You are here : EUR MANUTECH SLEIGHT > SLEIGHT's research projects

MALBOT - Research project

MAchine Learning for high definition BOne digital Twin

PhD student: Rehan JHUBOO, ED 488 SIS (Science, Engineering, Health)

ABSTRACT

The MALBOT aims at addressing the problem of understanding the aging of human’s bone tissue as well as bone loss dynamics by studying it through the lens of the effect of space exposure. We will make use of a dataset made available by the SAINBIOSE lab and composed of bone Computed-Tomography (CT) images of a cohort of astronauts collected before and after their flight to space. CT images often suffer from poor spatial resolution quality due to radiation dose regulation, hardening post-processing and image analysis of bone micro-structures. To address this spatial resolution issue, we aim at designing new deep learning-based super-resolution method. Because state of the art super-resolution techniques are failing in recovering morphometric and topological parameters of bone images, we plan to conceive a biology guided super-resolution model to improve the quality of images with respect to those bone parameters.

Result of a super-resolution task on one single bone image: LR (low-resolution), HR (high-resolution) and SR (super-resolution result)
Result of a super-resolution task on one single bone image: LR (low-resolution), HR (high-resolution) and SR (super-resolution result) - © R. Jhuboo - LabHC

 
PUBLICATIONS

CONFERENCES
  • Rehan Jhuboo, Ievgen Redko, Alain Guignandon, Françoise Peyrin, Marc Sebban. Why do State-of-the-art Super-Resolution Methods not work well for Bone Microstructure CT Imaging?. EUSIPCO 2022, Aug 2022, Belgrade, France. ⟨hal-03726811v1⟩
  • Rehan Jhuboo, Ievgen Redko, Alain Guignandon, Françoise Peyrin, Marc Sebban, Topology and Morphometry Guided Super-Resolution of Bone Microstructure CT Imaging, 22ème Journées Françaises de Biologie des Tissus Minéralisés (JFBTM) – La Baule-Escoublac - 27 au 29 Avril 2022