113 Etiology in Lesion-Symptom Mapping: Tumor vs. Stroke (1) the voxels included in the LSM (minimum threshold of 3 lesions), (2) the voxels included in the LSM for both the stroke and the tumor group (minimum threshold of 3 lesions in both groups), (3) the total MNI brain volume. Post-hoc analyses Post-hoc analyses were performed on multivariate LSM results to directly compare cognitive performance between stroke and tumor in specific areas where the LSM findings in stroke and tumor diverged. Therefore, the interaction effect between lesion status and etiology was tested for specific atlas areas. Atlas areas that showed divergent lesion-symptom associations between stroke and tumor, despite sufficient lesion coverage, were selected for post-hoc analyses. For each subject, we recoded voxel counts per atlas area into damaged (≥5% of voxels affected) and not damaged. Next, areas that were damaged in at least 5 subjects of both the tumor and the stroke sample were selected. Post-hoc analyses were performed per atlas area to directly test whether there was an interaction between lesion status (damaged versus not-damaged) and etiology on cognitive performance. We anticipated that the assumption of normally distributed data would be violated in our dataset and selected a non-parametric alternative. A studentized permutation version of the Wald-type statistic (WTS), as implemented in the GFD R package,45 was used to test both the main effects of lesion status and etiology as well as their interaction effect. This WTS does not require normally distributed data or variance homogeneity, contrary to the more regular ANOVA statistic. RESULTS Clinical characteristics In the period between January 2010 and July 2019 254 treatment-naïve diffuse glioma patients (WHO grade II-IV) were scheduled for awake brain surgery and included in the retrospective cohort study. Of these 254 patients, a subset of 196 glioma patients could be included for analyses. Of the 222 first-ever cerebral stroke patients, 147 were included for this analysis (Figure 1). Average time between ischemic infarct and cognitive assessment was 7.9 weeks (SD = 4.5). Patients from the tumor and stroke group did not differ from each other regarding sex distribution, level of education and hand preference (Table 1). On average, the stroke patients were older, despite a comparable age range. Lesion volume was significantly larger in the tumor group (Supplementary Materials). While the stroke group had an equal distribution of left and right hemisphere lesions, most tumor patients had a lesion in the left hemisphere. Most tumor patients had a grade IV glioblastoma, IDH-wildtype, followed by grade II + III astrocytoma, IDH-mutant. 5
RkJQdWJsaXNoZXIy MTk4NDMw