can predict dose distributions

“RapidPlan creates consistent, high-quality plans for personalized radiation therapy,” says Zankowski. “It moves beyond templates, leveraging clinical expertise to quickly build the right plan for virtually all types of radiation therapy.” To improve predictions for patients, Varian’s research and engineering teams are exploring distributed learning concepts in which algorithms are trained across several hospitals, but without the patient data leaving the hospital.

The capabilities of RapidPlan can be further enhanced by combining it with multi-criteria optimization (MCO), a tool available within Varian’s Eclipse treatment-planning software that allows clinicians to explore what happens when they alter different clinical criteria. This enables oncologists to better optimize each treatment plan based on the uniqueness of each patient and their condition, delivering high-quality results by combining human intelligence with the machine learning provided by RapidPlan.

“While RapidPlan helps us know what should be achievable for a patient based on previously planned patients, MCO can help us tailor that plan based on the individual patient’s unique clinical circumstances,” explains Suzanne Currie, a medical physicist and the lead clinical scientific lead at the Beatson West of Scotland Cancer Centre in Glasgow, UK. “Using these two tools we are able to sculpt the dose and optimize our plans in ways we were not able to before.”

Varian is bringing intelligence to the clinic in other ways too. Its 360 Oncology system allows cancer-treatment teams to collect and access all clinical data recorded for each of their patients, enabling better co-ordination between different specialities and enabling improvements in the quality of care. The system also supports clinical decisions by comparing treatment options and using predictive intelligence to make recommendations based on calculations of dose, toxicity and quality of life.

We can help hospitals to navigate this transformative technology, empowering them with the insights they need to improve clinical operations and workflow, and to optimize patient care and outcomes

Corey Zankowski

Varian’s role, says Zankowski, is to give clinicians the tools they need to unlock the power of their data at every step of the process – from pre-diagnosis through to treatment and post-treatment care. Its Noona software application, for example, is a smart, cloud-based software that allows patients to report any symptoms they experience outside of the clinic in a structured way. Importantly for medical teams, it incorporates intelligent algorithms to help identify those patients who need urgent care, while also keeping track of any patient concerns for discussion at future visits to the clinic.

Towards a smarter future

But Zankowski says that this is just the beginning of what will be possible with AI in healthcare. Already Varian is exploring the possibilities of adaptive radiotherapy, in which the treatment plan could be constantly updated in response to clinical changes in the patient. The first step towards this goal is to automate tasks that humans do slowly, such as image contouring in treatment planning. “AI could be used to perform mundane and time-consuming tasks in treatment planning, such as contouring the healthy organs and tissues for image-guided radiotherapy,” he explains. “That would allow the oncologist to concentrate on contouring the tumour.”

Intelligent software should also enhance the ability of clinicians to assess whether the treatment plan should be changed, allowing any adjustments to be made more quickly and so improve the efficacy of the treatment. Further in the future, AI could also be used to mine large clinical datasets to detect early signs of patients who are responding to treatment, and those who are not. “Patterns in the data might emerge from analyses done by intelligent machines that can look at data from hundreds of thousands or millions of cases,” he explains. “Such patterns couldn’t be discerned by humans because no human could simultaneously analyse data from enough cases.”

It might be too early to make detailed predictions of how artificial intelligence will change the way cancer is treated in the future, but Zankowski is convinced that it will have a positive impact for clinicians and for patients. “Varian products, with AI embedded within them, are already changing the practice of oncology,” he says. “We can help hospitals to navigate this transformative technology, empowering them with the insights they need to improve clinical operations and workflow, and to optimize patient care and outcomes.”

Copyright © 2024 by IOP Publishing Ltd and individual contributors