Influence of tractography termination criteria on structural connectivity matrices in glioma patients

Tractography is a powerful tool to investigate connectivity in glioma. State-of-the-art algorithms rely on the setup of several options: track termination criteria are among the most influencing parameters, in particular the cut-off value applied to the fibre orientation distributions’ peaks amplitude [1]. Although for healthy controls the suggested cut-off value is 0.1 [2][3], little is known about glioma patients. The aim of this work was to verify that the choice of this value does not significantly affect the estimates of Structural Connectivity (SC) matrices [4]. For each patient, the diffusion-weighted-images (b-values 0/710/2855 s/mm2, 100 directions) were processed using the MRtrix [5] software, exploiting the iFOD2 algorithm [6] and six progressively more permissive thresholds cut-offs [0.1:0.01:0.05]. Next, from the tractogram we computed connectivity matrix (SC) with two metrics (Number of streamlines (SCNS) and Mean streamline length (SCMSL)). Principal Component analysis (PCA) and Krzanowski test [7] were applied at the individual level to test whether there was a significant relationship between the connectivity structure and the cut-off value. PCA returned an average variance explained by the first principal component, computed among the patients, equal to (99.09 +/- 0.21%) (73.76 +/- 2.08%) for SCNS and SCMSL respectively. The Krzanowski test revealed no statistically significant differences between the two extreme-cut-off connectivity distributions (p-values>0.05 for both SCNS and SCMSL). Our results suggest that varying the cut-off value from the default does not produce significant changes in the structural connectivity matrix structure and thus, even in glioma, this value can be robustly exploited for the computation.