PhD Position at Mines Saint-Étienne: Inverse-approach 3D microscopy image analysis, application to the study of structure-function relationships in aortic aneurysm tissues.
This work will be carried out within the STBio department of the Centre for Biomedical and Healthcare Engineering (CIS) of Mines Saint-Etienne, and the DVH team of the INSERM SAINBIOSE laboratory, both on the Saint-Etienne Campus Santé Innovation site. The PhD candidate will be supervised by Claire Morin, Pierre Badel and Laurent Navarro.
As part of the work carried out on aneurysm mechanics within the STBio department of the Centre Ingénierie et Santé (CIS), the use of an inverse approach for the analysis of microstructure images and the reconstruction of models has been identified as very promising. However, many artifacts related to the image acquisition method make the exploitation of images complex: presence of shadows, fibers that appear thicker than in reality, strong attenuation of the signal in the thickness of the tissue. Consequently, conventional methods applied to these images make it possible to determine only a very limited number of microstructure characteristics, these characteristics being vitiated by errors. Thus, the proposed approach in this thesis consists in (1) using the acquisition physics of the confocal microscope to correct the acquired images and then (2) constructing synthetic images having the same characteristics as the microstructures in the sense of different criteria (morphological, spectral, statistics, ...) optimized to obtain a "clone" image of which one controls and knows all the characteristics.
This PhD project aims at developing an inverse approach methodology for the characterization of microstructures in aortic tissues, with the goal of reconstructing multiscale mechanical models. This involves integrating signal physics into the processing of microscope images. This first objective naturally induces a second and more fundamental objective which consists in the theoretical generalization of the reconstruction of complex images of biological tissues by image synthesis.
To address the first objective, conventional signal processing methods, driven by the specific physics of the multiphoton microscope, will be implemented to correct image artifacts. This objective is a prerequisite for the construction of synthetic images, in order to obtain a reliable comparison support.
For the second objective, the problem being strongly nonconvex, the work will be oriented towards methods embedding genetic algorithms and convolutional neural networks. In addition, it will be necessary to grasp the problem with multi-scale methods, given the intrinsic nature of multi-scale microstructures. Wavelets, and in particular the recently developed structured wavelets, will be an important asset.
Many images of microstructures of aneurysms have already been acquired in different mechanical configurations, which initially ensure a consistent database. The new CIS calculator will be a valuable asset as it will allow the implementation of algorithms on GPUs in particular; the methods envisaged at first glance will greatly benefit from this type of equipment.
Candidates are invited to send their CV before May 10th 2018 to firstname.lastname@example.org.
(c) GdR 720 ISIS - CNRS - 2011-2018.