Entitled “European Unified Cancer diagnostics and treatment AI agent,” or “EUcanAI” for short, the project is developing a multi-stage AI-based system to support precise diagnosis and treatment for patients with brain tumors.
Interconnected AI systems for better and faster management of brain tumor patients
“The workflows involved in treating patients with brain tumors and diagnosing these tumors are highly complex and largely rely on individual specialist expertise,” says the project’s coordinator Professor Felix Sahm from the Medical Faculty Heidelberg at Heidelberg University and Medical Director at the Department of Neuropathology at Heidelberg University Hospital. “Especially outside the highly specialist centers, inconsistent results are a common problem that can delay subsequent treatment.” This is where EUcanAI comes in: Interconnected AI systems, each specialized in one of the individual treatment steps—patient consultation, diagnostics and therapy planning—will help deliver rapid, reliable and compliant care to those affected. The researchers are planning to make the systems capable of interacting and building on one another in a modular fashion.
The plan is to have one AI model, with the patient’s consent, listening in on their medical consultations, asking its own questions and suggesting further investigations or examinations as appropriate. A second AI model will analyze and evaluate MRI scans and brain tumor tissue samples, while a third will identify and propose the most suitable options for the patient’s therapy either during or shortly after tumor surgery. All three models are designed to make clearly explained suggestions for the treating physicians, who will always have the final say on individual steps.
Bespoke AI models for medical image data
In Bonn, Professor Philipp Vollmuth, who is Else Kröner Professor for Artificial Intelligence in Medical Imaging at the University of Bonn and head of the Computational Radiology & Clinical AI Section at the Neuroradiology Clinic at the University Hospital Bonn, will be working with his team to develop foundation models for medical imaging that can be used as bespoke tools for agentic AI systems. In particular, they will support analyses of the MRI scans. “These models learn from vast quantities of medical image data, largely under self-monitoring, and can then be tweaked to suit different clinical tasks,” explains Prof. Vollmuth, who is also member of University of Bonn’s Life and Health Transdisciplinary Research Area. “Within EUcanAI, we’re tailoring them specifically for neuro-oncology together with our European and South Korean colleagues so that they can characterize brain tumors, assess the progress of therapy, and so on.” The image analysis techniques and models developed will subsequently be integrated as tools into the various specialized AI agents.
A single “master agent” bringing everything together
An overarching AI agent coordinates the specialized models and pools their findings to support clinical decision-making. This agent can draw on the existing AI system “Hetairos,” which can classify a large number of brain tumors in just a few minutes using conventional microscopic tissue samples. The AI system was developed by a team led by Sahm and Professor Moritz Gerstung froInstitutions involved and EIC Pathfinder Challenges funding
Link to EUcanAI:
https://ec.europa.eu/info/funding-tenders/opportunities/portal/screen/opportunities/projects-details/43108390/101306931