Webinar replay - The Holy Trinity of radiotherapy planning automation with AI
This morning, on October 23rd we had more than 50 radiotherapy clinical experts across the Nordics, Benelux and elsewhere dial in to listen to our speaker Prof. Nikos Paragios (Founder and CEO at TheraPanacea, Distinguished Professior of Applied Mathematics at Université Paris-Saclay) about automation and standardization of three vendor-agnostic treatment planning workflow steps with the power of Artificial Intelligence:
Throughout the webinar, Nikos delved into the current state of cancer treatment and how AI-driven automation can enhance efficiency, precision, and accessibility in radiotherapy.
Read the summary here or replay the entire webinar.
1. Introduction to AI in Radiotherapy
Nikos began by introducing the core theme of the webinar: how AI and machine learning are poised to bring the power of automation to radiotherapy. He noted that AI could optimize radiotherapy processes, starting with contouring (the precise mapping of treatment and avoidance areas) and advancing to treatment planning. These AI-driven solutions could, he explained, address some of the biggest challenges in radiotherapy, such as reducing time-consuming tasks and compensating for human bias.
He emphasized that while radiotherapy is essential for treating cancer, the current workflows are inefficient and slow. For instance, by 2030, around 15 million people are expected to require radiotherapy, which could generate up to 500 million treatment sessions globally. With this rising demand, AI could dramatically improve the capacity to deliver treatments faster and more efficiently.
2. The Role of Machine Learning
Nikos continued by elaborating on the benefits of machine learning in radiotherapy. He stated that one of the most significant advantages is time-saving, as AI can automate tasks that traditionally require hours of manual work. For example, generating treatment plans and contouring organs can be done much faster with AI, allowing the clinical staff to focus on more critical decision-making aspects of patient care.
He emphasized the need for standardization in radiotherapy practices. Currently, each hospital or center often follows its own protocols and guidelines, leading to inconsistencies in patient treatment. Nikos highlighted how AI could bring uniformity, ensuring that best practices upon the latest established international guidelines are applied consistently across different medical institutions. This would be especially valuable in smaller centers where staff may not have the same specialized expertise as larger, well-resourced facilities.
Moreover, he discussed how machine learning could improve treatment outcomes by reducing side effects and increasing the number of patients treated with the same infrastructure. By automating repetitive tasks, clinical radiotherapy professionals can deliver more treatments while maintaining quality care.
3. Current Challenges in Radiotherapy
Nikos then shifted to the challenges faced by modern radiotherapy. He pointed out that despite technological advancements in healthcare, radiotherapy has still a lot of potential for automation and efficiency gains. The process is still largely manual and involves two major stages: planning and treatment delivery.
One of the primary challenges, he explained, is the time-consuming nature of radiotherapy. Tasks like contouring, where clinicians must map out organs at risk and the tumor, can take hours, delaying the start of treatment. He also noted that expert bias plays a role, with each clinician making decisions based on their individual experience, which can lead to variations in treatment quality.
Nikos stressed the importance of adapting treatment to changes in the patient’s anatomy, something that is gaining reasonable traction in radiotherapy practices. He explained that often though, anatomical changes during treatment are ignored, and as a result, doctors often add larger margins around the tumor to compensate for potential movement or growth, which can increase toxicity and cause damage to healthy tissue.
4. AI’s Impact on Treatment Planning
Nikos provided insights into how AI is already making strides in treatment planning. He detailed how TheraPanacea 's AI algorithms can automatically contour organs at risk, saving valuable time for clinicians. He shared that AI could also help with tumor delineation, or outlining the tumor's boundaries, which is particularly challenging for complex tumors such as those found in the brain or head and neck.
He highlighted several AI tools developed by TheraPanacea that have already been implemented in clinical settings. For instance, their solution for automatically delineating brain tumors (pending MDR certification) has shown a +90% accuracy rate (measured using DICE) against the gold standard, which was found being a significant improvement when compared to contouring performed manually by clinicians. He continued by explaining how AI can reduce human error and variability, ensuring more accurate treatment plans, which are essential for delivering effective and safe radiation doses.
5. Automation and Adaptive Planning
Nikos talked extensively about the importance of adaptive radiotherapy, which allows radiation oncologists to adjust treatments based on changes in the patient's anatomy over their treatment cycle (ranging from days to weeks). He noted that despite its potential, adaptive radiotherapy remains underutilized due to its complexity and high costs. However, he emphasized that AI could make adaptive therapy more accessible by allowing existing infrastructure to be used for automation.
He explained that using generative AI, TheraPanacea transforms low-resolution CT scans into high-resolution images providing better data for daily treatment adjustments. This process, he said, could democratize adaptive therapy, making it available to a broader range of patients without the need for expensive (several million Euros), specialized adaptive machines.
Nikos then introduced a future vision where AI-driven systems could handle the entire process, from creating the digital twin of a patient to delivering a fully automated treatment plan without the need for a clinician’s continuous involvement. He emphasized that the ultimate goal is to reduce the cost and time required for adaptive treatments while maintaining or even improving treatment quality.
Recommended by LinkedIn
6. Future of AI in Radiotherapy
Looking toward the future, Nikos shared ongoing clinical trials that explore how AI can further personalize treatment. He discussed a trial focusing on brain tumors, where AI could identify different areas of the tumor - such as the core, necrotic zones, and proliferative areas - and adjust the radiation dose accordingly. This type of personalized treatment could reduce toxicity in healthy tissue while increasing the dose in areas of the tumor that are more aggressive.
He continued by describing another trial for head and neck tumors, aimed at improving the effectiveness of combining radiotherapy and immunotherapy. In past trials, this combination was unsuccessful because radiotherapy damaged lymphocytes, weakening the immune response needed for immunotherapy to work. The new trial will attempt to preserve circulating blood, particularly lymphocytes, by adding dose constraints on arteries. The trial will compare three approaches: standard radiotherapy, radiotherapy with preserved blood flow, and reduced dose radiotherapy combined with immunotherapy, with the goal of achieving better outcomes by protecting immune cells during treatment.
7. Conclusion
Nikos concluded by emphasizing the transformative potential of AI in radiation oncology. He reiterated the importance of data quality, cautioning that AI models are only as good as the data they're trained on. He highlighted the need for standardization and clinical validation to ensure that AI systems can truly improve treatment outcomes.
He also reminded the audience that while AI can bring efficiency and accuracy, it must be continuously refined in collaboration with domain experts to integrate seamlessly into the complex world of clinical medicine.
The presentation ended with a Q&A session, where Nikos addressed questions related to the economic feasibility of AI in radiotherapy, the integration challenges with existing hardware, and how AI can help standardize treatments across different centers. He closed by reinforcing his belief that AI has the potential to revolutionize radiotherapy, ensuring better, faster, and more personalized care for cancer patients worldwide.
👉 If you'd like to discuss the potential of AI automation for your radiotherapy department or get a demo...
...in the Nordics or Benelux, contact: András Szentmiklóssy from Human Bytes via as@humanbytes.ai.
...in other countries, contact Ludwig Roppelt at l.roppelt@therapanacea.eu
📢 Finally, make sure to register for our upcoming educational AI webinar:
How can an unstructured clinical conversation or patient consultation get directly documented into your local EHR, according to your local template with all the right ICD-10 codes?
How does world class dictation in e.g. radiology or pathology look like? And how does speech recognition work with various dialects.
This webinar will discuss the latest and greatest within speech recognition, and will be conducted in English, but Danish and Nordic language use-cases will be demonstrated.
When: November 11th @ 15.00 CET. Read more at
See you at one of our next educational webinars.
Team Human Bytes