123 Etiology in Lesion-Symptom Mapping: Tumor vs. Stroke cingulum, corpus callosum and ILF. For any two memory tasks this extended into the left fusiform gyrus, the temporal gyrus and optic radiations. Between the two different fluency tasks, there was an 8.6% overlap in significant voxels. These were located in and around the left insula, inferior frontal gyrus, rolandic operculum and arcuate fasciculus. For the lesion-symptom results in the stroke group there was a 4.9% overlap between voxels related to both memory and fluency task performance. Overlapping voxels were mainly located in the left putamen and the cortico-spinal tract. 1.3% of voxels were associated with all three memory tasks and 18.4% with any two memory tasks. For all three memory tasks the overlapping voxels were mainly located in the left IFOF and optic radiations. For any two memory tasks this extended into the left putamen and uncinate fasciculus and the right insula and IFOF. There were no voxels significantly related to both the phonological and semantic fluency task. Post-hoc analyses (Figure 7) Post-hoc interaction analyses were performed to further examine atlas areas that showed lesion-symptom associations for only one of the two etiologies, despite sufficient lesion coverage in both. A complete overview of the post-hoc results can be found in Figure 7. For the lesion-symptom associations found in the tumor group only, post-hoc analyses using a permutation version of the Wald-type statistic (WTS) revealed significant interaction effects between lesion status (damaged versus not damaged) and etiology (stroke versus tumor) in 6 out of 9 tested atlas areas. Interaction effects were found for semantic fluency performance in the left insula (WT = 5.08 p = .033), left rolandic operculum (WT = 9.40 p = .01), left Heschl’s gyrus (WT = 4.69 p = .048) and white matter areas of the left arcuate fasciculus (WT = 6.88, p = .016) (including the long segment (WT = 7.39, p = .015) and anterior segment (WT = 6.21, p = .023)), with worse performance for tumor lesions. For the association between the left internal capsule and semantic fluency there was only a significant main effect of lesion status (WT = 5.43, p = .04). This association was more pronounced in the tumor group, but a similar trend was present in the stroke group. The association between the left corticospinal tract and delayed recall performance was even stronger in stroke than in tumor, leading to a significant main effect of lesion (WT = 22.6, p = <.001) as well as a significant main effect of etiology (WT = 4.88, p = .045). The association between the left caudate and worse direct recall performance was not confirmed in post-hoc analysis. 5
RkJQdWJsaXNoZXIy MTk4NDMw