Efficient statistical methods for inference and model selection in diffusion-weighted MRI models
Publication Date
March 15, 2016
Creator
Abstract
Diffusion-Weighted Magnetic Resonance Imaging (DW-MRI) on the brain is a revolutionary method that provides in-vivo access to tissue macrostructure non-invasively (Basser et al., 1994). Recently, DW-MRI has been shown to have great potential in characterising brain microstructure, such as diameter and size distribution of neuronal fibres, features that were available so far only postmortem or through animal studies (Zhang et al., 2011). Using a process known as Tractography the existence of brain connections can be estimated using a set of DW images (Basser et al., 2000).
The main aim of this thesis is to develop efficient methods for studying Tractography within a Bayesian framework. In order to characterise the white matter in the brain we focus on the widely used partial volume model (Behrens et al., 2003). We describe methods that are both time and computationally efficient for estimating the parameters of the partial volume model, before reparametrising the model, so that parameter estimation is viable in some special cases. The partial volume model allows for multiple fibre orientations so we develop methodology to choose between the number of white matter fibres in a voxel. We then take into account the uncertainty in the number of fibre orientations and provide a Fully Probabilistic Tractography method as an alternative to existing Tractography algorithms. Finally we look into the Global Tractography model (Jbabdi et al., 2007) and develop efficient methods for inferring connections between brain regions by investigating methods based on Thermodynamic Integration.
Item Type
ethesis
Thesis Type
PhD
Supervisors
Subjects (LC)
Associated Schools / Departments
School of Mathematical Sciences (UK)
eprints ID
31173
UoN Repository URI
Except where otherwise noted, this item's license is described as
File(s)![Thumbnail Image]()
Name
Lisa_Mott_PhD_thesis.pdf
Type
Full-text
Description
Examined
Size
3.29 MB
Format
Adobe PDF
Checksum (MD5)
3536a984d9f114de0874ac45b31470de