I am an AI researcher specialized in medical image processing. I recently completed my PhD on the topic of Uncertainty Quantification for medical image segmentation at Pixyl and Grenoble Institut des Neurosciences (GIN). I was supervised by Michel Dojat (INSERM), Florence Forbes (INRIA, STATIFY) and Senan Doyle (Pixyl).

My thesis can be found here (click on the image):

My work focuses on Uncertainty Quantification techniques in order to increase the robustness and reliability of automated segmentation models. My research is divided in several areas:

• Probability calibration
• Quantifying uncertainty at the pixel (or voxel)-level: Ensembling techniques, heteroscedastic models
• Structural uncertainty: identifying uncertain structures (e.g lesions, tumors) in the segmentation
• Out-of-distribution detection, particularly from the latent space of trained models
• Predictive intervals using Conformal Prediction
• Risk control for medical image segmentation

As side projects, I also worked on the following topics:

• Unsupervised Anomaly Detection using reconstruction Autoencoders
• White Matter Hyperintensities synthesis using 3D Diffusion models

Short bio

I obtained a Biomedical engineering degree at Télécom Physique Strasbourg in 2020, along with a Master IRIV (Imagery, Robotics, Engineering for the Living) from the University of Strasbourg. I then joined Pixyl in 2020, first as a Research engineer, then as PhD student starting May 2021.

News

• August 2023: Our 2 papers have been accepted to MICCAI 2023 Unsure Workshop, respectively on the topic of Out-of-Distribution detection and Predictive Intervals.
• May 2023: We made it to the 2nd position of the SHIFT challenge medical track !