Digital Assistant for Cancer Treatment

The University Hospital of Zurich (USZ) is a hive of activity. It is nearly 2 p.m., and Andreas Wicki, Professor of Oncology at the University of Zurich, is walking briskly towards the consultation room. Today is Thursday, and the tumor board is about to convene for its weekly meet-up. There are more than 20 tumor boards at USZ, one for every organ. Similar meetings are also held weekly at the University Children’s Hospital Zurich. 

“Tumor boards have become an indispensable part of modern oncology,” says Wicki. They bring together selected medical specialists from the fields of surgery, radiology, oncology and pathology, as well as nurses. Their interdisciplinary collaboration enables them to develop treatment approaches tailored to the needs of their cancer patients.

In future, the basis for discussion at these meetings – patient data, medical treatment guidelines and imaging material – will be prepared and compiled by the AI tumor board. “The AI tumor board establishes AI, or artificial intelligence, in clinical practice,” says Professor Michael Krauthammer, Head of the Department of Quantitative Biomedicine at the University of Zurich.

The three-year research project AI Tumor Board is an incubator project of the research center The LOOP Zurich and is supported by the Promedica Foundation. Professors Andreas Wicki and Michael Krauthammer, as well as Jean-Pierre Bourquin from the University Children’s Hospital Zurich, have been working with their teams on the implementation of the project since fall 2024.

Soon, doctors will have a powerful digital ally when they first discuss the optimal treatment for a cancer patient. “The AI tumor board sifts through treatment guidelines, genetic profiles, molecular data, medical images, and previous treatment results at unprecedented speed,” explains Wicki. Of particular value is the system’s ability to take into account the specific biological characteristics of each patient’s cancer cells.

The AI tumor board can also uncover hidden patterns in clinical trials and patient data. It recognizes which therapies are most promising for given genetic profiles. “This will lead to faster and more accurate diagnoses and more personalized treatment recommendations,” Krauthammer says. This development gives hope to patients. They are more likely to receive the precise treatment that will deliver the optimal impact for their specific tumor. According to Krauthammer, this is a crucial step toward truly personalized medicine.

The new AI tumor board relies on a wealth of data – a clear case of the more, the better. “The biggest challenge is how to link the various databases from hospitals and scientific studies,” Krauthammer points out. Many of these systems do not communicate with each other or simply do not have suitable interfaces.

The solution that the team behind the AI tumor board is working on is both elegant and effective: “We’re going for a federal data structure,” explains Fabio Steffen, a postdoctoral researcher at the University Children’s Hospital in Zurich and a member of the development team. In practice, this means that valuable medical information stays where it is – on different, independent systems. However, it can still be linked and used without compromising privacy. This decentralized architecture makes it possible to break down data silos without actually moving them – a critical advantage in the sensitive context of patient data.

In a few years, the AI tumor board will be able to handle tasks that currently require time-consuming research by doctors. “This will take a lot of the workload off the medical staff, saving valuable resources and freeing up more time for patient care,” says Wicki.

Human experts will still make decisions about treatment, but they will be able to do so after a successful evaluation phase, based on much more comprehensive information and prepared in a much more systematic way. In the years to come, these digital assistants working in the background could become the decisive factor for cancer patients.

But one question remains: How reliable is the information that the AI tumor board will deliver? Although still in the development phase, the vision is clear: When the system is ready, it will not only compile relevant medical data, but also provide transparent explanations of how it reaches its conclusions.

“Unless the decision-making processes can be understood, even the most powerful AI is of little help. Physicians must be able to trust the results,” Bourquin explains. Another important aspect is that the decision will still ultimately be made by humans – a point that Bourquin emphasizes as essential for patient safety. “AI will not be making life-or-death decisions anytime soon. Humans will have the last word.” However, he acknowledges that artificial intelligence is becoming increasingly indispensable in medicine. Its ability to analyze huge amounts of data is opening up new perspectives.

As Wicki points out: “No physician can comprehend and compare the vast amounts of data available in modern medicine in such depth. Gone are the days when physicians used highlighters to skim through paper documents,” says Wicki. The biggest challenge for the developers of the AI tumor board is to create a system that not only provides excellent analyses but also communicates them in a way that people can understand. Only then will AI become a reliable partner and find its place in day-to-day clinical work.