Radiotherapy innovator Elekta harnesses AI technologies to help further its mission to bring hope to more cancer patients in more places throughout the world. Elekta AI specialists use the agility and power of on-demand infrastructure and services from Microsoft Azure to develop solutions that help empower clinicians to provide the next generation of personalized cancer treatments with the potential to improve patient treatment. Resulting increases in efficiency and effectiveness will help drive down the disparity between the many who are in need and the relatively few who have access to first-rate radiotherapy treatment.
“We rely heavily on Azure cloud infrastructure. With Azure, we can create virtual machines on the fly with specific GPUs. If that’s not enough, we can cancel that virtual machine, create a new one, and then scale up as the project demands.”
Silvain Beriault, Lead Research Scientist, Elekta
Broadening opportunities for care
Radiation therapy is an effective and affordable treatment for cancer. Its benefits are by now well established, yet its availability remains a problem. “Access to the required resources and expertise, especially in emerging markets, continues to be an issue even today,” says Rui Lopes, Director of New Technology Assessments at Elekta, which develops precision radiotherapy solutions. “It’s an unfortunate statistic that radiotherapy treatment remains unavailable to almost half of patients who could benefit from it, especially in emerging and developing areas of the world.”
Elekta makes innovative equipment, software, and services for cancer treatment providers and researchers. The company’s stated purpose is to provide hope for every person dealing with cancer. Says Lopes, “Our commitment is to a world where everyone has access to the best care possible.” Elekta considers technology, particularly breakthrough AI capabilities, as essential to expanding the use and availability of radiotherapy treatments. “It’s not just a question of purchasing equipment. You also need competent resources who know how to run it and to offer the right level of associated care,” Lopes says. “So, if you can embed some of that intelligence into the machines, you have a greater chance of democratizing cancer treatment and making it accessible everywhere.” Teams at Elekta, including Elekta Lead Research Scientist Silvain Beriault, are aiming to use AI to help rebalance the statistics and provide many more people around the world with access to the treatment they need.
Building hope, saving time
A pervasive problem at the root of many roadblocks to accessing cancer care is time. For the patient, this means time they need to spend away from their work and families to undergo their course of therapy, which often involves daily treatments over several weeks. For hospitals and clinics, this means time spent by scarce experts who are required to assess patients, identify radiotherapy targets and at-risk organs (known as contouring), and plan a course of treatment. Each visit involves more tests, scans, and assessments to confirm progress and ensure that vital contouring remains accurate as treatment progresses.
According to Beriault, Elekta technology saves time for all concerned. “We accelerate the overall treatment planning process with AI by automating time-consuming tasks such as analyzing and identifying issues, contouring targets and normal tissue to be spared, and optimizing the dose,” he says. And during a treatment session, AI is used for plan adaptation and dose re-optimization to help ensure the plan is up to date with the patient’s current anatomy. The equipment also monitors patient movement in real time during delivery to ensure the treatment remains safe, tuned, and targeted without the need for repeated manual intervention or lengthy MRI acquisitions that disrupt or delay the overall process. “We can get real-time MRIs while the patient receives treatment, tracking movement and keeping the target tumor in focus while minimizing risk to other healthy tissues in the area,” Beriault continues. “People find it hard to remain absolutely still for extended periods. Our system allows treatment to continue even if the patient moves a little, and that also reduces the number of minutes they must be in treatment.” The accuracy of the treatment also means that doctors can potentially increase the daily treatment dose to the target and reduce the duration of treatments from several weeks to several days.
The technology is used in Elekta’s flagship Unity system to provide a highly accurate and up-to-date view of a patient’s progress. Adds Lopes, “AI-enabled software means that Unity has the potential to dynamically adapt to every nuance and change in a patient’s anatomy and condition, whether that means changes in shape and size of organs from day to day or changes in the tumor due to the treatment. Before these breakthroughs, the best we could do was assume that everything stayed exactly the same during several weeks of daily treatments.”
Efficient development with strong security and streamlined compliance
Elekta doesn’t believe that AI technology can replace skilled physicians, but rather, it can make physicians more efficient so that they can spend more time focusing on valuable activities such as personalizing and adapting each treatment and interacting with their patients. The company’s AI-enabled solutions seek to help clinical, technical, and other specialized staff engage at the right moments and have all the information they need to be most effective.
The amount of information available is constantly growing. The Unity system, for example, generates a massive amount of imaging and telemetry data that can feed AI models to automate and streamline workflows. These models can evolve and advance over time as the system is used and more data is accumulated. Gathering this data and training machine learning models requires powerful and scalable storage and compute resources. “We rely heavily on Azure cloud infrastructure,” Beriault says. With Azure infrastructure, Azure Machine Learning, and Azure Virtual Machines, he found the agility Elekta needs to go where its research leads. “With Azure, we can create virtual machines on the fly with specific GPUs. If that’s not enough, we can cancel that virtual machine, create a new one, and then scale up as the project demands.” Elekta uses the PyTorch open-source machine learning framework on Azure to take full advantage of that scalability. “When it comes to optimizing our neural networks, we can easily launch experiments in parallel, which would be impractical without Azure,” Beriault continues. “We got the AI project underway quickly without any in-house hardware investment.”
Along with performance and efficiency, Elekta prioritizes the security of its IT infrastructure and the privacy of its data. The company must demonstrate compliance with healthcare data–related industry and government mandates. To address these concerns, the company makes sure that all data is anonymized, rests in location-specific encrypted blobs using Azure Blob Storage, and is accessed via the Azure portal across authorized, authenticated connections. Elekta’s positive experience led it to migrate one of its major cloud offerings, ProKnow, to Azure. ProKnow provides a comprehensive online repository for aggregated data, simplifying analysis and sharing. Elekta developers use ProKnow for key internal and external sharing and collaboration activities, including safeguarded, segregated customer workspaces for exchanging results and data. Says Lopes, “We recently migrated our ProKnow data discovery product from a different cloud platform to Azure because of its robust compliance capabilities for the many different regional jurisdictional requirements related to security and data privacy.”
Elekta manages these and similar development efforts across the company by using Azure DevOps. “We already used Azure DevOps for other projects, and that made it a natural fit for my team to use for our AI effort,” says Beriault. During development and testing, team members use Azure Monitor to track progress with model optimization and detect issues that might require them to halt or restart training or reconfigure the environment.
Successful partnerships driving successful outcomes
Partnerships are important for physicians and their patients—healthcare is a collaborative journey. And in its ongoing journey to bring hope to cancer patients, Elekta, too, collaborates with academic and healthcare professionals alike. “Elekta has always been a leader in supporting partnerships with academic centers and developing medical solutions that can eventually become marketable products,” Lopes says. “We’ve been working in ML and AI for more than a decade, using our extended Azure infrastructure to easily connect, share, and explore research opportunities with academic partners.”
Microsoft has become an important collaborator, and Elekta regards it as far more than simply a cloud vendor. “Our relationship with Microsoft is a strategic one,” Lopes says. Radiotherapy, and medicine more broadly, he notes, is an ever-changing, evolving environment. “To continue driving innovation, we need an IT partner that can keep pace, providing the support, services, and infrastructure to help us continue advancing and making a difference in patients’ lives—that’s what we get with Microsoft.”
Find out more about Elekta on Twitter, Facebook, and LinkedIn.
“When it comes to optimizing our neural networks, we can easily launch experiments in parallel, which would be impractical without Azure. We got the AI project underway quickly without any in-house hardware investment.”
Silvain Beriault, Lead Research Scientist, Elekta
Follow Microsoft