Figure 5.

Specific lesion atoms show strong lateralization for the Mini-Mental State Examination (A) and Boston Naming Test (B). Among all candidate lesion atoms, four distributed lesion patterns exerted pronounced hemispheric differences in predictive relevance to forecast future cognitive performance. The Boston Naming (BN) and Seoul Verbal Learning Tests (SVL, c.f., Fig. 6) were characterized by most lateralization effects of three lesion atoms. Two lesion atoms were substantially lateralized for the Mini-Mental State Examination (MMSE). Generally, we derived lateralization effects for lesion atoms by subtracting its marginal posterior distribution of one hemisphere from the marginal posterior distribution of the other hemisphere. We assumed a substantial lateralization when the distribution of the difference did not overlap with zero. In MMSE, hemispheric differences arose from right-hemispheric positive and left-hemispheric negative predictive relevances (positive: predictive of preserved function, negative: predictive of lost function). Left lateralization for BN originated from the difference between negative left-hemispheric and neutral right-hemispheric predictive relevances (exemption: lesion atom 3 was lateralized due to negative left- and neutral right-hemispheric relevances). Left columns: Plots visualize the joint density for combinations of parameter weights for age and each of the four lesion atoms. Age is plotted on the x-axis, weights of lesion atoms on the y-axis. Left-hemispheric lesion atom weights are shown in orange, right-hemispheric ones are shown in yellow. Right columns: Differences between left- and right-hemispheric lesion atom-wise predictive relevances. Figures are shown slightly transparent, in case of a non-defensible difference and thus absent lateralization (zero inside of 94% highest-posterior density interval).