We are honoured to join a consortium of seven European research institutions, two reference hospitals, and three industry partners selected under Horizon Europe for a €9M programme on explainable clinical AI. The programme runs for thirty-six months and addresses one of the most pressing operational questions in oncology today: how do we surface an AI model's reasoning in a way that a pathologist can audit in the seconds they actually have during case review?
Our role in the consortium
PathologyAI coordinates the computational pathology workstream. Concretely, that means we are responsible for designing the reference benchmarks, defining the evaluation protocol for explanations, and contributing a pre-trained foundation model to the consortium's shared infrastructure.
Why this programme matters
Explainability in pathology is often framed as a user-interface problem. It is not. It is a scientific problem about what a model's intermediate representations correspond to in the tissue, and whether those correspondences hold up across laboratories. The consortium brings together the methodological expertise and the clinical data to begin answering that question properly.
