127 Etiology in Lesion-Symptom Mapping: Tumor vs. Stroke Univariate lesion-symptom mapping results We performed univariate LSM with lesion volume control to corroborate results found in the multivariate LSM analyses. For the tumor population, worse performance on the RAVLT Direct Recall and Delayed Recognition was not related to specific lesion locations when using univariate analyses. For the RAVLT Delayed Recall, letter fluency and semantic fluency voxels related to worse performance on the univariate analyses largely overlapped with those from the multivariate analyses (see Supplementary Figure 7 and 8). Regions indicated by multivariate analyses seem to expand from the overlapping regions and encompass larger brain areas. For the stroke patients, univariate LSM did not indicate a significant relationship between lesion location and cognitive performance for the RAVLT direct recall, delayed recall or delayed recognition. For the two fluency tasks (letter and semantic) voxels related to worse performance on the univariate analyses largely overlapped with those from the multivariate analyses (see Supplementary Figure 9 and 10). Regions indicated by multivariate analyses seem to expand from the overlapping regions and encompass larger brain areas. As less voxels were significant in the univariate analyses for both the tumor and stroke group, the minimal overlap that was found with multivariate LSM decreased even further when using univariate LSM results. Statistical power To assess possible difference in power throughout the brain, statistical power maps were created for each univariate LSM for the tumor and stroke group separately (see Supplementary Figure 11). Adequate power was set at a maximum of 200 patients needed to replicate the results in 60% of the studies. The relative number of voxels with an adequate power was calculated relative to (1) the coverage for the specific cognitive task (minimum threshold of 3 lesions), (2) the overlap in coverage (minimum threshold of 3 lesions in both the stroke and tumor group), and (3) the total MNI brain volume. For each cognitive task, more voxels had adequate power in the tumor than in the stroke group (see Table 3). Lesion volume correction Both multivariate and univariate LSM were repeated without correcting for lesion volume (maps available in Repository). All LSM results were visually compared and no major differences between lesion volume corrected and uncorrected results were found. Corrected LSM tended to encompass smaller, but overlapping, areas with the uncorrected LSM. From the non-corrected LSM, voxels expanded from those regions, thereby indicating more voxels were significantly related to task 5
RkJQdWJsaXNoZXIy MTk4NDMw