Emotion self-regulation training in major depressive disorder using simultaneous real-time fMRI and EEG neurofeedback

Figures

• 

  Fig. 1. Experimental paradigm for emotion self-regulation training using simultaneous real-time fMRI and EEG neurofeedback (rtfMRI-EEG-nf). A) Real-time GUI display screen for Happy Memories conditions with rtfMRI-EEG-nf. The four neurofeedback signals are displayed on the screen as four variable-height bars. The two EEG-nf signals on the left are based, respectively, on changes in frontal alpha EEG asymmetry (FAA, magenta) and frontal high-beta EEG asymmetry (FBA, purple). The two rtfMRI-nf signals on the right are based, respectively, on fMRI activities of the left amygdala (LA, red) and the left rostral anterior cingulate cortex (L rACC, orange). The bar heights are updated every 2 s. The black horizontal bar in the middle of the screen specifies a target level. It is raised from run to run. B) Experimental protocol consisted of six runs, each lasting 8 min 46 s. It included a Rest run, four rtfMRI-EEG-nf runs – Practice, Run 1, Run 2, Run 3, and a Transfer run without nf. The names of the five task runs are abbreviated in the text and figures as PR, R1, R2, R3, and TR, respectively. The task runs consisted of 40-s long blocks of Rest, Happy Memories, and Count conditions. The…

• 

  Fig. 2. Target regions of interest (ROIs) and EEG channels used to provide the rtfMRI-EEG-nf. The ROIs were defined anatomically in the co-planar stereotaxic atlas of the human brain by Talairach and Tournoux, and transformed to each participant’s individual fMRI image space. A) Spherical 14-mm-diameter target ROI in the left amygdala (LA) region, projected onto the standard TT_N27 anatomical template in the Talairach space. B) Spherical 14-mm-diatemer target ROI in the left rostral anterior cingulate cortex (L rACC) region. C) Frontal EEG channels F3 (left) and F4 (right) used to provide the EEG-nf based on frontal EEG asymmetries in the alpha and high-beta EEG bands. Channel FCz was employed as reference. D) Custom modification of a standard MR-compatible 32-channel EEG cap to improve quality of the EEG-nf during fMRI. The modified cap includes four wire contours with resistors for acquisition of reference artifact waveforms.

• 

  Fig. 3. Real-time EEG-fMRI artifact correction procedure and its performance. A) The real-time procedure included three consecutive steps implemented in BrainVision RecView software: average artifact subtraction (AAS) of MR artifacts; regression of reference artifact waveforms approximating cardioballistic (CB) and random-motion artifacts (channels F3 and F4 only); AAS of cardioballistic artifacts. B) Illustration of performance of the regression procedure across an entire experimental run (left) and for a short time interval in the middle of that run (right). C) Performance of the regression procedure for channels F3 and F4 across all experimental (task) runs for all participants. Each histogram bar represents a probability of observing a change in signal variance within a given 1 dB-wide interval after the regression across one experimental run. D) Illustration of topographic properties of residual CB and random-motion artifacts after the real-time procedure (A). Topographies of independent components modeling such artifacts for a typical experimental run are shown. Note additional suppression of the artifacts for channels F3 and F4 compared to the surrounding EEG channels.

• 

  Fig. 4. Main performance characteristics of the rtfMRI-EEG-nf training for the experimental group (EG). Each bar represents a group mean of average individual results for a given run. The error bars are standard errors of the mean (sem). The experimental runs and condition blocks are depicted schematically in Fig. 1B. The NF at the bottom of each figure refers to group statistics (p-value from a t-test and effect size d, both relative to zero) for the individual results averaged across the four nf runs (PR, R1, R2, R3). A) Average changes in frontal alpha EEG asymmetry (FAA) between the Happy Memories and Rest conditions (H vs R). FAA = ln(P(F4)) − ln(P(F3)), where P is EEG power in the individual upper alpha band. B) Average changes in frontal high-beta EEG asymmetry (FBA) between the same conditions. FBA = ln(P(F3)) − ln(P(F4)), where P is EEG power in the high-beta band. C) Average fMRI percent signal changes for the LA target ROI (Fig. 2A) for the Happy Memories conditions with respect to the Rest baseline (H vs R). D) Average changes in fMRI functional connectivity between the LA and L rACC target ROI (Fig. 2B) during the Happy Memories conditions compared to the Rest conditions (H…

• 

  Fig. 5. Main performance characteristics for the control group (CG). Notations are the same as in Fig. 4.

• 

  Fig. 6. Average fMRI percent signal changes for the L rACC target ROI for the Happy Memories vs Rest condition contrast (H vs R) and for the Happy Memories vs Count condition contrast (H vs C). The black straight segments are linear fits across the four nf runs (PR, R1, R2, R3). The NF at the bottom of each figure refers to group statistics (p-value from a t-test and effect size d, both relative to zero) for the individual activity levels, corresponding to the H vs C contrast, averaged across the four nf runs. A) Results for the experimental group (EG). B) Results for the control group (CG).

• 

  Fig. 7. Correlations between changes in frontal alpha EEG asymmetry (FAA) during the rtfMRI-EEG-nf training and individual psychological measures. The individual FAA changes between the Rest and Happy Memories with rtfMRI-EEG-nf conditions (H vs R) were averaged across the four nf runs (PR, R1, R2, R3). The results are for the experimental group (EG), with each data point corresponding to one participant. A) Correlations between the FAA changes and severities of depression and anhedonia, assessed before the session. B) Correlations between the FAA changes and changes in state depression and total mood disturbance. The state measures were assessed both before and after the session, and their changes (after vs before) are included in the figures. Acronyms: MADRS – Montgomery-Asberg Depression Rating Scale, SHAPS – Snaith-Hamilton Pleasure Scale, POMS – Profile of Mood States.

• 

  Fig. 8. Illustration of BOLD laterality for middle frontal gyrus (MidFG) regions during the rtfMRI-EEG-nf training for the experimental group (EG) and its associations with individual psychological measures. A) The left and right MidFG ROIs as seen on the cortical surface. The ROIs were defined as described in the text. B) Average fMRI activity levels (percent signal changes) for the left and right MidFG ROIs, corresponding to the Happy Memories vs Rest condition contrast (H vs R) and to the Happy Memories vs Count contrast (H vs C). The individual fMRI activity levels were averaged across the four nf runs (PR, R1, R2, R3). C) Left: Correlation between the average individual MidFG laterality, i.e. the difference in fMRI activities between the left and right MidFG ROIs, for the Happy Memories vs Count contrast and after-vs-before changes in state depression. Right: Correlation between the average individual MidFG laterality and changes in total mood disturbance. POMS – Profile of Mood States.

• 

  Fig. 9. Correlations between changes in fMRI functional connectivity of the left amygdala with the left rACC during the rtfMRI-EEG-nf training and individual psychological measures. The fMRI connectivity changes for the target ROIs between the Rest and Happy Memories with rtfMRI-EEG-nf conditions (H vs R, psychophysiological interaction effect) are reported. The results are for the experimental group (EG), with each data point corresponding to one participant (however, n = 13 for the BAS). The individual results were averaged across the four nf runs (PR, R1, R2, R3). MADRS – Montgomery-Asberg Depression Rating Scale, BAS – Behavioral Activation System scale.

• 

  Fig. 10. Changes in temporal correlations between frontal EEG asymmetries and BOLD fMRI activity of the left amygdala during the rtfMRI-EEG-nf training. The correlation changes between Happy Memories and Count conditions (H vs C) were evaluated in EEG-fMRI psychophysiological interaction (PPI) analyses, based on time courses of frontal alpha EEG asymmetry (FAA) and frontal high-beta EEG asymmetry (FBA). Each bar represents a group mean of the individual PPI interaction values for a given run, averaged within the LA ROI. The error bars are standard errors of the mean (sem). The NF at the bottom of each figure refers to group statistics (p-value from a t-test and effect size d, both relative to zero) for the individual results averaged across the four nf runs (PR, R1, R2, R3). A) Average values of the FAA-based PPI interaction effect for the LA ROI for the experimental group (EG). B) Average values of the FBA-based PPI interaction effect for the LA ROI for the EG. C) Average values of the FAA-based PPI interaction effect for the control group (CG). D) Average values of the FBA-based PPI interaction effect for the CG.

• 

  Fig. 11. Enhancement in temporal correlation between frontal alpha EEG asymmetry (FAA) and BOLD activity during the rtfMRI-EEG-nf training. Statistical maps of the FAA-based PPI interaction effect for the Happy Memories vs Count condition contrast (H vs C) are shown for the experimental group (EG). The maps are voxel-wise FDR corrected and projected onto the standard TT_N27 anatomical template, with 3 mm separation between axial slices. The number adjacent to each slice indicates the z coordinate in mm. The left hemisphere (L) is to the reader’s right. Peak t-statistics values for the FAA-based PPI interaction effect and the corresponding locations are specified in Table 2.

• 

  Fig. 12. Comparison of the FAA- and FBA-based PPI interaction effects for the left amygdala region. A) FAA-based PPI interaction effects from Fig. 11. B) FBA-based PPI interaction effects from Supplementary Fig. S8. The green crosshairs (x = −21 mm, z = −16 mm) correspond to the center of the LA target ROI. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)