Annonce

Les commentaires sont clos.

Explainable deep models in cell imaging: application to the analysis of structural changes in human cells for diagnostic purpose

25 Mai 2022


Catégorie : Doctorant


PhD position: Explainable deep models in cell imaging: application to the analysis of structural changes in human cells for diagnostic purpose

PhD-XDeepCell.pdf

 

Context

In the framework of the French EUR project SLEIGHT, we propose a PhD position corresponding to the research program XDeepCell: Explainable deep models in cell imaging. This research program associates two labs of the University of Saint-Etienne: SAINBIOSE and Lab. Hubert Curien (LabHC).

The general objective of this project is to understand the structural modifications of cells after treatment (known drugs or biological samples of interest) and to propose a method of statistical quantitative characterization allowing to evaluate the degree of functional damage linked to the toxicity or to the considered pathology.

Scientific objectives

The goal is to propose explainable deep models to assess the degree of functional impairment related to the toxicity or the pathology. We plan to explore representation models encoding a vector quantization structure such as Fisher vectors that we will associate with self-learning methods to overcome the lack of annotations. Fisher coding is a generalization of visual bags of words that allows to represent the distribution of visual descriptors as a mixture of Gaussians. They have recently been integrated into deep models and are particularly well suited for describing textures 1 2. Self-learning consists in formulating a learning problem from unannotated data, for example by learning a representation space allowing to reconcile two distorted versions of the same image 3.

Keywords: Computer vision, deep learning, explainability, representation learning, morphology, cell structure, structure-function relationship, therapeutic drug screening.

Profile of the candidate

We are looking for an outstanding and highly motivated candidate with a Master degree in Computer Vision or Image processing with a strong knowledge in deep learning and machine learning and strong programming skills in Python. A background in the field of biology will be appreciated.

Information to apply

The application consists of a motivation letter, CV (with detailed list of courses related to computer science and computer vision), list of publications if applicable, names and contact details of two references, and transcripts of grades from under-graduate and graduate programs. Applications should be submitted via electronic mail to

Practical information

  • Duration: 36 months starting from October 2022
  • Location: Lab. Hubert Curien, campus Carnot, University of St-Etienne.
  • Supervisors: Prof. Christophe Ducottet, LabHC. Co-supervisors: Dr. Olivier Delézay, SAINBIOSE, Dr. Damien Muselet, LabHC.

  1. Tang, Peng, et al. “Deep FisherNet for image classification.” IEEE transactions on neural networks and learning systems 30.7 (2018): 2244-2250.↩︎

  2. Xu, Sixiang, Damien Muselet, and Alain Trémeau. “Deep Fisher Score Representation via Sparse Coding.” International Conference on Computer Analysis of Images and Patterns. Springer, Cham, 2021.↩︎

  3. Grill, Jean-Bastien, et al. “Bootstrap your own latent-a new approach to self-supervised learning.” Advances in Neural Information Processing Systems 33 (2020): 21271-21284.↩︎