PhD defense Rehan Jhuboo

PhD defense Rehan Jhuboo"Super-Resolution for bone microstructure CT imaging"

at 2pm

Room J020 
Telecom Saint-Étienne Building,
Campus Manufacture,
42000 Saint Etienne

"Super-Resolution for bone microstructure CT imaging"


Osteoporosis is a skeletal disease characterized by a decrease in bone mass, degradation of bone microstructure including increased porosity and cortical bone thinning. It represents a major public health issue, affecting 39% of women aged 65, rising to 70% for those over 80. While not fatal, it significantly increases fracture risk, compromising patients’ quality of life and potentially leading to severe complications. High-Resolution Peripheral Quantitative Computed Tomography (HR-pQCT) is a technology that was developed to clinically to predict fracture risk and study various therapies by measuring 3D parameters characterizing bone microarchitecture. HR-pQCT has a maximum resolution of 82µm per voxel, corresponding to a real spatial resolution of 130-150µm. However, the smallest trabeculae in humans can reach 100µm. Due to maximum radiation dose regulations, no in vivo technology is currently capable of achieving higher resolution. Increased resolution would better describe bone microstructure, improving osteoporosis prediction and understanding. Only ex vivo technologies such as micro-CT (µCT) can achieve a voxel size of 10.5µm. In this thesis, we propose MiCT-SR, a Deep Learning (DL) Super-Resolution (SR) method to artificially enhance the spatial resolution of low-resolution (LR) bone images. Assessing capabilities of state-of-the-art methods applied to bone imaging, we find them unable to faithfully reconstruct bone microarchitecture despite improved visual quality. We suggest guiding SR methods with intrinsic application-specific information: bone morphometry. This guidance is made possible by developing a convolutional neural network capable of estimating the morphometric parameters of an image. The goal of the method is to faithfully reconstruct a correct representation of bone microstructure while obtaining a visually enhanced image.



  • Marc Sebban, Laboratoire Hubert Curien, Supervisor
  • Alain Guignandon, Sainbiose, Co-Supervisor
  • Françoise Peyrin, Creatis, Co-Supervisor
  • Anne Tournadre, Centre hospitalier univ. de Clermont-Ferrand, INRAE, Examiner
  • Elisa Fromont, Université de Rennes, IRISA, Reviewer
  • Maria Zuluaga, EURECOM, Reviewer