Will AI Take Over Radiation Therapy by 2040?
Image Credit: National Cancer Institute | Unsplash
The closing debate at the ESTRO 2024 Congress in Glasgow focused on a thought-provoking motion about the future of radiation therapy care pathways. The central theme was whether AI could take over the delivery of radiation therapy entirely by 2040. The debate, playfully titled “Clash of the Tartans – the great AI debate”, brought together leading experts in the field to discuss the potential impact of AI on cancer treatment and whether human expertise would still play a crucial role.
The Case for Automation: Andrew Hope's Vision
Andrew Hope, from the Princess Margaret Cancer Center in Toronto, championed the idea that automation, driven by artificial intelligence, is not just inevitable but crucial. He painted a picture of an overburdened healthcare system, where the rising global cancer rates necessitate AI-driven interventions to meet the growing demand. Hope emphasized the exponential growth of AI, noting how advancements like AlphaGo have given way to applications transforming daily life. He argued that AI-driven robotics in radiation therapy could scan, plan, and treat patients in real-time, vastly improving efficiency. However, he assured that this wouldn't replace human jobs but would allow doctors to focus on patient care, handling complex cases, and improving AI models.
AI Isn’t Ready Yet: Dekker’s Counterargument
On the other hand, medical physicist Andre Dekker of MAASTRO Clinic disagreed with the notion that AI will dominate radiation therapy in the near future. Dekker emphasized that AI, despite its progress, remains "stupid" when compared to human intelligence. He argued that AI lacks the ability to generalize across different scenarios, such as identifying a liver in different scan orientations. Furthermore, Dekker highlighted AI’s inherent biases, noting that its shortcomings, such as generating stereotypical or incorrect results, could lead to suboptimal care for patients, particularly in low-income regions. His stance was clear: AI will not be able to fully replace the nuanced and specialized work done by humans in radiation oncology.
Can, Should, and Will We Automate? Stine Korreman Weighs In
Stine Korreman of Aarhus University acknowledged the rapid advancements in AI within radiation therapy, even showcasing AI systems that can automatically plan treatments. However, Korreman focused on the ethical and practical implications of relying entirely on AI. She noted that while AI can process vast amounts of data more effectively than humans, the errors it makes often stem from the people behind the AI models. Korreman also drew a comparison to self-driving cars, arguing that although AI has the potential to reduce errors, human oversight remains necessary. She foresaw that while the current generation might not fully adopt AI-driven radiation therapy, younger generations will push the boundaries, comfortable with the technology’s capabilities.
Human Touch Is Irreplaceable: Eliana Vasquez Osorio’s Perspective
Eliana Vasquez Osorio from the University of Manchester stood firmly against the notion of fully automated radiation therapy. She emphasized that while AI can optimize treatment delivery, the care pathway involves much more than algorithms. For example, oncologists make critical decisions by considering a patient’s physical condition, not just their medical data. Similarly, medical physicists ensure the safety and reliability of treatment systems, drawing on their expertise to prevent disasters, much like the failures seen with aviation automation systems. Vasquez Osorio highlighted that radiation therapy also relies on daily interactions between radiation therapists and patients, where empathy and human judgment play a key role — something bots simply cannot replicate.
The Verdict: A Future of Human-AI Collaboration?
At the end of the heated debate, the ESTRO delegates overwhelmingly voted against the motion, with 83% opposing the idea that radiation therapy will be fully bot-driven by 2040. While the debate illustrated the immense potential of AI in healthcare, the consensus suggested that human oversight, empathy, and expertise will continue to be vital parts of the equation. As Korreman predicted, full automation may come eventually, but it will be driven by future generations more at ease with the technology.
Source: PhysicsWorld
