Project: Accelerated Diffusion MRI tractography
Description
Diffusion MRI tractography is a non-invasive method of mapping the white matter connections of the brain. Current techniques rely on discrete voxelized data to track these connections which introduces artefacts due to interpolation.
In recent work, continuous representations have been explored for modelling the diffusion MRI signal. Preliminary testing shows that using these continuous representations may also prove helpful in tractography.
Alternatively, the project could focus on applying CPU/GPU optimisation techniques to tractography algorithms. Current implementations are simple compute kernels and lack sophisticated techniques for cache locality, data streaming, or occupancy. Building on the wavefront paradigm, this opens up many avenues for complex tracking methods on the GPU and your own ideas.
If you are interested in exploring the uses of continuous representations for tractography or would like to dive into GPU computed tractography, please let us know and we can discuss!
Details
- Student
-
ZDZitong Di
- Supervisor
-
Ruben Vink
- Secondary supervisor
-
Maxime Chamberland