LAMBERT Zoé

Zoé Lambert is an Associate Professor at the University of Rouen Normandy and a member of the LITIS laboratory. She completed her PhD in Applied Mathematics (2019–2022) at INSA Rouen Normandie, funded by the Normandy Region, with a thesis on hybrid variational and deep learning approaches for medical image segmentation. Prior to her PhD, she worked as a Research Engineer at INSA Rouen (2018–2019) on weakly supervised learning for medical imaging and as a Data Scientist Intern at AID in Paris (2018). Her current research focuses on integrating geometric and topological constraints into deep learning models for medical imaging and coastal cliff crack detection.

Research Interests

Zoé develops advanced image segmentation methods combining variational models and deep learning, with emphasis on:

• Topological prescriptions in CNNs
• Geometric constraints for anatomical accuracy
• Registration techniques via nonlinear elasticity
• Public dataset creation for crack segmentation
• Directional total variation decomposition models
• AI applications in energy transition (Prioreno tool)

Scientific Contributions

She co-organized the SegTHOR challenge at IEEE ISBI’19, which automated segmentation of thoracic organs-at-risk for radiotherapy. Her work on weighted total variation regularization and Mumford-Shah terms has improved segmentation reliability in scenarios with limited data. She also contributed to the DEPHY3GEO project at CEREMA (2023–2024), focusing on cliff crack detection using thermal infrared imagery.

Honors and Collaborations

Zoé won first place in the Normandy Datathon (2023) for her energy vulnerability decision support tool. She served on INSA Rouen’s Scientific Council (2020–2022) and laboratory council (LMI). Her collaborations span LITIS, LMI, CEREMA, INRIA, and international institutions like INSA Lyon and the University of Bordeaux.