Evaluating tractography algorithms for pre-surgical evaluation of patients with epilepsy

Tractography is a method that can be seen as a supplement to conventional imaging modalities of the brain. It is of relevance to different research and clinical areas, including pre-surgical planning for epilepsy. Although tractography was heavily researched in the past, there is no gold standard for algorithms, processing or interpretation of results. For reliable practical usage, better methods for calibrations of parameters have to be found.
This thesis aims to provide information about parameters of two different tractography algorithms, by determining the sensitivity of connectivity, measured as weighted streamline counts and the sensitivity of low-level geometry, measured as streamline densities and main tract directions, towards parameter variations. The analysis is done locally, using a reference constellation of parameters that made use of default values proposed by other researchers and a small prior qualitative assessment.
Analysis was successful for the parameters "cutoff" and "maximum fiber length" of the algorithm iFOD2, but showed several limitations for other parameters. In contrast, the parameters "maximum attempts" and "trials" of iFOD2 did not affect connectivity on a local scale more than chance. Similarly, the parameters "start temperature", "end temperature" and "iterations" of the Global Gibbs tractography were found to be irrelevant for both connection and low-level geometry. These findings may allow for dimensionality reduction. However, this would have to be confirmed through further investigation.