Fig 1.
Sternberg memory task for evaluation of WMem.
Table 1.
Spatial coordinates of local maxima in MNI space (mm).
Fig 2.
Results are represented by group means. Significantly different indexes between groups are labeled with an asterisk (*) for Bonferroni corrected p < 0.05. Nearly identical results are observed when analyzing the sub-groups used for each of the imaging analyses (S2 Fig).
Fig 3.
Functional connectivity differences between controls and TLE patients.
A): Adjacency matrix showing between-group pairwise connectivity differences. Full-colored cells depict group differences at an uncorrected p-value<0.05, while diamond-filled cells indicate significance after FDR correction. color-coded is used to depict the three components identified with NBS analysis (default-mode network: DMN; task-positive networks: TPN; and their interaction: DMN-TPN. Their corresponding spatial configuration is shown in B. C-E: Boxplots show group-wise values at the edge level for each component. Diamonds in A, and C-E represent significant between-group differences in pairwise node-node comparisons (FDR pcorr<0.05). Refer to Table 1 for ROI abbreviations. Brain slices and surfaces in this and Figs 4–6 correspond to the MNI 152 atlas.
Fig 4.
Functional connectivity and cognitive scores.
A) Significant associations between FC and cognitive scores in all participants (GLM) and B) associations between FC and cognitive scores with significant group interaction (ANCOVA). In the matrices, colored cells depict significant F-test values (p < 0.05). Scatter plots on the right side are shown to illustrate the linear relationships or group interactions between node-node pairs and cognitive scores (all scatter plots are available in S4C and S4D Fig). Functional connectivity is expressed as Fisher’s z. Pearson coefficients and p-values are shown color-coded for significant group-specific correlations. At the bottom, spatial representations of significant associations are depicted. Light-red colored spheres represent DMN nodes and blue colored spheres represent TPN nodes. Size of spheres (nodes) is proportional to the frequency of involvement of a given node in any significant association. The thickness of sticks (edges) is proportional to the frequency of involvement of a given node-node pair in any significant association. Results for GLM and ANCOVA models adjusted for clinical information are shown in S4E Fig.
Fig 5.
Sternberg’s working memory task.
A) Correct answers and reaction time (mean and standard error) as a function of memory load and its correlation with WMI derived from cognitive batteries. B) Group-level fMRI activation (controls and patients) during retention phase. Statistical map is shown referenced to the seizure onset hemisphere (I: ipsilateral; C: contralateral). C) Significant between-group voxelwise differences on fMRI activation during retention phase. Asterisks indicate pFDR < 0.05.
Fig 6.
A: Correlations between BOLD activity and cognitive scores. A scatter plot is shown for illustrative purposes for visual working memory index. B: ANCOVAs showed positive correlations only in TLE patients, and not in controls for three different memory indices. Analysis was restricted to regions involved in Sternberg’s task, and shown corrected for family-wise errors. Results of voxel-wise GLM and ANCOVA for the rest of cognitive scores are shown in S5 Fig. Color bars in A and B indicate the number of cognitive indexes that show significant correlations. C: Voxelwise analyses of seed-based RSFC for ROIs derived from the retention stage of Sternberg’s memory task. Top row shows connectivity of the whole brain RSFC of seed (red-outline diamonds). Bottom row shows group-wise differences of the corresponding connectivity maps.