Details of the internship: http://images.icube.unistra.fr/fr/index.php/Stage
Fluorescence microscopy has been revolutionized by super-resolution methods that overcame the resolution limits of conventional optical microscopes, and for which the Nobel prize was awarded in 2014. It is now possible to image the protein structure of small fundamental cellular units such as macromolecular assemblies, which were not observable in live cells until recent years. This opens a new field of investigation that have been growing very rapidly in recent years. However, intrinsic physical and biological limitations of fluorescence microscopy still limit the impact of these techniques: Firstly, the 3D resolution in fluorescence microscopy is strongly anisotropic (the axial resolution is 3 to 5 times lower than in the lateral plane), and secondly, the fluorescent proteins only label small parts of the structures of interest.
To overcome these issues, we have recently proposed a multiview reconstruction method based on the single particle reconstruction paradigm : we image several replicates of a given particle at random orientation, and we reconstruct a single particle that represents a model of these multiple observations. The combination of multiple views allows us to obtain high isotropic resolution, and to compensate for the partial labelling in the input particles. This method is restricted to a specific class of microscopy modalities, covering confocal and stimulated emission depletion (STED) microscopy. In this project, we want to extend the single particle reconstruction to a new class of modalities called single molecule localization microscopy (SMLM), which is able to reach the best resolution in fluorescence imaging.
The data acquired in SMLM differs from the images that are considered before: it is composed of point clouds with uncertainties associated to each point. The reconstruction has to be adapted to this new acquisition model. To this end, the trainee will have to implement a point clouds registration method inspired by computer graphics works, taking into account the individual uncertainties modeling the physics of the microscope. He will take advantage of the optimization framework developed in \cite. The reconstruction will be first realized on idealized synthetic data with gradually increasing complexity. The final goal is to apply the algorithm to real data to decipher the architecture of the centriole, an organelle present in most eukaryotic cells essential for cilia, flagella and centrosomes formation, in the context of the on-going collaboration with the Centriole Lab in University of Geneva.
The student will be a member of the IMAGeS team in the ICube laboratory in Illkirch. The internship will begin between January and May 2019, for a period of 6 months.
Profile of the candidate:
Contact: Denis Fortun, email@example.com
(c) GdR 720 ISIS - CNRS - 2011-2018.