DWI Pre-Processing, Tractography and Analysis for Glioma Patients

Abstract

Abstract Background and purpose: Glioma is the most common type of brain tumor, and there is a continuous need for research on the relations between glioma related effects on the brain and patient health related quality of life. Brain imaging gives the potential to investigate changes in the brain. This thesis aims to facilitate the future work of a research group at Sahlgrenska University Hospital by developing a workflow including pre-processing of diffusion-weighted magnetic resonance images (DWI), performing tractography, and analyzing the resulting brain tracts. Methods: The open-source white matter segmentation tool TractSeg was applied after image pre-processing steps including removal of noise, Gibbs ringing artifacts, susceptibility-induced distortions, eddy currents, motion artifacts, and bias fields, as well as registration to MNI space. Tract approximations of 16 glioma patients were compared to those of 42 healthy subjects from the Human Connectome Project using TractSeg’s Tractometry module. The comparison was based on fractional anisotropy and fiber orientation distribution peak length. Results: TractSeg successfully computed 99.0% of the tracts for the 16 glioma patients and demonstrated indications of ability to handle the skewed brain anatomies observed in glioma patients. The comparison of fractional anisotropy between patients and healthy subjects revealed that the most affected brain tracts in the cohort were the anterior thalamic radiation, corpus callosum, cingulum, and striato-frontoorbital tract. The comparison based on peak length indicated the most affected tracts to be the anterior thalamic radiation, corpus callosum, striato-fronto-orbital, thalamic-premotor, and thalamo-parietal tract. Conclusions: The most affected brain tracts, as determined by fractional anisotropy and peak length comparison, were mostly located in the frontal lobe, which is consistent with previous findings related to glioma localization. TractSeg exhibits potential for performing tractography on glioma patient images and generating accurate tract approximations. The absence of ground control brain tract outlining poses challenges in quantitatively evaluating TractSeg performance on glioma-induced shifted brain anatomy and the following tract analysis. To enable future research, it is advisable to conduct a comparison with a control method to evaluate the performance of TractSeg.