An RHU project links artificial intelligence and oncology
The Hospital-University Health Research Action (RHU) piloted by the ANR aims to support translational health research or clinical research projects, which may be based on fundamental research in biology, epidemiology, social sciences or health economics.
An RHU project links artificial intelligence and oncology
The RHU AI-Triomph: optimizing clinical trials in oncology using artificial intelligence thanks to multimodal approach is coordinated and led by Prof. Magali Svrcek, Department of pathology, Saint-Antoine Hospital, AP-HP, AP-HP.Sorbonne Université. The other two co-coordinators of this project are Pr Jean-Baptiste Bachet, Gastroenterology Department, Pitié-Salpêtrière Hospital, AP-HP. Sorbonne Université, and Prof. Xavier Tannier, Professor of Computational Sciences, LIMICS, Sorbonne Université.
The RHU AI-TRIOMPH (Artificial Intelligence - clinical TRIals Optimization for oncology with Multimodal PatHology) aims to develop a cutting-edge artificial intelligence system to revolutionize personalized therapeutic approaches and optimize the design of clinical trials for three poor-prognosis cancers that represent a major public health problem in France and worldwide: pancreatic ductal adenocarcinomas, esogastric adenocarcinomas and recurrent thyroid cancers.
The tripartite consortium (Sorbonne University, AP-HP and Owkin) will generate multimodal data sets, including data from the AP-HP clinical data warehouse, digitized slides - building on the deployment of digital pathology at AP-HP.SU - and, for the first time, liquid biopsies. This methodology will first be applied to pancreatic cancer, and then extended to other types of cancer. Finally, an AI platform based on OWKIN's open source privacy-preserving Federated Learning (FL) software Substra will be implemented within the AP-HP infrastructure. This platform will enable clients to build external control arms using the consortium cohorts for their single-arm trials, providing early assessments of the activity and efficacy of experimental drugs without the need for data transfer.
