Extensive functional connectivity between brain areas implicated in mental imagery production and phobic fear during both emotional and neutral mental imagery

Mental imagery is used by most people in their day-to-day cognition, for example, in planning, daydreaming, or remembering. Importantly, mental imagery has a powerful influence on emotion and is critically involved in many mental disorders. Thus, understanding the link between mental imagery and emotion is of clinical interest. For example, exposure therapy can be successfully conducted using mental imagery of fear-provoking stimuli, i.e., imaginal exposure. In this vein, accumulating evidence shows that mental imagery of a fearful stimulus produces a similar physiological and neural response as actual perception of the stimulus. Alas, knowledge of the neural processes underlying the link between mental imagery and emotion is limited. Functional magnetic resonance imaging data from a previous study on imaginal exposure (N = 30) was used to examine the functional connectivity during the production of phobic and neutral mental imagery. Regions of interest were selected from meta-analyses on brain regions consistently recruited during mental imagery production and phobic fear, respectively. Results showed that these regions were positively correlated during both phobic and neutral mental imagery production. Very few differences in functional connectivity between phobic and neutral imagery were found. Specifically, weaker functional connectivity between the supplemental motor area and a region including parts of the left insula and inferior frontal gyrus was observed during phobic (vs neutral) imagery. In conclusion, our findings suggest that brain regions previously implicated in mental imagery production and phobic fear are highly interconnected during the production of both phobic and neutral imagery.


Introduction
Mental imagery refers to sensory experiences created internally without corresponding sensory input [1,2].Mental imagery can have a powerful influence on our mood and motivation [3] and elicit emotions to a greater extent than verbal thoughts [4].Understanding the link between mental imagery and emotion is clinically relevant for several reasons.First, distressing intrusive mental imagery is present in many mental disorders, such as posttraumatic stress disorder (PTSD) [5].Secondly, mental imagery is widely used in psychological treatments.For instance, imaginal exposure is an effective treatment technique [6] and a key component in a golden standard treatment for PTSD (Prolonged exposure) [7].Third, the impact of mental imagery on emotion and motivation can be harnessed to improve psychological treatments [8].For example, producing mental imagery of positive future activities increases motivation, anticipated reward, and the probability that these will actually be performed [3] and thus provides a promising avenue to improve the effects of behavioral activation in depression [9].Moreover, mental imagery of an event's positive outcome reduces worry associated with the event [10].Thus, a better understanding of the link between mental imagery and emotion could help us further elucidate the role of emotional mental imagery in psychopathological conditions, promote treatment development, and ultimately alleviate human suffering.
The dominant theory underlying the use of mental imagery in psychological treatments is that mental imagery produces physiological and psychological responses similar to those of the actual stimulus, e.g., fear response [11] and memory activation [12].For instance, the ability of emotional mental imagery to produce a physiological arousal response [13,14] is part of the theoretical underpinnings of imaginal exposure [11].In support of this notion, it has been demonstrated that mental imagery can stand in for visual perception and create a similar physiological response in fear conditioning and extinction [15][16][17].Moreover, in a previous study, we observed that fear responses elicited by mental imagery of phobic stimuli produced subjective fear and activated brain areas similar to those activated by direct perception of phobic stimuli, such as the amygdala, insula, and anterior cingulate cortex (ACC) [14].Hence, accumulating evidence corroborates the notion that mental imagery of stimuli can produce a similar emotional response as direct perception at subjective, physiological, and neural levels.However, knowledge of the neural processes underlying the ability of mental imagery to elicit emotion is limited.
A meta-analysis of brain imaging studies of mental imagery was recently published, providing several brain regions that were consistently activated during mental imagery production [18].The results implicated that mental imagery production relies on higher visual areas (e.g., the fusiform gyrus) and fronto-parietal networks supporting high-level cognitive functions, notably challenging the influential theory that mental imagery uses the same neural system as perception, including lower-level primary visual cortices [19].
Research on the neural underpinnings of emotion has a rich history, spanning more than a century [20].Although neuroimaging studies have undeniably advanced the field, there is still an ongoing debate over whether emotions are the results of specific neural structures (the locationist account) or are constructed of more general neural networks (the constructionist account).Within modern affective neuroscience, these represent two different approaches: one focusing on localizing specific brain areas activated by each emotion [21] and the other on identifying functional hubs within the human brain that are used to construct emotions [22].Nevertheless, accumulating evidence for both accounts continues to implicate some key areas (e.g.amygdala, insular cortex, anterior cingulate cortex (ACC), and orbitofrontal cortex) [21][22][23].
Neuroimaging studies examining the neural underpinnings of emotional mental imagery are still relatively few.Several studies agree with the notion that emotional mental imagery engages both brain areas involved in emotion processes and brain areas involved in mental imagery.For example, as previously mentioned, results from a neuroimaging study in which participants afraid of spiders underwent a session of imaginal exposure showed that mental imagery of a phobic stimulus activated similar brain areas as those reported in previous imaging studies on visual phobic stimuli [24].These areas included the amygdala, thalamus, anterior insula, mid-cingulate gyrus (MCC), hippocampus, and other brain regions such as the cerebellum [14].Partly corroborating these results, another study reported increased activation in insula regions and MCC during mental imagery of risky (cf.emotional) compared to non-risky (cf.neutral) situations [25].Furthermore, the insular cortex was also implicated in two neuroimaging studies examining mental imagery to fear conditioned stimuli, suggesting the anterior insula as an important hub for the interaction between emotion and imagery [16,26].However, data also exist that do not agree with the notion that emotional mental imagery simply engages brain areas involved in emotion processes and mental imagery.One study, using a within-group design, compared neural activity when participants viewed or imagined emotional and neutral events.Results showed few overlapping brain activations during the perception of emotional stimuli and emotional mental imagery [27].Whereas viewing pictures activated the amygdala, thalamus, and anterior insula, emotional mental imagery was observed to activate the hippocampus, cerebellum, and dorsomedial prefrontal cortex.These results imply that different neural circuitries are employed when emotion is elicited by an external stimulus compared to when it is produced internally.In sum, conflicting results and the limited number of neuroimaging studies investigating emotional mental imagery highlight the need for further research.Specifically, more knowledge of the functional connectivity patterns could help bridge the gap between inconsistent results in activation studies and contribute to elucidating the neural mechanisms underlying the link between mental imagery and emotion.
Although little is known about brain functional connectivity during emotional mental imagery, Greening et al. [16] showed that mental imagery of the conditioned stimulus was associated with greater functional connectivity between the anterior insula and other areas previously associated with fear conditioning and emotion (e.g.bilateral amygdala, hippocampus) compared to the perception of the conditioned stimulus [16].An EEG-study using mental imagery as a method for mood induction, investigated effective connectivity to different emotional valence by means of the directed transfer function (DTF) methodology.It suggested three major connectivity hubs; the orbitofrontal cortex, which reduced its connectivity to right temporal and parietal areas during core affect; the right temporal area, which showed an increase in information flow to many other areas during negative affect; the right parietal area, which showed an increased information flow to bilateral temporal cortex during positive affect [28].Building on results from meta-analyses of neuroimaging findings, an avenue to extend our knowledge of the interplay between emotion and mental imagery is to examine the functional connectivity between brain areas found to be most consistently activated during emotional processing [24] on the one hand and those consistently activated during visual mental imagery on the other hand [18].
The aim of the current study was to elucidate the neural underpinnings of emotional mental imagery by examining the functional connectivity between brain areas previously implicated in mental imagery production and phobic fear, respectively.To this end, functional magnetic resonance imaging (fMRI) data from a previous study [14] on imaginal exposure in individuals fearful of spiders (N = 30) was used.The experimental procedure entailed producing mental imagery of different situations with spiders (phobic mental imagery) or corresponding situations with a neutral stimulus (a glove; neutral mental imagery).Region-to-region (ROI-to-ROI) connectivity (RRC) during the production of phobic and neutral mental imagery was examined using regions of interest (ROIs) identified in meta-analyses of brain areas activated by mental imagery production [18] and phobic fear [24], respectively.In addition, given the key role of the amygdala in fear research, seed-to-voxel analyses between the amygdala and the whole brain were also performed to explore differences in amygdala connectivity during phobic and neutral mental imagery.In line with the observed higher activation in emotion-processing brain areas during phobic vs neutral mental imagery in Hoppe et al. [14], we expected to find a stronger functional connectivity between brain areas associated with mental imagery production and brain areas implicated in phobic fear during phobic mental imagery as compared to neutral.

Participants
Thirty participants (age: M = 24.0,SD = 5.6 years: 22 women and 8 men) fearful of spiders were recruited through social media and billboards [14].Participants were included if they had a score of 19 or higher on the spider phobia questionnaire (SPQ) [29].Exclusion criteria consisted of current psychiatric disorders other than spider phobia (screened using the Mini International Neuropsychiatric Interview [30]), having received psychological treatment or psychotropic medication within six months, neurological disease, substance abuse, and contraindications for magnetic resonance imaging.Participants were reimbursed 500 Swedish kronor, corresponding to $50 for their participation.Participants also consented to anonymized data being made publicly available and used to examine other research questions.

Procedure
Before entering the scanner, participants underwent general and task-specific mental imagery training (mental imagery protocol can be found here: https://osf.io/7mg2r/)and the Plymouth Sensory Imagery questionnaire (PSI-Q) [31] was collected.Participants were asked to listen to the descriptions and to try to imagine the scenes as vividly as possible while keeping their eyes closed throughout the procedure.A recorded voice provided short descriptions (2-4 s) of situations containing (phobic/neutral) content.Phobic and neutral situations were the same, but in the neutral ones, spiders were replaced with gloves (e.g., 'you touch a spider'; 'you touch a glove').After hearing the description, participants produced mental imagery of the described situation for 7 s, after which a bell signaled that the participant could stop producing mental imagery and relax (for 6 s) until the next description.There were 13 descriptions of situations with spiders, and 13 corresponding situations including gloves.Stimuli were presented in pairs, where the situation involving phobic imagery always preceded the situation with the neutral situation.This approach was taken to minimize expectancy effects, as presenting the neutral situation first would pose the risk of eliciting expectancy fear as participants could fret about how the corresponding situation would play out when a spider is brought into the scenario in the upcoming trial.The situations grew increasingly fearful during the session, mimicking exposure therapy.Three times during the session, participants were reminded to imagine the scene as vividly and with as much detail as possible and to regenerate the image if it was lost.The entire experimental fMRI procedure lasted for 10 min, during which skin conductance responses (SCRs) to the mental imagery production were also measured.More details on the procedure and a complete list of mental imagery situations can be found in the original article and its supplementary materials [14].
After the exposure session, the participants reported their estimated imagery vividness (1 = not at all vivid; 5 = extremely vivid) and subjective (0 = no fear at all; 100 = extreme fear) fear during the procedure.The mean imagery vividness was high (M = 3.68; SD = 0.70) and correlated with the reported subjective fear, r S = 0.48, p = 0.008, suggesting that participants complied with the instructions and that the produced mental imagery was the source of subjective fear.
For this study, BOLD images were preprocessed using CONN [32] (RRID:SCR_009550) release 21.a [33] and SPM 12 (http://fil.ion.ucl.ac.uk/spm/) using the default preprocessing pipeline in CONN toolbox release 21.a.BOLD images were realigned and unwarped, slice-time corrected, normalized to the Montreal Neurological Institute (MNI) standard space (voxel size 2 × 2 × 2), and denoised using the CONN toolbox standard denoising pipeline.The denoising involves estimating and removing potential confounding effects on the BOLD signal (e.g.noise components from cerebral white matter and cerebrospinal areas, estimated subject-motion parameters) using regression, and temporal band-pass filtering of frequencies below 0.008 Hz or above 0.09 Hz.Data was not spatially smoothed.

Regions of interest 2.4.1. Mental imagery
Regions of interest (ROIs) for brain areas involved in mental imagery production were drawn from the meta-analysis by Spagna et al. [18] using the contrast 'Visual Mental Imagery vs Control condition' [18].This approach yielded eight clusters numbered Visual Mental Imagery (VMI) 1-8.

Phobic fear
ROIs for phobic fear were drawn from the meta-analysis by Gentili et al. [24], and the contrast 'Phobic individuals vs Healthy controls' was calculated from both the studies that used ROI analyses and the studies using whole-brain analysis [24].This approach yielded four clusters numbered Phobic Fear (PF) 1-4.
A conjunction analysis performed on all the chosen ROIs revealed a small overlap of the ROIs associated with mental imagery and the ROIs associated with phobic fear (VMI 6 and PF 4, 8 voxels; VMI 3 and PF 3, 29 voxels).In order to avoid artificially inflated connectivity between ROIs, that is, to avoid the same voxels being included in more than one ROI, the overlap was removed from the ROIs associated with phobic fear.See Supplementary Materials Fig. 1 for the location of the removed overlaps.The final ROIs can be found in Table 1 and Fig. 1.
Weighted ROI-to-ROI connectivity (wRRC) uses user-defined temporal weights to identify which volumes belong to which experimental condition, and then calculates correlation maps for a specific set of ROIs for each condition, respectively.Thus, wRRC was used to produce correlational maps of connectivity for phobic mental imagery and neutral mental imagery, respectively.The connectomes were evaluated using the CONN toolbox's functional network connectivity (FNC) multivariate parametric statistics using the default cluster threshold of p < 0.05 FDR-corrected (MVPA omnibus test) with p < 0.05 connection threshold [34].In addition, single connections were explored using both p < 0.05 FDR corrected and p < 0.001 uncorrected, i.e., the CONN toolbox default setting for single connection analysis.
Generalized psycho-physiological interaction (gPPI) analysis is used to test functional connectivity between experimental conditions by studying the interaction term between psychological variables (experimental conditions) and a physiological variable (imaging data) [35].Thus, gPPI was used to study whether there was a difference in RRC between phobic and neutral mental imagery production.Moreover, another method to test functional connectivity differences between experimental conditions was also explored, that is, to contrast wRRC correlational maps with each other.Again, the produced connectomes were evaluated using the CONN toolbox's functional network connectivity (FNC) multivariate parametric statistics using the default cluster threshold (see above).Single connections were explored using both p < 0.05 FDR corrected and p < 0.001 uncorrected.
Weighted seed-based connectivity (wSBC) employs user-defined temporal weights to identify which volumes belong to which experimental condition, and then calculates correlation maps between the seed and the whole brain for each condition, respectively.Thus, wSBC analyses were used to evaluate seed-to-voxel connectivity maps during the production of phobic and neutral mental imagery, respectively, using the mean of the activity in both amygdala as seed.In addition, a gPPI analysis was used to assess change in amygdala connectivity between the production of phobic and neutral mental imagery in the whole brain.

Functional connectivity during phobic and neutral mental imagery
A wRRC analysis was performed on both phobic and neutral mental imagery, respectively.Analyses showed that brain areas associated with mental imagery and areas associated with phobic fear were extensively connected during the production of both phobic and neutral mental imagery (Fig. 2, Supplementary Table 1 and 2).In depicting the connectomes below, ROIs were grouped according to previously noted functions of their included brain areas (VMI 1, 3, 4, & 5 include parts of the premotor cortex and the supplemental motor area; VMI 2, 7 & 8 include parts of the occipital cortex, the angular gyrus, and fusiform gyrus; VMI 6 is grouped with the PF ROIs as it contains the left insula).Note that this grouping is simplified and does not convey the ROIs' full complexity.Please consult Table 1 and the original sources [18,24] for the full details.

Differences in functional connectivity between phobic and neutral mental imagery
A gPPI analysis (phobic vs neutral mental imagery) did not produce any significant cluster-based inference.However, the single connection approach (132 connections among 12 ROIs) produced a significant difference in the connection between the ROIs VMI 6 and VMI 3, i.e., between the left insula and the supplemental motor area, t (29) = − 4.12, p unc = 0.0003, p FDR = 0.04.A wRRC approach was also explored, which showed a few more connections, all between the ROIs associated with phobic fear (i.e., emotion) and the premotor areas (Fig. 3, Supplementary table 3), indicating that the functional connectivity between these areas is somewhat weaker during phobic mental imagery, than during neutral mental imagery.
A gPPI analysis, examining the change in amygdala connectivity between the production of phobic and neutral mental imagery did not produce any significant results.

Amygdala functional connectivity during phobic and neutral mental imagery
A wSBC analysis, using the bilateral amygdala as seed region, showed functional connectivity between the amygdala and large parts of the brain during both phobic and neutral mental imagery (Fig. 4).

Discussion
The aim of the study was to further elucidate the neural underpinnings of emotional mental imagery.To this end, we assessed the functional connectivity during the production of mental imagery of phobic and neutral content, as well as the difference between the two, in individuals fearful of spiders.Brain regions of interest were drawn from two independent meta-analyses, one identifying core brain regions involved in mental imagery production [18], and the other delineating brain regions consistently activated by phobic fear [24].Our results showed that brain regions previously implicated in mental imagery production and phobic fear were extensively interconnected both during phobic and neutral mental imagery.The only differences in functional connectivity between phobic and neutral mental imagery were observed between the supplemental motor area (SMA) and the precentral gyrus on one side, and the left insula and areas previously shown to be activated by phobic fear, on the other side.
We hypothesized that the functional connectivity between ROIs implicated in mental imagery [18] production and ROIs associated with phobic fear [24], i.e., emotion, would be more pronounced when mental imagery included phobic stimuli compared to neutral stimuli.Instead, results showed that phobic and neutral mental imagery displayed similar functional connectivity patterns between ROIs, even though many of the regions associated with phobic fear were significantly activated during phobic imagery [14].Similarly, even though significantly higher activation of the amygdala was noted during phobic mental imagery, as compared to neutral [14], whole-brain analyses of amygdala connectivity likewise showed a similar functional connectivity pattern during both phobic and neutral mental imagery.Thus, amygdala functional connectivity with other brain regions was not  1.2.5.

Table 1
Regions of Interest (ROIs) used in the analyses.From left to right: ROI, the brain regions included in these clusters (aal v3), the MNI coordinates (x, y, z) where peak ALE was found, Peak ALE, and volume in mm 3  significantly altered by the emotional content of mental imagery.These results imply that regions associated with mental imagery show continuous functional connectivity with regions associated with emotion and do not recruit additional brain regions to convey emotion when imagery content goes from neutral to emotional.However, the content of imagery may direct the activity in this circuit.For example, based on the observed network of positive correlation between ROIs, it is possible that activation elicited in one, or a number of key regions (e.g., the amygdala) within the circuit may drive the activation level of the remaining brain areas.Alas, correlations do not allow for causal conclusions, and it is therefore not possible to conclude which brain areas may be driving emotion based on the present analyses.
Results revealed a weaker, positive functional connectivity during phobic, as compared to neutral, mental imagery between regions in the SMA, the precentral gyrus, and regions previously associated with phobic fear, especially the left insula.The precentral gyrus (Brodmann area 6) is commonly thought of as the premotor cortex, and the SMA is believed to contribute to motor responses and planning [36].However, the SMA is also involved in action response and action inhibition [37], as well as emotional tasks, such as emotion regulation through reappraisal [38].Thus, a weakening of the functional connectivity between SMA and areas associated with phobic fear may reflect many different processes.One possibility is that the consistent activation of the SMA during mental imagery partly reflects the preparation of action responses associated with the content of mental imagery.That is, the SMA may be recruited for action response and inhibition to a stimulus regardless of whether the stimulus is simulated in mental imagery or perceived through direct perception.If so, the activity needed for directing action responses may simply not be proportionate to the strength of the emotional response, leading to a seemingly weaker connection when emotion-processing areas kick into gear.Another possibility, if SMA is concerned with action response and inhibition, is that the weaker functional connectivity between it and emotion-processing brain areas during phobic imagery represents participants' difficulty inhibiting their action response to the phobic content.Given the involvement of SMA in emotion regulation, the difference in connectivity may represent the effort to inhibit the emotional response elicited by mental imagery of phobic stimuli.Future studies should further examine the link between emotional mental imagery and the SMA, including the functional connectivity between SMA and emotion-processing brain areas.Moreover,  previous results have tied activity in the right insula to emotional mental imagery [16,26].Even though the functional connectivity changes found in the present study involved the left insula, the present results add to previous findings indicating the involvement of the insular cortex in emotional mental imagery.
The present study has some limitations that should be considered when interpreting the results.First, the limited sample size may have precluded the detection of more subtle differences in the functional connectivity between phobic and neutral mental imagery.Nevertheless, the sample size was large enough to reveal several statistically significant functional connections.Second, mental imagery ROIs selected from the meta-analysis [18] focused only on visual mental imagery.This approach may have occluded differences in emotion-related functional connectivity associated with brain regions specific to other mental imagery modalities.For example, it was pointed out in early research on emotional mental imagery that the inclusion of somatic responses in imagery instructions increased psycho-physiological responses, presumably as a form of interoceptive mental imagery [39].Participants in the study from which the present fMRI data was taken [14], included somatic details in its mental imagery instructions.However, the subject of somatic responses in mental imagery production is not touched upon in the meta-analysis from which mental imagery ROIs were taken [18].Future studies should include ROIs associated with mental imagery modalities beyond visual mental imagery.Third, our selection of ROIs may not be optimal.We chose to include only ROIs identified from studies that compared visual imagery to a control condition [18], because it corresponded to the methodology of the data used in the current study.However, results from other groups of studies, or from another meta-analysis [40], yielded varied sets of ROIs, which may characterize visual mental imagery better.Fourth, results showed that all ROIs were positively correlated, meaning they were all activated in concert to some degree during both phobic and neutral imagery.Thus, although the selected brain regions may likely play a role in the link between mental imagery and emotion, the similar functional connectivity during both conditions suggests that other types of connectivity analyses may be required to untangle the specific mechanisms at work.Fifth, the results are calculated on participants afraid of spiders, and may not hold for other populations.Finally, as imagery vividness was correlated with subjective fear ratings, it is possible that functional connectivity may differ with the vividness of mental imagery.This is an interesting line of research for future studies, where perhaps a group of aphantasics (i.e., individual's not able to produce visual mental imagery [41]) could be compared with a group from the normal population, or a task could be used that produces large variations in imagery vividness.T. Agren and J.M. Hoppe

Conclusions
In conclusion, our findings suggest that brain regions previously implicated in mental imagery production and those activated by phobic fear show high and similar functional connectivity during mental imagery of both emotional (phobic) and neutral content.However, further research is warranted to unravel the specific neural mechanisms underlying the link between mental imagery and emotion.

Fig. 2 .
Fig. 2. Weighted ROI-to-ROI connectivity analysis for phobic and neutral imagery.Statistics for single connections can be found in supplementary tables 1 & 2.

Fig. 3 .
Fig. 3. Differences in connectivity between the production of phobic and neutral imagery (phobicneutral) explored with a (left) gPPI, and a (right) wRRC approach.

Fig. 4 .
Fig. 4. Amygdala connectivity during phobic and neutral imagery.Images show correlations between the mean of bilateral amygdala activity and voxels of the whole brain (cluster threshold p FDR <0.05, voxel threshold p < 0.001 uncorrected; MNI coordinates for the images: left X = 5, middle Z = − 15, right Y = 0).See Supplementary tables 4 & 5 for statistics and brain areas included in the clusters.