PhD position: Machine learning for understanding the aging of a human bone due to space exposure

Hubert Curien, CREATIS and Sainbiose laboratories seek to recruit a PhD student working on an interdisciplinary project at the crossroads of medical imaging and machine learning.

Keywords Medical Imaging, Machine Learning, Generative Adversarial Networks, Transfer Learning, Bone Aging, Tomography, Astronauts.

Background In this interdisciplinary project, we aim at addressing the problem of understanding the aging of human bone tissue as well as bone loss dynamics by studying it through the lens of the effect of space exposure. This latter is known to be a good case study to analyze the accelerated aging of human bone highlighting the degradation that occurs in the bone tissue over a short period of time. To this end, we will benefit from a dataset made available by Sainbiose lab and composed of longitudinal low resolution bone images of a cohort of astronauts collected before and after the flights to the space. From this valuable data, we plan to learn a robust model that will be able to correctly predict the potential structural characteristics of the aging bone for previously unseen subjects.

Work Plan From a methodological perspective, the objective would be to design new machine learning tools able to address the major challenges associated to the above-mentioned task in two complementary ways. First, our goal is to propose new super-resolution methods that would allow to enhance the quality of the observed images as the resolution of conventional clinical X-Ray scanners now critically limits the exploration of the 3D architecture and porosities of human bone samples. Second, we plan to take advantage of the learned model for a better clinical prediction of bone porosities. To this end, we will leverage the recent advances in transfer learning, a subfield of machine learning that aims at training and deploying models on samples from different distributions, in order to allow the generalization of the learned model to patients in rheumatology. The generalization of such a model will be tested on the real-world CT images of a varying resolution or doses of radiation.

Skills We are looking for a student with a strong background in machine learning with a general understanding of transfer learning, generative modelling and super-resolution methods. Knowledge in medical imaging would be appreciated but is not mandatory. Also, strong coding skills in Python and proficiency in using machine learning toolboxes relevant to the indicated research techniques is a must.

How to apply? Send your CV, a motivation letter, and your academic records to Ievgen Redko (ievred.github.io) and Marc Sebban (https://perso.univ-st-etienne.fr/sebbanma/) using the following contact information: ievgen.redko@univ-st-etienne.fr, marc.sebban@univ-st-etienne.fr

Machine Learning group, Data Intelligence team
Hubert Curien laboratory
18 rue du professeur Benoit Lauras
42000 Saint-Etienne

Salary ≈ 1700 e/month