Research ArticleNEUROSCIENCE

The physiological effects of noninvasive brain stimulation fundamentally differ across the human cortex

See allHide authors and affiliations

Science Advances  31 Jan 2020:
Vol. 6, no. 5, eaay2739
DOI: 10.1126/sciadv.aay2739
  • Fig. 1 Study design

    (A) Each participant underwent three counterbalanced TMS-fMRI sessions on three different days. During each session, one target region (FRO, OCC, CTR; colored circles) was stimulated for 20 min with rTMS (1 Hz), and we acquired resting-state fMRI data during preTMS and postTMS. Colored overlays on the brain surface illustrate cortical template networks (31). (B) For each subject, we derived individual target spots (green spheres) within target areas from a functional network analysis of the preTMS fMRI data. (C) Brain slices with statistical parametric maps (PFWE < 0.05, corrected at cluster level, one-sample t tests) of the group average functional connectivity of each target region during preTMS calculated from the individual TMS targets. Color bars, t values.

  • Fig. 2 Opposite effect of TMS on brain functional connectivity

    (A) Statistical parametric maps of significant changes in whole-brain functional connectivity after OCC- (top), FRO- (bottom), and CTR- (right) TMS (PFWE < 0.05, corrected at cluster level, voxel-wise repeated-measures ANOVAs). Color bar (yellow-blue), t values. OCC-TMS increased not only the functional connectivity of the OCC target (direct) but also the functional connectivity of the nonstimulated, FRO target (indirect). Conversely, FRO-TMS decreased functional connectivity to widespread cortical areas. Again, FRO-TMS had an impact not only on the functional connectivity of the stimulated, FRO target (direct) but also on the nonstimulated, OCC target (indirect). Neither of the stimulation protocols changed functional connectivity of the CTR target nor did CTR-TMS. Horizontal bar plots (violet) indicate the laterality effect of stimulation, i.e., the ratio of significant voxels in left and/or right hemispheres. (B) Summary of whole-brain functional connectivity changes as found in (A) overlaid on the cortical surface of a standard brain. Horizontal color bars indicate the ratio of significant voxels in each of the template networks as illustrated by colored outlines on the cortical surface.

  • Fig. 3 Effect of stimulation on global functional integration

    (A) Overview of consensus modularity analysis on the individual and group levels resulting in condition-specific parameters of local (z) and global (h) functional integration for each brain node. (B) Group average coclassification matrix (left) and its corresponding force-directed topological representation of preTMS (middle). Nodes with higher coclassification values are located closer to each other, and node size represents h. Node color indicates the modular affiliation, and abbreviations indicate assignment to template networks. Yellow circles indicate TMS targets, and insets highlight only those nodes with direct functional connectivity to the target nodes. (C and D) Group average coclassification matrices and topological representations after (C) OCC-TMS and (D) FRO-TMS with target nodes (yellow circles). Scatterplots of local (z) versus global (h) integration before (gray) and after (violet) stimulation. Note that only OCC-TMS increased global integration of the OCC target and of the entire graph. **P = 0.004, Wilcoxcon signed-rank test. Bar plot in (D) illustrates h during preTMS for all voxels that showed changes in pairwise functional connectivity. Baseline h was higher for voxels with spreading effects after FRO-TMS compared to OCC-TMS. ***P = 0.00004, Wilcoxon signed-rank test. n.s., not significant.

  • Fig. 4 Brain functional integration is a predictive marker for spreading effects of TMS

    (A) Individual classification between OCC- and FRO-TMS based on h values of a whole-brain parcellation yielded an overall prediction accuracy of 67%. (Left) Confusion matrix with the prediction accuracies for every class. (Middle) receiver operating characteristic (ROC) curve and discrimination probability [area under the curve (AUC) of 0.68]. (Right) Model (green line) significantly (P = 0.027, permutation testing) deviated from the null distribution (black dashed line) after permutation testing. (B) (Left) Spatial distribution of h across the entire cortex and (right) averaged across template networks.

  • Fig. 5 No effect of TMS on local brain activity

    Bar plots indicate group average (A) amplitude of ALFF, (B) ReHo of local functional connectivity, and (C) SD of the fMRI signal averaged across all voxels of each target region. PostTMS values (colored bars) after direct stimulation did not significantly differ from any other session (PFWE > 0.05 corrected at cluster level, voxel-wise ANOVA for repeated measures). Error bars represent the 95% CI of variance across subjects.

  • Fig. 6 Biophysical modeling of the stimulation effect on directional signaling

    (A) We modeled directional signaling between regions with functional connectivity changes using spectral DCM. The model incorporates local self-inhibition (red), excitatory feedforward (green), and inhibitory feedback (yellow) signaling. Connectivity schemas illustrate directional signaling along an anterior-posterior (left-to-right) axis of the human cortex. During preTMS, the model indicated specific feedforward and feedback, as well as balanced (gray) connectivity pathways. (B) OCC-TMS strengthened feedforward signaling of occipital (R1 and R2) and parietal (R6) regions, while other connections remained in place (gray arrows). (C) FRO-TMS uncoupled 11 of 16 directional pathways (dotted lines) equally affecting feedforward and feedback connections. Violet numbers indicate changes in parameter estimates after PEB procedures with a posterior probability of >95%.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/6/5/eaay2739/DC1

    Table S1. Participants demographic information and individual TMS target coordinates in MNI space.

    Table S2. Statistical significant changes in whole-brain functional connectivity after OCC-TMS as presented in Fig. 2 and fig. S2A.

    Table S3. Statistical significant changes in whole-brain functional connectivity after FRO-TMS as presented in Fig. 2 and fig. S2B.

    Fig. S1. Image quality assessment.

    Fig. S2. Cross-sectional representation of Fig. 2A.

    Fig. S3. Statistical significance of the modularity results.

    Fig. S4. Classification results using linear support vector machine (SVM).

    Fig. S5. Effect of the parameter τ on the global functional integration results.

  • Supplementary Materials

    This PDF file includes:

    • Table S1. Participants demographic information and individual TMS target coordinates in MNI space.
    • Table S2. Statistical significant changes in whole-brain functional connectivity after OCC-TMS as presented in Fig. 2 and fig. S2A.
    • Table S3. Statistical significant changes in whole-brain functional connectivity after FRO-TMS as presented in Fig. 2 and fig. S2B.
    • Fig. S1. Image quality assessment.
    • Fig. S2. Cross-sectional representation of Fig. 2A.
    • Fig. S3. Statistical significance of the modularity results.
    • Fig. S4. Classification results using linear support vector machine (SVM).
    • Fig. S5. Effect of the parameter τ on the global functional integration results.

    Download PDF

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article