From data to architecture in the neonatal brain : tractography development and comparison

• Main Author: Downing, Angela
• Other Authors: Rueckert, Daniel ; Edwards, David ; Hajnal, Jo
• Published: Imperial College London 2013
• Subjects: 610
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.570048

Reconstruction of white matter tracts in the neonatal brain offers valuable insight into the development and function of the brain. An accurate tractography algorithm is desirable for improving diagnosis and therapy for problems associated with premature birth. An investigation into tractography, beginning with the development of a minimal-model tractography algorithm and progressing to a comparison of established tractography methods is presented. A novel method of tractography is proposed for use with clinical diffusion weighted MR images for the reconstruction of white matter tracts in neonates. The principle behind the algorithm is to simulate the diffusion process directly, using diffusion measurements obtained in the acquisition to guide the tractography. Reconstructions of simulated and phantom data proved successful and were easily verified as the ground truth was known. However, two significant problems presented themselves with in vivo data. Discrimination between tissues in the brain proved to be a confounding factor to a tractography algorithm reliant on the diffusivity properties. Secondly, the lack of ground truth for comparison does not allow an accurate assessment of the reconstructed tracts. The need for a method of verification motivated the second part of the project which presents a comparison of established tractography methods in the parcellation of the thalamus in neonates. Four tractography algorithms were considered with different methods of initiating the tracks. Normalisation of the connectivity distributions used in the thalamus parcellation is proposed to eliminate bias in the results arising from variation in the topology of the cortex. The STAPLE algorithm was used to estimate the true segmentation from the candidate segmentations and this was used as the gold standard in the analysis. A whole brain approach to tractography with normalised connectivity was found to give the most consistent results in this group of subjects.