Academic/Research

Partners of the project :
Laboratory GREYC - School of Engineer ENSICAEN
Laboratory SIC - University of Poitiers

Project FOGRIMI
Post Type: Full-time, fixed-term for 36 months

Pathologist have been using their tool of trade, the microscope, since the early 17th century and much of their analysis is visual. Therefore, pathology imaging has become a medical imaging environment of growing significance which presents a number of challenges. Recent imaging pathology workstations provide interfaces for virtual microscopy by Whole Slide Imaging (WSI). WSI produces however huge images for a unique slide (around 30 Go) and managing such masses of data is therefore a innovative challenge for this new era of digital microscopy. WSI images are provided as a single image at full resolution and by a sequence of images coding decreasing resolutions of the base image. WSI Images have the particular property to encode different structures at different resolutions. Such images thus combine intrinsically multi resolution and multi scale representations.

The Phd student will develop a new model encoding a multi-scale and multi resolution representation of WSI images using a top-down analysis scheme. Based on this model the PhD student will have to develop efficient region based segmentation algorithms using the MDL formalism and semi supervised learning methods. The model will be composed of two parts : a topological part based on 2D combinatorial maps (which may be roughly understood as 2D planar graphs) and a geometrical component based on discrete segments. The segmentation algorithms will be based on split and merge operations. The split operations will be based on graph based semi supervised learning methods applied within the feature space of the region. The over segmentation of the region will then be corrected using an MDL approach using apriori information about the content of the WSI images. Once a partition is obtained at one resolution, one of its region is down projected at a greater resolution and further processed.

The project is based on a collaboration between 4 laboratory (LUSAC, GREYC,SIC,LaBRI). The PhD student will perform half of its thesis in the SIC laboratory in Poitiers and the other half to the GREYC laboratory in Caen.

The PhD student should have a Master degree in computer science with a first research experience(e.g. through a project). He should be able to assimilate quite different formalisms (topology, graph theory, statistics) and to provide insights in some of these fields thought the project.A strong scientific background in computer science and an experience with programming in C/C++ under Linux/Unix is thus requested. A first experience in image processing and/or Graph representations would be appreciated. He will be able to demonstrate a keen interest in state-of-the-art research and development, working alongside academic partners to achieve goals within tight time constraints.

Contact:
For more information, please contact Luc Brun luc.brun@greyc.ensicaen.fr and Guillaume Damiand damiand@sic.univ-poitiers.fr.