Bioimage Mining Group

We study how molecules and cellular structures move and interact inside living tissues. Our goal is to unravel intracellular dynamics in deep tissues by combining stochastic modeling, AI models, and the co-design of microscopy experiments and analysis methods.

The group is led by Philippe Roudot, CNRS researcher at Institut Fresnel and group leader at CENTURI (Turing Centre for Living Systems), in Marseille, France. It is part of the Information and Photonics theme within the Phyti group at Institut Fresnel.

Research

Tracking intracellular motion is difficult when signals are weak, structures overlap, and light scatters in tissue. We work on three complementary approaches:

Research diagram showing statistical models for trajectory uncertainty, attention networks for particle associations, and tailored illumination for signal detection in tissue.

Publications

See Philippe Roudot’s Google Scholar profile for publications, or HAL for deposited manuscripts.

Microscopy volumes shown from several viewpoints with colorful three-dimensional trajectory and bounding-box reconstructions.

Current team

Philippe Roudot — Group leader & Principal Investigator
CNRS researcher at Institut Fresnel and CENTURI group leader. Stochastic modeling, bioimage analysis, and microscopy co-design.
Sophie Carneiro — Postdoctoral fellow
Bioimage Mining Group, Institut Fresnel / CENTURI.
Piyush Mishra — PhD student
Deep learning for weakly supervised inference of molecular dynamics in fluorescence images of physiological environments. Joined for a CENTURI CMB master’s project in 2023, followed by a PhD at Institut Fresnel.
Mohamed Bacar Abdoulandhum — PhD student
EndoEmbryo: hybrid AI to understand how intracellular trafficking shapes embryo morphogenesis. Co-supervised with Nicolas Bertaux and Claudio Collinet (IBDM); PhD started in 2025.

Group profile at CENTURI

Alumni

Elhamou Ali — Master’s internship, six months, 2025
Worked on the MetaMorphogenesis project, using stochastic filtering to study cellular dynamics during Drosophila metamorphosis. She evaluated StarDist segmentation models on noisy 3D in vivo microscopy data and developed a hybrid approach combining deep-learning segmentation with Kalman filtering to improve coherent nucleus reconstruction and cell tracking. Master 2 in Complex Systems Engineering, Université de Picardie Jules Verne.
Jules Vanaret — PhD, 2021–2025
OrganoDyn: reconstructing organoid dynamics across scales to understand how cell motions build tissue shapes, at Institut Fresnel, I2M, and IBDM.
Subsequently joined EPFL’s CIME and Center for Imaging as an engineer, working on image analysis and electron microscopy.
Laura Neschen — Master’s internship, 2023
Automated hyperparameter optimization for a tracking algorithm using Bayesian optimization, at I2M (April–October 2023).
Subsequently worked as a computer vision engineer at IBDM, then started a PhD at Inria / Université Grenoble Alpes in October 2024 on 3D reconstruction and correspondence estimation for deforming objects.
Agentan Parthipan — Master’s internship, 2022
Simulation of fluorescence tomography experiments and inference in silico. Master 2 student at Université Paris-Saclay.
Théo André — Master’s internship, 2022
Multiscale stochastic filtering for detecting endocytic events in the fly embryo. Master 1 student at Aix-Marseille Université.
Continued with a PhD at Heidelberg University, studying mathematical models of biological dynamics.
Marwa Bouhadouza — Master’s internship, 2021
Multiple particle tracking under structural constraints, applied to intermediate-filament dynamics. Master’s student at Aix-Marseille Université.
Sandeep Manandhar — PhD collaborator, 2016–2019
Worked with the group through the Inria / UT Southwestern collaboration on estimating and assessing 3D optical flow in light-sheet fluorescence microscopy, including methods for evaluating motion fields without ground truth.
He is now a Research Scholar at Emory University School of Medicine, applying deep learning and motion analysis to medical imaging.

Funding

Earlier funding and fellowships

Teaching

Community activities

Research software and transfer

Contributions include u-track 3D, particle-tracking methods transferred to Amira (Thermo Fisher / FEI, 2016–2018), and consulting on 3D navigation and multidimensional data visualization for Arivis (2017–2019).

Contact

Philippe Roudot
philippe.roudot@univ-amu.fr
Institut Fresnel / CENTURI, Marseille, France