Project Launch of IMPACT-OC: Interpretable AI and Digital Twins for Ovarian Cancer
The four-year project IMPACT-OC (Interpretable Multimodal Predictive Agents and Digital Twins for Ovarian Cancer) started on 1 October 2026.
The European Innovation Council (EIC) funds it as a Pathfinder Challenge project. Ten partners from five European countries take part, led by Univ.-Prof. Priv.-Doz. Dr. med. Ramona Woitek from Danube Private University (DPU, Austria).
Most ovarian cancers are diagnosed at an advanced stage. Clinicians have to make difficult choices early on: operate first or start with chemotherapy, decide which patients to refer for genetic BRCA/HRD testing, and judge whether an unclear mass on the ovary is benign or malignant. The information they need is spread across imaging, pathology, genomics and clinical records.
IMPACT-OC develops an interpretable AI platform that brings these data together. Specialised AI agents each analyse one type of data: ultrasound, CT and MRI images, pathology slides, genomic data and clinical notes. A super-agent combines their results, checks them against clinical guidelines and reports how uncertain they are. A digital twin models how each patient's disease develops. A dashboard prepares the results for the tumour board. The system is tested only in shadow mode: it runs in the hospitals alongside routine care, but its outputs do not influence treatment. Patients and patient organisations are involved from the start.
Hospitals cannot simply pool patient data in one place, and each site has relatively few cases. At IBT, our team (Maria Francesca Spadea, Ciro Benito Raggio, Jonathan Frömel) therefore leads the development of the federated learning infrastructure. Each site trains models on its own data, and only model updates are shared and combined into a common model. Patient data never leaves the hospital. We also lead the deployment of a secure, read-only test environment in which the consortium evaluates the integrated system in shadow mode. In addition, we help generate synthetic imaging data for dry runs and benchmarking.
We look forward to finding out, together with our partners, how well these tools work across clinical sites.
The IMPACT-OC consortium: Danube Private University (DPU, Austria), Duervation GmbH (DUE) and ecoplus (Austria); Karlsruhe Institute of Technology (KIT, Germany), Heidelberg University (UHEI, Germany) and NEC Laboratories Europe (Germany); Fondazione Policlinico Universitario Agostino Gemelli IRCCS (FPG, Italy) and Politecnico di Milano (POLIMI, Italy); University of Cambridge (CAM, United Kingdom); Institut régional du Cancer de Montpellier (ICM, France).
Funded by the European Union under Horizon Europe (EIC), Grant Agreement No. 101306266. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Innovation Council and SMEs Executive Agency (EISMEA). Neither the European Union nor the granting authority can be held responsible for them.