Dynamic fMRI networks of human emotion
- Department of Psychology, Universidad de la Laguna, Spain
- Institute of Biomedical Technologies, Universidad de La Laguna, Spain
- Institute of Neurosciences, Universidad de la Laguna, Spain
- Department of Child and Adolescent Psychiatry, Institute of Psychiatry and Mental Health, Hospital General Universitario Gregorio Marañón, Spain
- Ciber del Área de Salud Mental, Instituto de Investigación Sanitaria Gregorio Marañón, Spain
- Department of Psychiatry, UMCU Brain Center, University Medical Center Utrecht, Netherlands
- Clinical Translational Neuroscience Laboratory, Department of Psychiatry and Human Behavior, University of California, Irvine, United States
- Center for the Neurobiology of Learning and Memory, University of California, Irvine, United States
Figures
Figure 1
Overview of quality control of structural data with Euler number (A), functional MRI (fMRI) head motion scores (B), analysis of name agreement using binomial GLM (logit link) with condition as a fixed effect and Bonferroni-adjusted pairwise contrasts.
Error bars reflect standard deviations for the scores of 22 participants per condition, (C) and reaction times using mixed effect model with condition as a fixed effect and a subject random intercept and Bonferroni-corrected pairwise contrasts Error bars reflect the standard errors of the model coefficients (D), and overview of the task (E).
© 1994, Paramount Pictures. Screenshots in panel E are taken from 'Forrest Gump' (1994). These are not covered by the CC-BY 4.0 license.
Figure 2
Slice-based functional MRI (fMRI) explanation (A, B), results (C), and subsequent processing steps (D).
In a hypothetical fMRI data acquisition protocol (A), three slices are acquired with TR = 2 s resulting in Slice-Acquisition Times (SAT) of 0, 0.7, 1.3 s, etc. The key point is that stimulus presentations are combined to extract the fMRI signal at high temporal resolution (Janssen et al., 2018). Three stimulus trials (T1.RT, T2.RT, T3.RT and red, yellow, green bars) whose onset coincides with each slice results in slices sampling the brain at all time points during a 6 s epoch relative to the onset of the stimulus. Thus, for a specific voxel on slice 2 (B), after these three stimulus trials, signal intensities at a temporal resolution equal to the ΔSAT (0.7 s) are acquired. Statistical models can extract the fMRI signal (B, gray dashed line) at this or at a higher powered binned, but lower temporal resolution. Results from group-level Slice-Based analysis of movie clips at 2 s resolution for the separate happy, fear, and sad trials (not all time points shown). Note that results reveal complex increases (hot colors) and decreases (cool colors) of signals across the whole brain during the 26 s epoch (C). The slice-based datasets for the happy, fear, and sad trials for each subject were then concatenated and entered into group spatial independent component analysis (ICA) for the detection of component maps and time courses (D).
Figure 3
Results from group spatial independent component analysis (ICA) with six dimensions.
Plotted are both the spatial maps and the component time courses averaged across subjects and conditions associated with each IC. Consideration of both spatial and temporal properties of these six ICs led to the identification of four ICs as signal (IC0, IC1, IC2, IC4) and two as noise (IC3, IC5). Note the noise components had spatial distributions associated with non-gray-matter regions (CSF, draining veins). Note also the networks are labeled and ordered as explained in the text. Bottom bar denotes the different stages of the behavioral task. The amplitude of the component time courses reflects the strength of engagement of each independent component over time, in z-scored units. Modulation of these time courses by emotions is investigated further below.
Figure 4
Basic regional functional connectivity in the four large-scale networks.
Note the scale differences between networks. Black bars denote the top 10 regions with the highest functional connectivity within each network. All results are Bonferroni-corrected.
Figure 5
Differences in functional connectivity between IC0 vs IC1 (A), IC1 vs IC0 (B), IC0 vs IC2 (C), IC2 vs IC0 (D), IC2 vs IC4 (E) and IC4 vs IC2 (F).
Note that here we focus on three main contrasts to highlight differences between IC0 and IC1, IC0 and IC2, and between IC2 and IC4.
Figure 6
Functional connectivity modulation of emotion conditions across networks IC1 “input” (A), IC0 “meaning” (B), IC2 “response (C), and IC4 “dmn” (D).
Note all large-scale networks involved in the task are modulated by emotions in all major lobes of the brain. H=happy, F=fear, S=sadness. All results are Bonferroni-corrected.
Figure 7
Explanation of curve-fitting procedure (A), goodness of fit results from Gaussian curve fitting (B), and graphical overview of how peak value (C), time to peak (D), and duration (E) differ between the four IC networks.
Note that goodness of fit values were relatively equal across IC networks, and that there were earlier time-to-peak values for IC1 compared to IC0. Error bars reflect standard error of mode coefficients, Different letters above bars indicate a significant difference in mixed effect models with component as a fixed effect and subject as a random intercept (p<0.05, Bonferroni-corrected) from 22 participants.
Figure 8
Results of the pairwise contrasts of the happy (H), sadness (S), and fear (F) conditions for the peak value (pv), time to peak (ttp), and duration (dur) across all four IC networks.
Only significant (p<0.05, Bonferroni-corrected) values are plotted.
Tables
Table 1
Order of the movie clips in the experiment.
| Clip# | Emotion | Description |
|---|---|---|
| 1 | Happy | Young Forrest meets Jenny |
| 2 | Fear | Forrest runs from bullies |
| 3 | Sad | Forrest says goodbye to Jenny |
| 4 | Fear | Forrest gets shouted at |
| 5 | Sad | Forrest is alone in bed |
| 6 | Fear | Forrest gets attacked |
| 7 | Happy | Forrest sees Jenny |
| 8 | Fear | Forrest sees a dangerous situation |
| 9 | Sad | Forrest sees a sad person |
| 10 | Sad | Forrest sees sick Jenny |
| 11 | Fear | Forrest sees Jenny's ghost |
| 12 | Happy | Forrest sees Jenny |
| 13 | Sad | Forrest knows what love is |
| 14 | Happy | Forrest is with Jenny |
| 15 | Happy | Forrest stays with Jenny |
Table 2
Overview of the ANOVA table from the statistical modeling of the regional functional connectivity data.
| Sum sq. | Mean sq. | NumDF | DenDF | F value | Pr(>F) | |
|---|---|---|---|---|---|---|
| Hemisphere | 0.17 | 0.17 | 1.00 | 21155.06 | 6.15 | 0.0131 |
| Head motion | 0.03 | 0.03 | 1.00 | 19.03 | 0.98 | 0.3346 |
| Euler | 0.07 | 0.07 | 1.00 | 18.97 | 2.53 | 0.1279 |
| IC_network | 26.37 | 8.79 | 3.00 | 21155.54 | 309.64 | <0.0001 |
| Emotion_condition | 0.09 | 0.04 | 2.00 | 21155.53 | 1.58 | 0.2060 |
| Region | 66.83 | 1.63 | 41.00 | 21155.13 | 57.43 | <0.0001 |
| IC:cond | 0.50 | 0.08 | 6.00 | 21155.46 | 2.95 | 0.0070 |
| IC:region | 226.84 | 1.84 | 123.00 | 21155.08 | 64.98 | <0.0001 |
| cond:region | 1.81 | 0.02 | 82.00 | 21155.07 | 0.78 | 0.9313 |
| IC:cond:region | 12.00 | 0.05 | 246.00 | 21155.05 | 1.72 | <0.0001 |
Table 3
Overview of the ANOVA table from the statistical modeling of the curve-fitting data.
| Variable | Sum sq. | Mean sq. | NumDF | DenDF | F value | Pr(>F) | |
|---|---|---|---|---|---|---|---|
| Peak value | RSE | 2583.06 | 2583.06 | 1.00 | 206.16 | 77.37 | <0.0001 |
| IC | 740.74 | 246.91 | 3.00 | 190.48 | 7.40 | 0.0001 | |
| cond | 1123.70 | 561.85 | 2.00 | 187.37 | 16.83 | <0.0001 | |
| IC:cond | 1058.63 | 176.44 | 6.00 | 186.05 | 5.28 | <0.0001 | |
| Time to peak | RSE | 1.72 | 1.72 | 1.00 | 205.18 | 1.67 | 0.1981 |
| IC | 3403.49 | 1134.50 | 3.00 | 189.38 | 1096.81 | <0.0001 | |
| cond | 26.45 | 13.22 | 2.00 | 187.35 | 12.78 | <0.0001 | |
| IC:cond | 14.83 | 2.47 | 6.00 | 186.55 | 2.39 | 0.0301 | |
| Duration | RSE | 9.43 | 9.43 | 1.00 | 160.08 | 3.25 | 0.0734 |
| IC | 1443.68 | 481.23 | 3.00 | 195.80 | 165.69 | <0.0001 | |
| cond | 24.82 | 12.41 | 2.00 | 191.79 | 4.27 | 0.0153 | |
| IC:cond | 46.73 | 7.79 | 6.00 | 189.42 | 2.68 | 0.0160 |
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Dynamic fMRI networks of human emotion
eLife 14:RP106070.
https://doi.org/10.7554/eLife.106070.3
