Skip to main content
. 2021 May 22;3(2):fcab110. doi: 10.1093/braincomms/fcab110

Figure 7.

Figure 7.

Cognitive impairments caused by stroke were predicted by atlas regions with unique lateralization effects. (A) Heatmaps indicate the associations of each anatomical region (Harvard-Oxford atlas)28 with lost (blue colour) or preserved (red colour) clinical outcomes in patients. The Rey Complex Figure Test (RC) exclusively featured right-hemispheric predictive relevances. Lesions of right pre- and post-central gyrus, lateral occipital cortex, the occipital pole and hippocampus had the greatest predictive relevance of lost function. The Mini-Mental State Examination (MMSE) stuck out amongst all four scores as it featured predictive relevances that were equally prominent in the left and right hemisphere. However, these relevances were mainly indicative of preserved function (red) in the right hemisphere and of lost function on the left hemisphere (blue). The Boston Naming (BN) and Seoul Verbal Learning Test (SVL) were both characterized by more pronounced left-hemispheric predictive relevances. These two tests differed in the exact region-wise distribution of the predictive relevances: lesions predictive of poorer naming function primarily extended from the mediotemporal hippocampal to further occipital areas, while verbal learning impairments were better predicted by lesions affecting opercular and insular cortices, as well as planum polare and temporale. (B) Prediction accuracy (R2 estimated based on posterior predictive checks) for the four cognitive scores along with scatterplots of actual (x-axis) and predicted (y-axis) cognitive performances. The MMSE model achieved the highest R2 prediction performance (55.8%). In view of 52.4% and 47.1% explained variance, the BN and RC model reached slightly lower scores. The explained variance of the SVL model totalled 39.7%.