Comparison of parameter threshold combinations for diffusion tensor tractography in chronic stroke patients and healthy subjects

Abstract

Background: Although quantitative evaluation of diffusion tensor imaging (DTI) data seemed to be extremely important for clinical research its application is under debate. Besides fractional anisotropy (FA) the quantitative comparison between hemispheres of the number of fibers reconstructed by means of diffusion tensor tractography (DTT) is commonly used. However, the tractography-related parameters FA, minimum tract length (LENGTH) and the angle between two contiguous tracking steps (ANGLE) are inconsistently applied. Using 18 combinations we tested for the influence of parameter thresholds on the amount of reconstructed fibers for the posterior pyramidal tract in both hemispheres in order to obtain meaningful thresholds for DTT.

Results: In 14 chronic stroke patients with unilateral lesions of the pyramidal tract around the height of the internal capsule and considerable motor deficits a 3-way repeated-measures ANOVA showed a significant interaction between the effects of FA and ANGLE level on reconstructed fiber lateralization, F (2.9, 37.67) = 3.01, p = 0.044, and a significant main effect FA, F (1.4, 18.1) = 11.58, p = 0.001. Post-hoc pairwise comparisons showed that this interaction was completely driven by FA. In 22 right-handed healthy subjects no significant interactions or main effects could be found.

Conclusion: The parameter threshold combinations with highest FA showed highest effect. ANGLE and LENGTH insofar influenced the lateralization effect when selected as liberal as possible, short LENGTH and large ANGLE thresholds. The DTT approach should be used with great care since results are highly dependent on the thresholds applied.

Figure 1. Repeated-measures ANOVA: Interaction graphs for ANGLE*FA.

The graph illustrates that the FA threshold is essential for modulating the lateralization effect.

Figure 2. Repeated-measures ANOVA: Interaction graphs for FA*ANGLE.

The interaction graph, similar to Figure 1, but in the opposite way, again illustrates the crucial influence of the FA threshold.

Figure 3. Effect sizes of parameter threshold combinations.

The plot shows the effect size of each tested parameter threshold combination utilizing Cohens' d. The data were sorted by the effect sizes of the patient group.

Figure 4. Diffusion tensor tractography in a stroke patient.

The results of diffusion tensor tractography (left) in a patient who experienced subcortical stroke, causing relevant unilateral hand motor impairment. The lesion in the left posterior limb of the internal capsule is marked by a red circle on a T1-weighted anatomical slice image (middle), the de-normalized ROIs are shown as colored overlays on an FA slice image (right).