Effects of simultaneous real-time fMRI and EEG neurofeedback in major depressive disorder evaluated with brain electromagnetic tomography
1 Effects of simultaneous real-time fMRI and EEG neurofeedback in major depressive disorder evaluated with brain electromagnetic tomography
Vadim Zotev and Jerzy Bodurka Laureate Institute for Brain Research, Tulsa, OK, USA; Stephenson School of Biomedical Engineering, University of Oklahoma, Norman, OK, USA
Abstract:
Recently, we reported an emotion self-regulation study (Zotev et al., 2019), in which patients with major depressive disorder (MDD) used simultaneous real-time fMRI and EEG neurofeedback (rtfMRI-EEG-nf) to upregulate two fMRI and two EEG activity measures, relevant to MDD. The target measures included fMRI activities of the left amygdala and left rostral anterior cingulate cortex, and frontal EEG asymmetries in the alpha band (FAA) and high-beta band (FBA). Here we apply the exact low resolution brain electromagnetic tomography (eLORETA) to investigate EEG source activities during the rtfMRI-EEG-nf procedure. The analyses reveal significant changes in hemispheric lateralities of upper alpha and high-beta current source densities in the prefrontal regions, consistent with upregulation of the FAA and FBA during the rtfMRI-EEG-nf task. Similar laterality changes are observed for current source densities in the amygdala. Prefrontal upper alpha current density changes show significant negative correlations with anhedonia severity. Comparisons with results of previous LORETA studies suggest that the rtfMRI-EEG-nf training is beneficial to MDD patients, and has the ability to correct functional deficiencies associated with anhedonia and comorbid anxiety. Our findings confirm the poten-tial of the rtfMRI-EEG-nf for treatment of major depression.
Keywords: depression, neurofeedback, EEG-fMRI, amygdala, frontal EEG asymmetry, alpha band, high-beta band, EEG source analysis, LORETA
1. Introduction
Simultaneous real-time fMRI and EEG neurofeedback (rtfMRI-EEG-nf) is an advanced neuromodulation tech-nique that combines real-time fMRI neurofeedback (rtfMRI-nf) and EEG neurofeedback (EEG-nf) to enable simultaneous regulation of both hemodynamic (BOLD fMRI) and electrophysiological (EEG) brain activities (Zotev et al., 2014). The main promise of rtfMRI-EEG-nf for treatment of neuropsychiatric disorders is its ability to significantly alter both fMRI and EEG activity measures relevant to a specific disorder. However, implementation of rtfMRI-EEG-nf is technically challenging, and its mechanisms of action remain insufficiently investigated (Lioi et al., 2020; Mano et al., 2017; Perronnet et al., 2017; Zotev et al., 2014). Recently, we completed an emotion self-regulation study, in which patients with major depressive disorder (MDD) used rtfMRI-EEG-nf to simultaneously upregulate two EEG-nf and two rtfMRI-nf target measures, while inducing happy emotion (Zotev et al., 2019). The target measures for EEG-nf included right-vs-left frontal alpha EEG asymmetry (FAA) and left-vs-right frontal high-beta EEG asymmetry (FBA) for EEG channels F3 and F4. The rtfMRI-nf target measures included fMRI activity of the left amygdala (LA) and fMRI activity of the left rostral anterior cingulate cortex (L rACC). We selected these four target measures, because each of them is relevant to MDD (Zotev et al., 2019). During the rtfMRI-EEG-nf procedure, the MDD patients were able to significantly increase the FAA, the FBA, and the LA fMRI activation. They also achieved significant enhancement in fMRI functional con-nectivity between the LA and the L rACC through simultaneous upregulation of these regions’ fMRI activi-ties. After the rtfMRI-EEG-nf session, the MDD participants showed significant mood improvements, in-cluding reductions in state depression, anxiety, confusion, and total mood disturbance, and increase in state happi-ness (Zotev et al., 2019). These findings demonstrated the potential of the rtfMRI-EEG-nf for treatment of MDD. The purpose of the present follow-up study is to evalu-ate EEG source activity during the rtfMRI-EEG-nf procedure. We employed the exact low resolution brain electromagnetic tomography (eLORETA) (Pascual-Marqui, 2007). The eLORETA method provides a distrib-uted source solution to the electromagnetic inverse problem in neuroimaging, which enables exact, zero-error localization of point current sources in the presence of measurement and structured biological noise (Pascual- ___________________
Corresponding authors. E-mail: [email protected]; [email protected] voxels, and magnitude of current source density is computed for each voxel. It has been shown that the eLORETA method is superior to other EEG/MEG source analysis techniques in terms of localization accuracy and reliability of functional connectivity estimates (Pascual-Marqui et al., 2018). Re-sults of LORETA source localization are generally consistent with activation patterns revealed by simultane-ous fMRI (e.g. Mulert et al., 2004, 2005). We conducted the present study to investigate the fol-lowing. First , we wished to examine to what extent the modulation of FAA and FBA by means of EEG-nf during the rtfMRI-EEG-nf procedure is reflected in hemispheric lateralities of cortical EEG sources.
Second , we aimed to better understand potential therapeutic effects of the rtfMRI-EEG-nf by comparing our eLORETA results with findings from previous LORETA studies in MDD patients and healthy individuals. All analyses conducted in our work are exploratory, because the EEG-nf procedure tar-geted scalp EEG activity, while the eLORETA method estimates sources of neuronal activity.
2. Methods
The study, described in detail in Zotev et al., 2019, was conducted at the Laureate Institute for Brain Research. It was approved by the Western Institutional Review Board (IRB). Twenty four unmedicated MDD patients completed one rtfMRI-EEG-nf training session. The participants were right-handed and met the criteria for MDD laid out in the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV, American Psychiatric Association, 2000). They underwent a psychological assessment, that included the Montgomery-Asberg Depression Rating Scale (MADRS, Montgomery and Asberg, 1979), the Snaith-Hamilton Pleasure Scale (SHAPS, Snaith et al., 1995), the State-Trait Anxiety Inventory (STAI, Spielberger et al., 1970), and other instruments. During the session, participants in the experimental group (EG, n =16) were provided with the rtfMRI-EEG-nf, based on their real-time brain activity measures. Participants in the control group (CG, n =8) were provided, without their knowledge, with computer-generated sham feedback signals, unrelated to any brain activity (Zotev et al., 2018a, 2019). The rtfMRI-EEG-nf was implemented using the custom real-time control system for integration of simultaneously acquired EEG and fMRI data streams (Zotev et al., 2014). The system utilizes real-time features of AFNI (Cox, 1996) and BrainVision RecView software (Brain Products, GmbH). The multimodal graphical user interface (mGUI) was used to display four variable-height bars (Fig. 1A). The bar heights, updated every 2 s, represented the four neurofeedback signals as follows. First , the magenta EEG-nf bar on the left represented a change in relative alpha EEG asymmetry for channels F3 and F4 (Fig. 2A, see e.g. Allen et al., 2001). The relative right-vs-left alpha asymmetry was defined as A = ( P (F4) – P (F3)) / ( P (F4) + P (F3)), where P is EEG power in the alpha frequency band [7.5-12.5] Hz. In offline EEG analyses, normalized frontal alpha EEG asymmetry, commonly defined as FAA = ln( P (F4)) – ln( P (F3)), was employed. Second , the purple EEG-nf bar on the left represented a change in relative left-vs-right high-beta EEG asymmetry, defined as B = ( P (F3) – P (F4)) / ( P (F3) + P (F4)), where P is EEG power in the high-beta (beta 3) frequency band [21-30] Hz. Normalized FBA = ln( P (F3)) – ln( P (F4)) was used in offline EEG analyses. Third , the red rtfMRI-nf bar on the right represented BOLD fMRI activity of the left amygdala (LA) target ROI (Fig. 2B). This spherical ROI with R =7 mm was centered at (−21, −5, −16) locus (Zotev et al., 2011) in the Talairach space (Talairach and Tournoux, 1988). Fourth , the orange rtfMRI-nf bar on the right represented fMRI activity of the left rACC target ROI (Fig. 2C). This ROI, also with R =7 mm, was centered at (−3, 34, 5) locus (Zotev et al., 2013). For a detailed explanation of selection of these four target measures, see Zotev et al., 2019. All study participants followed the experimental protocol depicted in Fig. 1B. It included six EEG-fMRI runs, each lasting 8 min 46 s. During the Rest run, the participants were asked to relax and rest while looking at a fixation cross. The five task runs – the Practice run, Run 1, Run 2, Run 3, and the Transfer run – consisted of alternating 40-s-long blocks of Rest, Happy Memories, and Count conditions (Fig. 1B). For the Rest conditions, the participants were instructed to relax and rest looking at a fixation cross. For the Happy Memories with rtfMRI-EEG-nf conditions, the participants were asked to induce happy emotion by retrieving happy autobiographical memories, and simultaneously raise the levels of all four neurofeedback bars (Fig. 1A). The bar heights represented changes in the target measures for the current Happy Memories condition (fMRI volume and EEG segment) relative to the baselines corresponding to the preceding Rest condition block. For the Count conditions, the participants were instructed to mentally count back from 300 by subtracting a given integer. The integers were 3, 4, 6, 7, and 9 for the five task runs, respectively. No bars were displayed during the Happy Memories conditions in the Transfer run, and during the Rest and Count conditions in all runs. For details of the experimental protocol and instructions given to the participants, see Zotev et al., 2019. The experiments were performed using the General Electric Discovery MR750 3T MRI scanner with a standard 8-channel head coil. A single-shot gradient echo EPI sequence with 34 axial slices, TR/TE =2000/30 ms, and SENSE R =2 was employed for fMRI. It provided whole-brain fMRI images with 1.875×1.875×2.9 mm voxels. A T1-weighted 3D MPRAGE sequence yielded high-resolution anatomical brain images with 0.94×0.94× 1.2 mm voxels. EEG recordings were conducted simultaneously with fMRI using a 32-channel MR-compatible EEG system from Brain Products, GmbH. MR-compatible EEG caps (EASYCAP, GmbH) were custom modified to enable ac-quisition of four reference artifact waveforms for improved real-time EEG-fMRI artifact correction (Zotev et al., 2019). Because of this modification, the number of EEG channels was 27. Raw EEG data were acquired using BrainVision Recorder software (Brain Products, GmbH) with 0.2 ms temporal and 0.1 µV measurement resolution in [0.016-250] Hz frequency range with FCz reference. BrainVision RecView software was used to perform real-time EEG-fMRI artifact correction and export the correct-ed EEG data to the mGUI software for further processing. Technical details of the rtfMRI-EEG-nf implementation, real-time artifact correction, and real-time data processing were described previously (Zotev et al., 2014, 2019). Pre-processing of raw EEG data acquired during fMRI was performed using BrainVision Analyzer 2.1 software (Brain Products, GmbH), as described in Zotev et al., 2019. Briefly, it involved average artifact subtraction (AAS) for MR and cardioballistic (CB) artifacts and identification of bad intervals, showing intense random-motion artifacts. Independent component analysis (ICA) was applied to identify residual EEG-fMRI artifacts and various EEG artifacts, and to separate them from neuronal activity. The pre-processed EEG data included time courses of 27 EEG channels with single-electrode (FCz) reference and 250 S/s sampling (4 ms interval). The upper alpha EEG frequency band was defined individually for each participant as [IAF, IAF+2] Hz, where IAF is the individual alpha peak frequency. The IAF was determined by inspection of average EEG spectra for the occipital and parietal EEG channels across the Rest condition blocks in the four nf runs (Fig. 1B).
Figure 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. B) Experimental protocol consists of six runs, each lasting 8 min 46 s. It includes a Rest run, four rtfMRI-EEG-nf runs – Practice, Run 1, Run 2, Run 3, and a Transfer run without nf. The task runs consist of 40-s long blocks of Rest, Happy Memories, and Count conditions. In preparation for the source analysis, the EEG data were transformed to common average reference, and the original reference was restored as the regular FCz channel, yielding 28 EEG channels. The data were lowpass filtered at 56 Hz (24 dB/octave), and downsampled to 125 S/s sampling rate (8 ms interval). The data for each task run were then split into three datasets, corresponding to the Happy Memories, Count, and Rest conditions (Fig. 1B), with exclusion of the bad intervals, and segmented into 4096-ms-long (512 time points) epochs.
The LORETA-KEY software was used to conduct the eLORETA source analyses. The software employs a realistic head model (Fuchs et al., 2002) and the probabilistic Montreal Neurological Institute (MNI) brain atlas (Mazziotta et al., 2001). The eLORETA transformation matrix was computed using MNI coordinates of the 28 EEG electrodes arranged according to the international 10-20 system (Fp1, Fp2, F3, F4, F7, F8, FC5, FC6, C3, C4, CP1, CP2, CP5, CP6, T7, T8, P3, P4, P7, P8, O1, O2, Fz, FCz, Cz, Pz, POz, Oz). To enable EEG source analysis in the frequency domain, we computed an EEG cross spectrum using each EEG channel’s data for all the epochs corresponding to a given condition (Happy Memories, Count, Rest) in a given run. The cross spectra were calculated separately for the individual upper alpha band [IAF, IAF+2] Hz and for the high-beta band [21-30] Hz. The eLORETA transformation was then applied to the cross spectra, yielding current density magnitudes j for 5×5×5 mm voxels ( n =6239) as functions of frequency in the selected frequency bands. The eLORETA results were normalized as ln( j ) without any additional scaling. Changes in the normalized current density values between the Happy Memories and Rest conditions were computed for each run for each participant. Group statistical analyses were performed on changes in the normalized current density ln( j ) between the two conditions. To evaluate task-specific activity, a single t -test relative to zero was applied to average individual current density changes in a given frequency band (upper alpha, high-beta). Correction for multiple comparisons across the eLORETA solution space was based on the randomization SnPM procedure (Nichols and Holmes, 2001), implemented in the LORETA-KEY software. The procedure yielded critical t -thresholds and corrected p -values. To evaluate correlations between localized activity and a psychological measure, a regression procedure was performed for average current density changes in a given frequency band and the psychological measure as an independent variable. The randomization SnPM procedure in this case yielded critical r -thresholds and corrected p -values. To evaluate hemispheric laterality of eLORETA results, we defined, a priori, pairs of ROIs in the corresponding brain regions on the left and on the right. Because the EEG-nf signals were based on FAA and FBA for channels F3 and F4 (Fig. 2A), we chose ROIs in the middle frontal gyrus (MidFG) and superior frontal gyrus (SFG), located approximately underneath these two electrodes. Each ROI center was selected as a center of mass of an anatomical overlap between a gyrus and a Brodmann area (BA) in the same hemisphere, as defined in the Talairach-Tournoux Daemon (Lancaster et al., 2000). The centers of mass were determined using the 3dclust AFNI program with -mni option. The resulting ROI centers are specified in Table 1. The ROIs were then defined as collections of all eLORETA voxels within R =10 mm distance from the selected centers. In addition to the MidFG and SFG ROIs, we defined ROIs in the lateral orbitofrontal cortex, i.e. inferior frontal gyrus (IFG), BA 47, which plays an important role in emotion regulation, and in the amygdala (Table 1). The left amygdala ROI was centered at (−21, −4, −19) in MNI coordinates, corresponding to the center of the LA target Figure 2.
EEG channels and target regions of interest (ROIs) used to provide the rtfMRI-EEG-nf. A) Frontal EEG channels F3 (left) and F4 (right) used to generate EEG-nf signals based on frontal EEG asymmetries in the alpha and high-beta EEG bands. Channel FCz is the reference. B) Spherical target ROI for rtfMRI-nf in the left amygdala (LA) region. C) Spherical rtfMRI-nf target ROI in the left rostral anterior cingulate cortex (L rACC) region. The two ROIs, defined in the Talairach space, are transformed to each participants’s individual fMRI image space. ROI (Fig. 2B). The right amygdala ROI was centered at (21, −4, −19). Each ROI included approximately 15-20 voxels. Importantly, all the ROIs were defined independently of any results in the present study.
3. Results
Figure 3 exhibits whole-brain statistical maps for changes in the normalized current density ln( j ) in the upper alpha band during the Happy Memories with rtfMRI-EEG-nf conditions relative to the Rest conditions (H vs R) for the experimental group (EG). The upper alpha band was defined individually for each participant, as described above (Sec. 2.2). The individual-subject results were averaged across the four rtfMRI-EEG-nf runs (Practice, Run 1, Run 2, Run 3). The statistical results for voxels with | t |>2.78, two-tailed, are significant with p <0.05, corrected for multiple comparisons across the eLORETA space. The scale exponent for the color scale in Fig. 3 is 2.0. The most significant statistical results are reported in Table 2. For each anatomical area, the largest t -statistics value and the number of voxels in that area with statistics exceeding the critical threshold are specified. Figure 4 illustrates hemispheric laterality of the normalized upper alpha current density changes, exhibited in Fig. 3. Each sub-figure shows group mean (±sem) values for individual current density changes averaged within the corresponding ROIs on the left and on the right, defined a priori as described above (Sec. 2.4, Table 1). The results of exploratory paired t -tests (Fig. 4) show that the reductions in upper alpha current density during the rtfMRI-EEG-nf task were significantly stronger, with medium effect sizes, for the prefrontal ROIs on the left, than for their counterparts on the right (MidFG, BA8: t (15)=2.30, p <0.036, d =0.57; MidFG, BA9: t (15)=2.40, p <0.030, d =0.60; SFG, BA 9: t (15)=3.08, p <7.7e−3, d =0.77). Significant positive change in the right-vs-left laterality of current densities is also observed for the amygdala ROIs (Amygdala: t (15)=2.27, p <0.038, d =0.57). For the lateral orbitofrontal cortex ROIs, the laterality change trends toward significance (IFG, BA 47: t (15)=2.06, p <0.057, d =0.51). Significant across-subjects correlation is found between the amygdala laterality changes and the MidFG, BA 9 laterality changes ( r =0.55, p <0.034), as well as the IFG, BA 47 laterality changes ( r =0.79, p <5.0e−4). For the control group (CG), the laterality effects, corresponding to those described above for the EG (Fig. 4), were non-significant with small Figure 3. eLORETA statistical maps for changes in normalized upper alpha current source density ln( j ) between the Happy Memories with rtfMRI-EEG-nf and Rest conditions (H vs R) for the experimental group (EG). The maps are projected onto the MNI152 template. The arrows above the colorbar denote the critical threshold from the randomization SnPM procedure. Statistics are summarized in Table 2. effects sizes. The eLORETA results in the present study showed the most significant associations with the MDD patients’ anhedonia severity ratings (SHAPS). Correlations with other psychological measures were less significant. Note that there were significant correlations among the depression severity (MADRS), anhedonia severity, and trait anxiety (STAI-t) ratings for the EG participants (MADRS vs SHAPS: r =0.67, p <4.5e−3; SHAPS vs STAI-t: r =0.75, p <8.7e−4; MADRS vs STAI-t: r =0.78, p <3.5e−4). Figure 5 reports statistical maps for correlations between the changes in the normalized upper alpha current density for the Happy Memories with rtfMRI-EEG-nf conditions relative to the Rest conditions (H vs R) and the EG participants’ SHAPS ratings. The voxels with | r |>0.66 correspond to p <0.05, corrected. Statistics for the results in Fig. 5 are included in Table 3. We also conducted exploratory analyses examining correlations between changes in right-vs-left laterality of upper alpha current densities for the a priori selected ROIs (Table 1) and the participants’ anhedonia severity ratings. Figure 4.
Laterality of the upper alpha current source density changes (Fig. 3) for corresponding ROIs on the left and on the right. The ROIs were selected a priori as described in the text. The statistics at the bottom of each figure ( p -value from a paired t -test and effect size d ) refer to comparison of the upper alpha current density changes on the right and on the left. MidFG – middle frontal gyrus, SFG – superior frontal gyrus. Figure 5. eLORETA statistical maps for correlations between the changes in normalized upper alpha current source density ln( j ) for the Happy Memories with rtfMRI-EEG-nf conditions relative to the Rest conditions (H vs R) and anhedonia severity (SHAPS) ratings for the experimental group (EG). The maps are projected onto the MNI152 template. The arrows above the colorbar denote the critical threshold from the randomization SnPM procedure. Statistics are summarized in Table 3. SHAPS – Snaith-Hamilton Pleasure Scale. We considered partial correlations, controlled for the average current density changes for the corresponding left and right ROIs, as well as for the patients’ trait anxiety ratings (STAI-t). The partial correlation trended toward significance for the MidFG, BA 8 (Laterality vs SHAPS: r (12)=0.47, p <0.088), and was significant for the IFG, BA 47 (Laterality vs SHAPS: r (12)=0.69, p <5.9e−3). For the amygdala ROIs, the partial correlation was also positive, but not significant ( r (12)=0.37, p <0.191). Figure 6 shows statistical maps for changes in the normalized current density ln( j ) in the high-beta band between the Happy Memories with rtfMRI-EEG-nf conditions and Rest conditions (H vs R) for the experimental group (EG). The maps correspond to the first rtfMRI-EEG-nf run (Practice; average results across the four nf runs were not significant after the whole-brain correction). The statistical results for voxels with | t |>2.86, two-tailed, trend toward significance ( p <0.10) after the correction. The scale exponent in Fig. 6 is 1.0. The statistical results are summarized in Table 4. Figure 7 reveals prefrontal hemispheric laterality of the normalized high-beta current density changes (Fig. 6). The individual current density changes were averaged within the ROIs on the left and on the right (Table 1). The exploratory paired t -tests indicate that the reductions in high-beta current density during the rtfMRI-EEG-nf task were significantly stronger, with medium or large effect sizes, for the SFG ROIs on the right, than for their counterparts on the left (SFG, BA 8: t (15)=2.44, p <0.028, d =0.61; SFG, BA 9: t (15)=3.63, p <2.4e−3, d =0.91). The mean high-beta current density change is positive for the left amygdala ROI, and negative for the right amygdala ROI, but the positive left-vs-right laterality change is not significant (Amygdala: t (15)=1.15, p <0.267, d =0.29). However, significant across-subjects correlations are found between the amygdala laterality changes and the laterality changes for the prefrontal ROI pairs, e.g. for the MidFG, BA 9 ( r =0.57, p <0.022), the SFG, BA 9 ( r =0.60, p <0.013), the IFG, BA 47 ( r =0.68, p <3.6e−3). For the CG, the laterality effects, corresponding to those described above for the EG (Fig. 7), were non-significant with small negative effects sizes. Figure 8 exhibits statistical maps for correlations between the changes in the normalized high-beta current density during the Happy Memories with rtfMRI-EEG-nf conditions relative to the Rest conditions (H vs R) and the EG participants’ SHAPS ratings. The corresponding statistics are reported in Table 5. Because the statistical Figure 6. eLORETA statistical maps for changes in normalized high-beta (beta 3) current source density ln( j ) between the Happy Memories with rtfMRI-EEG-nf and Rest conditions (H vs R) for the experimental group (EG). The maps are projected onto the MNI152 template. The arrows above the colorbar denote the critical threshold from the randomization SnPM procedure. Statistics are summarized in Table 4. results did not reach the critical threshold in this case, only the largest correlation values are included in the table, without voxel numbers. We also performed exploratory analyses evaluating correlations between changes in left-vs-right laterality of high-beta current densities for the a priori selected ROIs (Table 1) and the patients’ trait anxiety ratings (STAI-t). We considered partial correlations, controlled for the av-erage current density changes for the corresponding left and right ROIs, as well as for the participants’ anhedonia severity ratings (SHAPS). The partial correlations were significant for the MidFG, BA 9 (Laterality vs STAI-t: r (12)=0.63, p <0.017), and for the amygdala (Laterality vs STAI-t: r (12)=0.60, p <0.022).
4. Discussion
In this paper, we reported the first application of EEG source analysis (eLORETA) to evaluate effects of simultaneous real-time fMRI and EEG neurofeedback. Our results show that the eLORETA analyses provide valuable new insights into mechanisms of rtfMRI-EEG-nf training, and complement the fMRI and EEG-fMRI analyses.
The eLORETA results for the upper alpha band (Fig. 3, Table 2) show widespread activation (indicated by reduction in the upper alpha current density) of the frontal, temporal, and parietal brain regions during the rtfMRI-EEG-nf task. These regions are consistent with those identified in the fMRI and EEG-fMRI analyses (Zotev et al., 2019). Fewer activation centers are detected in the limbic and sub-lobar regions (Table 2), compared to the fMRI results. Activation in the occipital lobe appears less prominent (Table 2), because the common average signal, subtracted prior to the eLORETA application, is dominated by occipital alpha activity. The results in Fig. 3 and Table 2 exhibit pronounced hemispheric laterality: reductions in the upper alpha current density during the rtfMRI-EEG-nf task are more significant for the frontal and temporal brain regions on the left than for the corresponding regions on the right.
Figure 7.
Laterality of the high-beta current source density changes (Fig. 6) for corresponding ROIs on the left and on the right. The ROIs were selected a priori as described in the text. The statistics at the bottom of each figure ( p -value from a paired t -test and effect size d ) refer to comparison of the high-beta current density changes on the left and on the right. Figure 8. eLORETA statistical maps for correlations between the changes in normalized high-beta current source density ln( j ) for the Happy Memories with rtfMRI-EEG-nf conditions relative to the Rest conditions (H vs R) and anhedonia severity (SHAPS) ratings for the experimental group (EG). The maps are projected onto the MNI152 template. Statistics are summarized in Table 5. Within the prefrontal cortex, the most significant lateralized current density changes are observed for the left MidFG (BA 9, 8) and for the left IFG (BA 45, 47). The premotor cortex (precentral gyrus, BA 6) is activated bilaterally with similar lateralization. A LORETA study by Pizzagalli and colleagues demonstrated that reward responsiveness, a measure of approach motivation, is associated with reduced resting upper alpha (alpha 2) current density in the left prefrontal regions, including the left MidFG, SFG, and precentral gyrus (Pizzagalli et al., 2005). Similarly, an fMRI study by Spielberg and colleagues revealed an association between trait approach motivation and activation of the left MidFG during a cognitive task (Spielberg et al., 2011). Consistent with these findings, we interpret the left-lateralized activation of the MidFG and adjacent brain regions in our study as indicating enhanced approach motivation during the rtfMRI-EEG-nf task, as we argued previously (Zotev et al., 2016, 2019). The eLORETA results also reveal significant activations in two adjacent areas of the left parahippocampal gyrus, corresponding to BA 34 and BA 28 (Table 2). Both activated areas include voxels in the left amygdala, targeted by the rtfMRI-nf (Fig. 1A), more specifically − in its superficial (SF) subdivision. The SF amygdala subdivision is involved in reward processing and “modulation of approach-avoidance behavior in social interaction” (Bzdok et al., 2013). The observed SF activation is consistent with results of our previous studies that showed enhanced temporal correlations between the FAA in the upper alpha band and BOLD activity in the SF subdivision of the left amygdala during the LA-based rtfMRI-nf training (Zotev et al., 2016) and during the rtfMRI-EEG-nf training (Zotev et al., 2019). Importantly, the laterality effects in Fig. 3 are consistent with the significant positive changes in FAA for channels F3 and F4 (Zotev et al., 2019), targeted with the EEG-nf in our study (Fig. 1A). This fact is illustrated in Fig. 4, which shows significant positive changes in right-vs-left laterality of average current densities for the a priori selected MidFG and SFG ROIs, located approximately underneath EEG electrodes F3 and F4. These findings cannot be predicted in advance based on the FAA results, because the transformation of scalp EEG data to common average reference can obscure frontal alpha asymmetry effects (Hagemann et al., 2001). A recent eLORETA study by Smith and colleagues demonstrated that the right-vs-left intracranial asymmetry scores for resting alpha current density are significantly lower in MDD patients than in healthy individuals for the MidFG and precentral gyrus (Smith et al., 2018). Similar lateralization was found for the upper alpha (alpha 2) band in an earlier LORETA study (Lubar et al., 2003). Our results in Fig. 4 show that such source asymmetry/ laterality becomes more positive as MDD patients upregulate the FAA during the rtfMRI-EEG-nf procedure. Note that more approach-related emotional states are associated with more positive FAA levels (e.g. Stewart et al., 2014). Remarkably, the laterality of upper alpha current density in the amygdala is also significantly increased during the rtfMRI-EEG-nf task (Fig. 4). Moreover, the amygdala laterality changes significantly correlate with the prefrontal laterality changes (Sec. 3.1). These findings confirm the connection between the left-lateralized activations of the prefrontal cortex and the amygdala, we reported previously (Zotev et al., 2016). The variations in upper alpha current density during the rtfMRI-EEG-nf task show significant negative associations with the participants’ anhedonia severity (SHAPS) ratings (Fig. 5, Table 3). These correlations are most pronounced for the prefrontal areas, including the MidFG, the SFG, and the IFG (Table 3). Similar effect is observed for the dorsal anterior cingulate (dACC, BA 24, Table 3). The negative correlations mean that the MDD patients with higher anhedonia severity show stronger activations (greater reductions in upper alpha current density) of these brain areas during the rtfMRI-EEG-nf task. This result is consistent with findings reported in the literature. It has been observed that MDD patients, performing a cognitive task and achieving the same level of cognitive performance as healthy participants, exhibit higher activations of the dorsolateral prefrontal cortex (DLPFC) and dACC, than healthy participants (Fitzgerald et al., 2008; Harvey et al., 2005). The reason is that MDD patients need to recruit activities of these regions to a greater extent to achieve similar performance. Therefore, the results in Fig. 5, demonstrating positive associations between the anhedonia severity and the DLPFC and dACC activations, can be interpreted as showing the ability of the rtfMRI-EEG-nf to correct (i.e. reverse or normalize) functional deficiencies related to anhedonia. This interpretation was proposed and explained in our previous works (Zotev et al., 2016, 2018b, 2019). The correlation effects in Fig. 5 are most significant for the left MidFG (Table 3), suggesting that the MDD patients with more severe anhedonia achieve greater enhancement in approach motivation during the rtfMRI-EEG-nf task. Therefore, the rtfMRI-EEG-nf procedure may be effective, in particular, at correcting approach motivation deficits associated with anhedonia in MDD. Notably, changes in the right-vs-left laterality of upper alpha current densities show significant or trending toward significance positive correlations with the anhedonia severity (controlled for trait anxiety, Sec. 3.1) for the MidFG (BA 8) and for the IFG (BA 47). These results are consistent with the significant positive correlation between the FAA changes and the anhedonia severity (Zotev et al., 2019). They suggest the potential of the rtfMRI-EEG-nf for correcting the FAA and upper alpha source laterality deficits associated with anhedonia. The eLORETA results for the high-beta band (Fig. 6, Table 4) show pronounced reduction in high-beta activity of the fronto-centro-parietal cortical area, extending down to the cingulate gyrus, during the rtfMRI-EEG-nf task. We interpret this effect as an indication of reduction in anxiety. Indeed, the most common anxiety-related pattern revealed by LORETA is the elevated beta activity localized along the anterior cingulate or the midline cortex (Price and Budzynski, 2009). This includes elevated activity in the high-beta band (e.g. Sherlin and Congedo, 2005; Velikova et al., 2010). A LORETA study by Pizzagalli and colleagues showed that MDD patients, compared to healthy individuals, exhibit increased resting high-beta current density in the right prefrontal regions, particularly the right SFG and IFG (Pizzagalli et al., 2002). The high-beta activity in those regions correlated with trait anxiety, accompanying depression (Pizzagalli et al., 2002). An sLORETA study by Paquette and colleagues demonstrated that alleviation of depressive symptoms after treatment is associated with reduction in high-beta current density in several brain regions on the right, including the right medial prefrontal cortex/dACC (BA 9/32) (Paquette et al., 2009). In our study, the reductions in high-beta current density are more widespread for the fronto-centro-parietal regions on the right, than for those on the left (Fig. 6). This observation suggests that the changes in high-beta activity, achieved during the rtfMRI-EEG-nf task, may benefit MDD patients. Importantly, the laterality of the high-beta activity changes in Fig. 6 is consistent with the significant positive changes in FBA for channels F3 and F4 (Zotev et al., 2019), upregulated using the EEG-nf (Fig. 1A). This is demonstrated in Fig. 7, which shows significant positive changes in the left-vs-right laterality of average current densities for the a priori selected SFG ROIs. The occurrence of these effects in the SFG, i.e. along the cortical midline, is consistent with the previous findings (Paquette et al., 2009; Pizzagalli et al., 2002). Interestingly, positive, though non-significant, high-beta laterality change is also observed for the amygdala ROIs. Similar to the upper alpha laterality results, the amygdala laterality changes for the high-beta band significantly correlate with the prefrontal laterality changes (Sec. 3.2). The high-beta current density variations for the fronto-centro-parietal regions during the rtfMRI-EEG-nf task show negative correlations with the MDD patients’ anhedonia severity ratings (Fig. 8, Table 5). These negative correlations mean that the MDD patients with higher anhedonia severity, associated with higher depression and trait anxiety severities for the EG (Sec. 3.1), exhibit stronger reductions in high-beta activity during the rtfMRI-EEG-nf task. The negative correlations are more pronounced for the areas on the right (Fig. 8, Table 5), particularly for the right parietal regions, which may show elevated beta activity in anxiety (e.g. Hammond, 2010). Therefore, the rtfMRI-EEG-nf may have the ability to correct the abnormally elevated right-lateralized high-beta activity in MDD patients. Such activity can conceivably be attributed to comorbid anxiety and avoidance motivation in MDD (Bruder et al., 1997, 2017; Trew, 2011). Interestingly, changes in the left-vs-right laterality of high-beta current densities show significant positive correlations with the trait anxiety severity (controlled for anhedonia severity, Sec. 3.2) for the MidFG (BA 9) and for the amygdala. These results suggest the potential of the rtfMRI-EEG-nf for correcting the FBA and high-beta source laterality deficits related to trait anxiety.
5. Conclusion
The eLORETA source analysis results, reported in this paper, lead to the following conclusions.
First , performance of the rtfMRI-EEG-nf task is associated with significant positive changes in hemispheric lateralities of current source densities in the prefrontal cortical regions. These laterality changes are consistent with the significant positive changes in the upper alpha and high-beta frontal EEG asymmetries during the rtfMRI-EEG-nf task.
Second , the EEG source activities during the rtfMRI-EEG-nf procedure are beneficial to MDD patients. Specifically, the MDD patients with higher symptom severity (depression/anhedonia/anxiety) demonstrate larger reductions in upper alpha current density in the left prefrontal regions, indicating an enhancement in approach motivation. They also exhibit larger reductions in high-beta current density in the right fronto-centro-parietal regions, indicating a reduction in comorbid anxiety. These eLORETA findings suggest that the rtfMRI-EEG-nf may become an effective tool for treatment of major depression.
Availability of data
The data that support the findings and the data analysis scripts used in this study are available from the corre-sponding authors upon reasonable request.
Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Funding
This work was supported by the Laureate Institute for Brain Research and the William K. Warren Foundation and in part by the P20 GM121312 award from National Institute of General Medical Sciences, National Institutes of Health.
References
Allen, J.J.B., Harmon-Jones, E., Cavender, J.H, 2001. Manipulation of frontal EEG asymmetry through biofeedback alters self-reported emotional responses and facial EMG. Psychophysiology 38, 685-693. American Psychiatric Association, 2000. Diagnostic and Statistical Manual of Mental Disorders , n support for the capability model. Psychophysiology 51, 446-455. Talairach, J., Tournoux, P., 1988. Co-Planar Stereotaxic Atlas of the Human Brain. Thieme Medical Publishers, New York, NY. Trew, J.L., 2011. Exploring the roles of approach and avoidance in depression: an integrative model. Clinical Psychology Review 31, 1156-1168. Velikova, S., Locatelli, M., Insacco, C., Smeraldi, E., Comi, G., Leocani, L., 2010. Dysfunctional brain circuitry in obsessive-compulsive disorder: source and coherence analysis of EEG rhythms. NeuroImage 49, 977-983. Zotev, V., Krueger, F., Phillips, R., Alvarez, R.P., Simmons, W.K., Bellgowan, P., et al., 2011. Self-regulation of amygdala activation using real-time fMRI neurofeedback. PLoS ONE 6, e24522. Zotev, V., Phillips, R., Young, K.D., Drevets, W.C., Bodurka, J., 2013. Prefrontal control of the amygdala during real-time fMRI neurofeedback training of emotion regulation. PLoS ONE 8, e79184. Zotev, V., Phillips, R., Yuan, H., Misaki, M., Bodurka, J., 2014. Self-regulation of human brain activity using simultaneous real-time fMRI and EEG neurofeedback. NeuroImage 85, 985-995. Zotev, V., Yuan, H., Misaki, M., Phillips, R., Young, K.D., Feldner, M.T., et al., 2016. Correlation between amygdala BOLD activity and frontal EEG asymmetry during real-time fMRI neurofeedback training in patients with depression. NeuroImage: Clinical 11, 224-238. Zotev, V., Misaki, M., Phillips, R., Wong, CK, Bodurka, J., 2018a. Real-time fMRI neurofeedback of the mediodorsal and anterior thalamus enhances correlation between thalamic BOLD activity and alpha EEG rhythm. Human Brain Mapping 39, 1024-1042. Zotev, V., Phillips, R., Misaki, M., Wong, C.K., Wurfel, B.E., Krueger, F., et al., 2018b. Real-time fMRI neurofeedback training of the amygdala activity with simultaneous EEG in veterans with combat-related PTSD. NeuroImage: Clinical 19, 106-121. Zotev, V., Mayeli, A., Misaki, M., Bodurka, J., 2019. Emotion self-regulation training in major depressive disorder using simultaneous real-time fMRI and EEG neurofeedback. Preprint https://arxiv.org/abs/1909.05764 Table 1.
Locations in the corresponding brain regions on the left and on the right used to define regions of interest (ROIs) in the eLORETA space. The ROIs were defined as collections of voxels within 10 mm distance from the specified centers.
Region Brodmann area Left center x , y , z (mm) Right center x , y , z (mm) Middle frontal gyrus 8 −36, 24, 47 37, 24, 47 Middle frontal gyrus 9 −43, 25, 37 45, 25, 37 Superior frontal gyrus 8 −18, 32, 53 19, 32, 53 Superior frontal gyrus 9 −20, 49, 36 22, 49, 36 Inferior frontal gyrus 47 −39, 24, −11 39, 24, −11 Amygdala 34 −21, −4, −19 21, −4, −19 x , y , z – MNI coordinates. Table 2.
Changes in normalized upper alpha current source density between the Happy Memories with rtfMRI-EEG-nf and Rest conditions (H vs R) for the experimental group (EG).
Region Laterality x , y , z (mm) t -score Voxels Frontal lobe Medial frontal gyrus (BA 6) L,R −10, −25, 50 −3.41 68 Precentral gyrus (BA 6) L −55, 0, 10 −3.37 85 Middle frontal gyrus (BA 9) L −45, 25, 40 −3.30 31 Inferior frontal gyrus (BA 45) L −55, 15, 5 −3.30 28 Inferior frontal gyrus (BA 47) L −50, 15, 0 −3.29 77 Precentral gyrus (BA 6) R 50, −5, 25 −3.28 84 Middle frontal gyrus (BA 6) L −30, −5, 55 −3.27 63 Precentral gyrus (BA 4) L −15, −30, 60 −3.22 60 Precentral gyrus (BA 4) R 50, −10, 45 −3.19 38 Middle frontal gyrus (BA 8) L −40, 25, 45 −3.18 16 Middle frontal gyrus (BA 6) R 40, −5, 50 −3.15 53 Middle frontal gyrus (BA 9) R 50, 25, 40 −3.03 23
Temporal lobe
Superior temporal gyrus (BA 22) L −50, 5, 0 −3.38 75 Middle temporal gyrus (BA 21) L −60, 0, −5 −3.30 93 Superior temporal gyrus (BA 38) L −55, 5, −10 −3.30 69 Inferior temporal gyrus (BA 20) L −55, −5, −35 −3.30 54 Middle temporal gyrus (BA 39) L −45, −80, 20 −3.15 33
Parietal lobe
Precuneus (BA 7) L,R −5, −35, 45 −3.57 144 Paracentral lobule (BA 5) L,R 0, −35, 50 −3.54 52 Precuneus (BA 31) L,R −10, −50, 35 −3.53 35 Superior parietal lobule (BA 7) R 25, −65, 45 −3.35 33 Postcentral gyrus (BA 3) L −20, −30, 50 −3.24 48 Superior parietal lobule (BA 7) L −25, −60, 45 −3.20 34
Occipital lobe
Middle occipital gyrus (BA 19) L −45, −85, 10 −3.08 36
Limbic lobe
Cingulate gyrus (BA 31) L,R 0, −35, 40 −3.59 81 Cingulate gyrus (BA 24) L,R 0, −25, 40 −3.52 66 Cingulate gyrus (BA 23) L,R 0, −25, 35 −3.47 23 Parahippocampal gyrus (BA 34) L −30, 5, −20 −3.00 9 Parahippocampal gyrus (BA 28) L −15, −5, −15 −2.90 7
Sub-lobar
Insula (BA 13) L −45, 5, 5 −3.39 97 Insula (BA 13) R 45, −5, 15 −3.11 29 Corr. p <0.05 for | t |>2.78; BA – Brodmann areas; L – left; R – right; x , y , z – MNI coordinates. Table 3.
Correlations between the changes in normalized upper alpha current source density for the Happy Memories with rtfMRI-EEG-nf conditions relative to the Rest conditions (H vs R) and anhedonia severity (SHAPS) ratings for the experimental group (EG).
Region Laterality x , y , z (mm) Corr. coeff. r Voxels Frontal lobe
Middle frontal gyrus (BA 9) L −30, 20, 35 −0.744 34 Inferior frontal gyrus (BA 9) L −35, 5, 30 −0.732 19 Middle frontal gyrus (BA 8) L −30, 15, 45 −0.731 28 Superior frontal gyrus (BA 8) L −20, 15, 50 −0.712 30 Precentral gyrus (BA 6) L −35, 0, 30 −0.711 18 Middle frontal gyrus (BA 6) L −30, 10, 50 −0.710 25 Middle frontal gyrus (BA 9) R 45, 30, 40 −0.703 36 Medial frontal gyrus (BA 6) L,R −5, 15, 50 −0.703 30 Middle frontal gyrus (BA 8) R 40, 30, 45 −0.702 30 Superior frontal gyrus (BA 6) L −20, 10, 55 −0.699 37 Superior frontal gyrus (BA 8) R 40, 20, 55 −0.684 39 Superior frontal gyrus (BA 6) R 5, 10, 55 −0.680 36 Inferior frontal gyrus (BA 47) L −50, 45, −10 −0.678 24
Parietal lobe
Inferior parietal lobule (BA 40) R 45, −35, 40 −0.677 14
Limbic lobe
Cingulate gyrus (BA 32) L,R −15, 15, 35 −0.752 50 Cingulate gyrus (BA 24) L,R −10, 15, 30 −0.748 40
Sub-lobar
Insula (BA 13) L −35, 5, 20 −0.726 24 Corr. p <0.05 for | r |>0.66; BA – Brodmann areas; L – left; R – right; x , y , z – MNI coordinates. Table 4.
Changes in normalized high-beta current source density between the Happy Memories with rtfMRI-EEG-nf and Rest conditions (H vs R) for the experimental group (EG).
Region Laterality x , y , z (mm) t -score Voxels Frontal lobe Medial frontal gyrus (BA 6) L,R −5, −15, 65 −3.02 58 Superior frontal gyrus (BA 6) L −5, −10, 70 −2.99 10 Superior frontal gyrus (BA 6) R 5, −10, 70 −2.99 9
Parietal lobe
Paracentral lobule (BA 31) L,R −5, −15, 50 −2.96 13
Limbic lobe
Cingulate gyrus (BA 24) L,R −5, −15, 45 −2.91 13 Corr. p <0.10 for | t |>2.86; BA – Brodmann areas; L – left; R – right; x , y , z – MNI coordinates. Table 5.
Correlations between the changes in normalized high-beta current source density for the Happy Memories with rtfMRI-EEG-nf conditions relative to the Rest conditions (H vs R) and anhedonia severity (SHAPS) ratings for the experimental group (EG).
Region Laterality x , y , z (mm) Corr. coeff. r Frontal lobe
Precentral gyrus (BA 4) R 20, −30, 70 −0.531 Precentral gyrus (BA 6) R 20, −20, 70 −0.524 Superior frontal gyrus (BA 6) R 10, −20, 70 −0.513
Parietal lobe
Precuneus (BA 7) R 30, −55, 50 −0.605 Superior parietal lobule (BA 7) R 25, −55, 45 −0.598 Inferior parietal lobule (BA 40) R 35, −55, 55 −0.593 Paracentral lobule (BA 5) R 20, −45, 50 −0.571 Postcentral gyrus (BA 3) R 30, −35, 55 −0.534
Limbic lobe
Cingulate gyrus (BA 31) R 15, −45, 40 −0.535 Corr. p <0.10 for | r |>0.63; BA – Brodmann areas; L – left; R – right; x , y , z – MNI coordinates.– MNI coordinates.