Olivier Humbert

“It often surprises people, but nuclear medicine is a major branch of oncology. We are the direct heirs of Marie Curie. The first nuclear medicine treatments were developed in the 1950s and have continued to evolve ever since,” says Olivier Humbert. After much deliberation between this and a career in engineering, this physics and film photography enthusiast ultimately chose a path where hard sciences and human commitment converge. During his medical studies, he specialized in nuclear imaging and therapies, which combine patient contact, technology, and cutting-edge research. Following a PhD in science on the identification of biomarkers through imaging to assess the aggressiveness of breast cancers and their response to treatments, Olivier Humbert arrived in Nice, collaborating with researchers at Inria from 2018 onward, before joining the Côte d'Azur Interdisciplinary Institute for Artificial Intelligence (3IA). He explains: "I didn't come here just for the sun, but also because I saw that the French Riviera was a land of AI that could open up multiple possibilities for collaboration on projects with engineers."

A leading federated learning project in France

“When you first encounter Artificial Intelligence, it seems magical. But you quickly realize that it doesn't work perfectly the first time; projects require real expertise, and it's a tedious process. AI will thus change our practices and ways of being doctors, but developing reliable, robust, and ethical models will require a lot of work and overcoming obstacles, particularly regarding data access,” continues Olivier Humbert. For this researcher, meeting Marco Lorenzi, who was working on collaborative AI software for hospitals, was pivotal. The idea of ​​this mathematician from the Epione team at Inria? To move away from centralized AI, which gathers all the data on a common server, and adopt a federated approach: “This way, patient data remains in each hospital, and only mathematical formulas are sent.”

A decision based on patient feedback: “For lung cancer, we have fantastic immunotherapies that stimulate the immune system, with incredible results even for patients in advanced stages. However, in half of the cases, this can actually accelerate tumor growth. We are therefore reaching a limit, which underscores the need to develop even more personalized medicine. AI could provide us with this precision by cross-referencing data from national and international patients who have already received this type of treatment,” the doctor continues. Currently, the in-hospital platform developed in Nice connects 2200 patients across 10 hospitals in the Alpes-Maritimes region and throughout France. He concludes: “We are going to initiate an international collaboration with Canada and Quebec. Developed for breast and lung cancer, this project could then be replicated in other medical fields.”