Abstract
The dissertation describes the research work on the inference of cerebral white matter fibres from diffusion tensor magnetic resonance images (DT-MRI), derived from the high angular resolution diffusion-weighted imaging (HARDI) data.
A novel framework for inferring cerebral white matter fibres from diffusion MR images is presented. It includes feature extraction using graph based methods; feature selection with statistical pattern recognition techniques; and the inference of the white matter fibres applying machine learning methods.
Four similarity measures are adopted or proposed for the fibre characterisation. Very good results are produced and a comparison is made. An evaluation of the methodology is conducted on real diffusion MRI data.