Deep-learning cross-modality super-resolution imaging of the cellular network in bone

Many organisms, such as bone, depend on fluid-filled network structures for cellular transport, signaling and mechanosensing. Bone is made of a network called the lacunocanalicular network (LCN), which consists of lacunae that house osteocyte cells connected together by a network of 300-nm wide canals called canaliculi. An understanding of this vast network is incredibly important and could provide valuable insight into bone function and disease. While many techniques exist to visualize LCN, the large size of the network makes acquiring images at sufficient resolution while covering a large field of view difficult.

The proposed research aims to use and develop deep learning image-to-image translation to improve the resolution of microscopic images of bone and enable graph network extraction of the LCN directly from low-resolution (LR) images. It is hypothesized that deep learning can be used to perform same-modality super-resolution (SR) of LR confocal microscope images, as well as cross-modality SR of LR confocal microscope images to high-resolution Focused Ion Beam-Scanning Electron Microscopy (FIB-SEM), and to directly extract large area graph networks more efficiently.

Faculty Supervisor:

Kathryn Grandfield

Student:

Partner:

Université Grenoble Alpes

Discipline:

Engineering

Sector:

Education

University:

McMaster University

Program: