130 Chapter 5 coverage in the stroke group. The majority of memory-regions found in either sample have been identified or are adjacent to regions found previously in tumor and stroke populations.24,46,54 The location of the fluency-related voxels, mainly in the left temporal and frontal lobe, aligns well with previous knowledge of the language system, and fluency in particular.26,47 Regions critical for letter fluency performance are located mostly in sensorimotor regions, while regions located around Broca’s area were critical for semantic fluency. This distinction in neuroanatomical fluency correlates was also found in a large-scale study in 1231 stroke patients.55 Similar to our results, they found lesion-symptom associations between frontotemporal regions and letter fluency, while more posterior temporal regions were crucial for semantic fluency. The importance of white matter to language functions in tumour patients was also found in a recent LSM study in patients suffering from left perisylvian gliomas.26 So, our study shows function-specific neural correlates for the memory and language functions that are corroborated by prior literature, supporting the validity of our LSM results in both populations. That said, the LSM presents large differences between tumor and stroke. This can partly be explained by the etiology-specific brain coverage that allows for investigation of distinct brain areas in tumor versus stroke. However, also in brain areas affected in both groups the overlap was minimal; we found no directly overlapping voxels between the LSM for tumor and stroke on any task. The similarities we did find mostly involved associated white matter tracts, in accordance with previous literature.24,46,55,56 For example, verbal memory tasks were associated with the left IFOF in both populations. Additionally, the left corticospinal tract and precentral gyrus were associated with semantic fluency performance for both populations. These results were confirmed in the combined LSM analyses in which the relation between lesion location and cognitive performance, irrespective of etiology, was tested. Differences did exist, however, between these combined and separate etiology-specific analyses with some brain areas being critical areas in only one of the two analyses. Again, correction for lesion etiology in the combined analyses was only possible in regions where both populations had adequate lesion coverage. Moreover, as was also shown by the statistical power of the univariate LSM, more voxels had adequate power in the tumor group. Thereby, the results of the combined analyses are likely mainly driven by the lesion data available in the tumor population, due to the uneven sample distribution at the voxel level.
RkJQdWJsaXNoZXIy MTk4NDMw