Voluntary control of anterior insula and its functional ...

NeuroImage 130 (2016) 230–240

Contents lists available at ScienceDirect

NeuroImage journal homepage: www.elsevier.com/locate/ynimg

Voluntary control of anterior insula and its functional connections is feedback-independent and increases pain empathy Shuxia Yao a,1, Benjamin Becker a,b,1, Yayuan Geng a, Zhiying Zhao a, Xiaolei Xu a, Weihua Zhao a, Peng Ren a, Keith M. Kendrick a,⁎ a b

Key Laboratory for NeuroInformation of Ministry of Education, Center for Information in BioMedicine, University of Electronic Science and Technology of China, Chengdu, Sichuan 610054, China Department of Psychiatry, Division of Medical Psychology, University of Bonn, 53105 Bonn, Germany

a r t i c l e

i n f o

Article history: Received 5 August 2015 Accepted 11 February 2016 Available online 17 February 2016 Keywords: Empathy Insula Neurofeedback Real-time fMRI Resting state functional connectivity

a b s t r a c t Real-time functional magnetic resonance imaging (rtfMRI)-assisted neurofeedback (NF) training allows subjects to acquire volitional control over regional brain activity. Emerging evidence suggests its potential clinical utility as an effective non-invasive treatment approach in mental disorders. The therapeutic potential of rtfMRI-NF training depends critically upon whether: (1) acquired self-regulation produces functionally relevant changes at behavioral and brain network levels and (2) training effects can be maintained in the absence of feedback. To address these key questions, the present study combined rtfMRI-NF training for acquiring volitional anterior insula (AI) regulation with a sham-controlled between-subject design. The functional relevance of acquired AI control was assessed using both behavioral (pain empathy) and neural (activity, functional connectivity) indices. Maintenance of training effects in the absence of feedback was assessed two days later. During successful acquisition of volitional AI up-regulation subjects exhibited stronger empathic responses, increased AI-prefrontal coupling in circuits involved in learning and emotion regulation and increased resting state connectivity within AIcentered empathy networks. At follow-up both self-regulation and increased connectivity in empathy networks were fully maintained, although without further increases in empathy ratings. Overall these findings support the potential clinical application of rtfMRI-NF for inducing functionally relevant and lasting changes in emotional brain circuitry. © 2016 Elsevier Inc. All rights reserved.

Introduction While functional magnetic resonance imaging (fMRI) has greatly advanced our understanding of the neurobiological basis of brain-based disorders, particularly psychiatric conditions, the clinical application of fMRI remains limited (Botteron et al., 2012). Recent developments in fMRI technologies have the potential to promote the translation of fMRI approaches from basic scientific research to clinical applications, including innovative and non-invasive treatments for psychiatric disorders (Linden, 2014; Stoeckel et al., 2014). One of the most promising technologies is real-time fMRI (rtfMRI), which allows real-time assessment of regional brain activity (see Weiskopf et al., 2003, 2004, 2007; Weiskopf, 2012 for a detailed overview). rtfMRI is a powerful technique to transform the real-time activity of brain regions into visualized feedback, which enables subjects to learn volitional control over regional brain activity. To date this neurofeedback

⁎ Corresponding author at: No. 2006, Xiyuan Ave., West Hi-Tech Zone, Chengdu, Sichuan 611731, China. Fax: +86 28 83201358. E-mail address: [email protected] (K.M. Kendrick). 1 Contributed equally to this work.

http://dx.doi.org/10.1016/j.neuroimage.2016.02.035 1053-8119/© 2016 Elsevier Inc. All rights reserved.

(NF) approach has been used in the context of a number of important nodes in cognitive and emotional processing networks, such as the inferior frontal gyrus (Rota et al., 2008), amygdala (Brühl et al., 2014; Paret et al., 2014; Posse et al., 2003; Zotev et al., 2011), anterior cingulate cortex (ACC; Weiskopf et al., 2003; deCharms et al., 2005), amygdala–insula networks (Johnston et al., 2010) and even the mesolimbic dopamine system (Sulzer et al., 2013). Importantly, initial clinical studies have also produced promising results, suggesting that rtfMRI-NF is a safe and efficient non-invasive strategy for helping to modify aberrant brain activity patterns in psychiatric populations (Hawkinson et al., 2012; Linden, 2014; Stoeckel et al., 2014), including patients with major depression (Linden et al., 2012; Young et al., 2014), contamination anxiety (Scheinost et al., 2013) and schizophrenia (Ruiz et al., 2013). Nevertheless, research has only just begun to explore the therapeutic application of rtfMRI-NF as a neuromodulatory strategy and several important issues need to be clarified (Linden, 2014; Stoeckel et al., 2014; Sulzer et al., 2013). As with other novel neuromodulatory treatment strategies, such as transcranial magnetic stimulation (TMS) (Rossi et al., 2009), the therapeutic potential of rtfMRI-NF particularly depends on: (1) whether the alterations in brain activity are of functional relevance (i.e. affect emotional or cognitive processing of the individual both in terms of behavior and alterations in functional circuitry in the brain), and (2) whether the neuromodulatory

S. Yao et al. / NeuroImage 130 (2016) 230–240

effects last beyond the duration of initial training in the absence of further feedback. To specifically address these questions, we conducted a shamcontrolled between-subject study that used rtfMRI-NF to train healthy subjects in the volitional control of the anterior insula (AI), and assessed neural (activity and functional connectivity) and behavioral indices of success during initial training and then subsequently their maintenance after 2 days in the absence of feedback. The AI insula is involved in a broad range of cognitive and social emotional functions which support successful social interaction and flexible behavioral control (Adolphs, 2003; Engen and Singer, 2013; Simmons et al., 2013; Koban and Pourtois, 2014). Previous studies indicate a crucial contribution of the AI in empathic processing (Decety and Jackson, 2006; Engen and Singer, 2013; Lamm et al., 2007; Singer, 2006). Particularly the empathic response during the observation of others suffering from pain (pain empathy) seems to specifically and crucially depend on the AI (Gu et al., 2010, 2012; Jackson et al., 2005, 2006; Lamm et al., 2011; Singer et al., 2004). Moreover, aberrant AI processing has been observed across the most common psychiatric disorders, including major depression (Hamilton et al., 2012), anxiety (Etkin and Wager, 2007) and autism spectrum disorders (ASD) (Di Martino et al., 2009) and initial studies have revealed deficient empathic processing in psychiatric disorders with social emotional deficits, such as ASD and schizophrenia (Derntl et al., 2009; Jones et al., 2010) that might contribute to the social interaction difficulties commonly observed in these patient populations. Given the highly specific role of the AI in pain empathy and its relevance for a range of psychiatric disorders, pain empathy ratings were chosen as a sensitive behavioral read-out to evaluate the functional relevance of the volitional AI regulation. While previous studies have demonstrated successful AI modulation in healthy and psychiatric populations after brief rtfMRI-assisted NF training, reports of functional effects have been inconsistent (Caria et al., 2010; Veit et al., 2012; Berman et al., 2013; Lawrence et al., 2014; Ruiz et al., 2013). Thus one study found associations between successful modulation of AI activity and valence ratings (Caria et al., 2010), whereas others failed to (Berman et al., 2013; Lawrence et al., 2014). Valence processing is a continuous and partly implicit evaluative process that critically relies on additional brain regions, particularly the amygdala (Berntson et al., 2011). In contrast, pain empathy processing has been specifically associated with the AI (Lamm et al., 2011; Bernhardt and Singer, 2012) and particularly relates to pain observation in others. It seems reasonable to hypothesize therefore that pain empathy processing might have a higher sensitivity for demonstrating behavioral effects of a focal AI modulation using rtfMRI-NF. Similarly, while some previous rtfMRI studies have suggested that subjects can maintain volitional control of their brain activity in the absence of NF (e.g., Scheinost et al., 2013; Young et al., 2014), others have not (Berman et al., 2013). However, assessments of training success have usually been made immediately after training without a strict control for non-specific training effects by using a sham group and only in terms of regional activation rather than functional indices (overview in Linden, 2014; Stoeckel et al., 2014). Furthermore, no studies have investigated the effects of learned volitional control of the AI on its patterns of functional connections during effective regulation or in terms of its resting state functional connectivity (RSFC). The present study used a randomized double-blind sham-controlled between-subject design to evaluate the neurotherapeutic potential of rtfMRI-NF. Given that meta-analytic data indicates predominant left AI (LAI) activity during emotional processing and a previous reported altered valence rating after LAI rtfMRI-training, the LAI was chosen as the target ROI (Caria et al., 2010; Wager et al., 2003). Neural and behavioral effects (pain empathy ratings) of rtfMRI-NF training of the AI were assessed during the training and after two days without feedback being given. In line with previous studies (Hampson et al., 2011; Scheinost et al., 2013), resting state fMRI scans were included before and after

231

the training to capture complex changes at the network level (see Fig. 1a for an overview). Materials and methods Participants 37 healthy students (19 male, mean age = 21.97 years, SD = 1.40) from the University of Electronic Science and Technology of China (UESTC) participated in the study. 21 subjects (10 male) were randomly assigned to the experimental group and 16 subjects (9 male) to the control group. All subjects gave informed consent before the experiment and were free of past or current psychiatric or neurological disorders. 3 subjects were excluded due to excessive head movement larger than 3 mm or 3 degrees (1 male in each group and 1 female in the control group). Another 2 subjects were excluded due to reporting being unable to perform the regulation strategies when debriefed after the whole experiment (2 females in the experimental group). In a post-scan interview these latter two subjects reported that they could not recall negative memories or felt nothing when recalling negative personal events. Thus 18 subjects in the experimental group and 14 subjects in the control group were included in the final analysis. To exclude potential confounding effects from personality traits or current mood states, all subjects completed the Chinese versions of validated psychometric questionnaires before the start of the experiment, including the Positive and Negative Affect Schedule (PANAS — Watson et al., 1988), Behavioral Inhibition System (BIS) and Behavioral Activation System (BAS — Carver and White, 1994), Autism Spectrum Quotient (ASQ — Baron-Cohen et al., 2001), Beck Depression Inventory-II (BDI-II — Beck et al., 1996), State–Trait Anxiety Inventory (STAI — Spielberger et al., 1983), and Empathy Quotient (EQ — Baron-Cohen and Wheelwright, 2004). Written informed consent was provided to all subjects before study inclusion. The study and all procedures were approved by the local ethical committee at UESTC and were in accordance with the latest version of the Declaration of Helsinki. Experimental group fMRI localizer task In line with previous rtfMRI-NF studies (Caria et al., 2010; Ruiz et al., 2013; Veit et al., 2012), the LAI was initially localized functionally for each subject. To this end subjects were presented with color pictures depicting individuals in painful situations and neutral pictures in a blocked design session. All painful pictures (mean valence 2.90 ± 1.15, mean arousal 5.07 ± 1.03) in the present study were from Meng et al. (2012). These pictures depicted painful situations that occasionally happen in daily life, such as a hand cut by a knife or stabbed by a syringe (see Fig. S1 for examples; Meng et al., 2012). All neutral pictures were from the International Affective Picture System (IAPS; Lang et al., 2005) and depicted neutral objects such as a book or a chair (mean valence 4.95 ± 0.15, mean arousal 2.77 ± 0.44). Subjects were instructed to imagine how painful the person feels when presented with a painful picture, and to just watch when presented with a neutral picture. The localizer consisted of 8 blocks (4 blocks of neutral pictures and 4 blocks of painful pictures) alternating with a 30 s baseline interval (fixation cross). Each picture was presented for 2 s and there were 10 pictures within each block. The pictures within each block were presented randomly to subjects. The localizer session lasted about 7 min. NF training task The NF training task was adopted from a procedure used by Caria et al. (2010) and consisted of 4 training sessions each lasting about 10 min. During each session, 5 regulation blocks (30 s) alternating with 6 baseline blocks (30 s) were presented. During NF training subjects in the experimental group viewed a display showing the activity within the LAI region of interest (ROI) as a graphical thermometer

232


Fig. 1. (a) Experimental protocol. (b)The paradigm for the rtfMRI-based NF training task.

composed of graduating bars (the more bars were filled, the stronger the AI was activated and vice versa). A regulation block was indicated by an upward arrow while a baseline block was indicated by a downward one. In line with previous studies (Berman et al., 2013; Caria et al., 2007, 2010), during regulation blocks subjects were instructed to increase the number of filled bars of the thermometer by recalling or imagining negative emotional and personally relevant events. For the baseline blocks, subjects were instructed to rest with their eyes open to return the AI activity back to the baseline level. Thermometer bars were updated every 2 s. To assess the functional relevance of the AI modulation both regulation and baseline blocks were followed by a painful picture (8 s) and 2 rating tasks (each 6 s). The first rating assessed pain empathy and subjects had to rate how much they felt the pain being experienced by the person in the picture. The second rating assessed arousal using the SelfAssessment Manikin (Bradley and Lang, 1994). Ratings were assessed on a 9-point Likert scale (pain intensity: from 1 = not at all to 9 = very painful; arousal: from 1 = calm to 9 = very excited) and indicated using a two-button response pad (Fig. 1b).

Resting state task In total, 3 resting state scans were acquired in the present study on each of the test days. The first one preceded the functional localizer (Rest 1), the second one followed the functional localizer (Rest 2) and the third one followed the NF training task on day 1, or following performance of the task without feedback on day 3 (Rest 3). During the resting

scan subjects were asked to stay relaxed, close their eyes, and think of nothing in particular but without falling asleep.

Follow-up assessment Two days after the NF training subjects underwent a follow-up assessment to test whether the neural and behavioral effects of NF training could be maintained. The procedures were identical to the training sessions, except that during follow-up no feedback was provided and only two training sessions were included. Subjects were instructed to modulate the LAI activation according to the learned experience during the NF training task. The thermometer bars remained blank during the whole task.

Control group To control for non-specific effects of repeated scanning sessions, and other potential confounding factors, a control group was included. The control group underwent identical procedures and task instructions during training as well as follow-up assessment to the experimental group, except that subjects in the control group received sham feedback from a control ROI. In line with previous studies, the control ROI was a large region in a reference slice distant from the AI encompassing the whole brain to rule out global effects and to average out any unspecific activation (Fig. S1; Berman et al., 2013; Caria et al., 2007, 2010; Veit et al., 2012).


Image acquisition Images were collected using a 3 T, GE Discovery MR750 scanner (General Electric Medical System, Milwaukee, WI, USA). During each fMRI scan, a time series of volumes was acquired using a T2*-weighted echo planar imaging pulse sequence (repetition time, 2000 ms; echo time, 30 ms; slices, 32; thickness, 3.4 mm; gap, 0.6 mm; field of view, 220 × 220 mm; resolution, 64 × 64; flip angle, 90°). The localizer run consisted of 221 volumes. Each NF training session consisted of 303 volumes and each of the resting state scans consisted of 240 volumes. High-resolution whole-brain volume T1*-weighted images were acquired obliquely with a 3D spoiled gradient echo pulse sequence (repetition time, 6 ms; echo time, minimum; flip angle, 9°; field of view = 256 × 256 mm; acquisition matrix, 256 × 256; thickness, 1 mm; number of excitations, 2; 128 slices) to control for any anatomic abnormalities and increase normalization accuracy during pre-processing. Data analysis Online real-time analysis Functional images were analyzed online using Turbo Brain Voyager (TBV) 3.2 (Brain Innovation, Maastricht, The Netherlands). The functional images were collected and immediately transferred to the local hard disk of the TBV-installed computer. The TBV software conducted online preprocessing including 3D motion correction and drift removal. Then the BOLD signal of the target ROI was extracted and translated into visual graphs. Finally, the visual graphs were transferred to the local hard disk of the stimulus presentation system in real-time and were presented to subjects (Goebel, 2001; Weiskopf et al., 2007). Specifically, the individual ROI for each subject used in the training sessions was selected functionally in the localizer task by using the maximum activity within the LAI (MNI: mean ± standard deviation = − 37.18 ± 4.48, 13.65 ± 5.53, 0.82 ± 5.62) in the real-time activation maps based on the contrast “pain N neutral”. The ROI was placed on the LAI extending 3 axial slices using a t-value threshold of 2. During the training task, there were 15 volumes for each block. When calculating the values of baseline conditions, the first 3 volumes of the baseline block were excluded to avoid the impact of the hemodynamic delay from a previous regulation block. In the same vein, the first volume in the next block was included into the baseline condition, because the signal started to rise after about 2–3 s. Thus the feedback presented to subjects was calculated by computing the BOLD signal difference of the subject-specific ROI between the averaged last 3 volumes (the current volume plus the previous two) and the last 12 volumes of the previous baseline block plus the first following volume. The feedback value was calculated by the following equation: feedback value = (the current time points value − baseline value) / baseline ∗ 100. The feedback value was a percent signal value used to fill the thermometer display. Then the visual graphs transformed by TBV were presented to subjects in real-time using E-Prime 2.0. The delay due to data processing from image acquisition to feedback display was about 2 s. Subjects were informed of the data processing and the intrinsic hemodynamic response delay of 4–6 s. Offline analysis The SPM8 software package (Wellcome Department of Cognitive Neurology, London, UK, http://www.fil.ion.ucl.ac.uk/spm/spm8; Friston et al., 1994) was used for offline data processing. The first five images were excluded to achieve magnet-steady images and the remaining functional images were realigned to correct for head motion. After co-registering the mean functional image and the T1 image, the T1 image was segmented to determine the parameters for normalizing the functional images to Montreal Neurological Institute (MNI) space. Normalized images were spatially smoothed with a Gaussian kernel (8 mm full-width at half maximum, FWHM). The first level design matrix included 5 regressors (regulation, baseline, stimuli, pain rating, arousal rating) and the six head-

233

motion parameters convolved with the canonical hemodynamic response function (HRF). For the 1st-level analysis, session-specific contrast images between regulation and baseline (regulation N baseline) were created for each subject. Regions of interest (ROI) analysis To examine the training effect, the averaged BOLD signal change in the LAI during regulation and baseline was extracted for each session using MarsBar (Brett et al., 2002). Then the BOLD signal difference was calculated by subtracting the BOLD signal change value in the baseline condition from the regulation condition. The LAI ROI for both groups was defined as a 6 mm sphere using coordinates centered on the maximally activated voxel within the LAI during the localizer tasks on day 1. The same ROI coordinates were used in the follow-up assessment to ensure comparable activity being extracted. For statistical analysis a repeated-measures ANOVA was performed on BOLD signal change difference with session as a within-subject factor (sessions 1–4) and group (experimental vs. control) as a betweensubjects factor using SPSS (version 18, SPSS Inc., Chicago, IL, USA). Post-hoc tests were conducted to clarify differences contributing to significant interactions. Furthermore, a linear regression analysis on BOLD signal change difference was conducted to examine the linear learning effect over sessions. For the follow-up assessment a two-sample t-test on the averaged BOLD signal change difference across the 2 sessions was conducted between the two groups to test whether subjects in the experimental group maintained their ability to regulate AI activity. PPI analysis A psychophysiological interaction (PPI) analysis was conducted in the experimental group to clarify the neural mechanisms underlying AI regulation during NF training using the approach described in Veit et al. (2012). Briefly, the ROI was defined as a 6-mm sphere centered on the maximally activated voxel within the LAI. The LAI coordinate used was derived from a combination of sessions 3 and 4, in which subjects could regulate their AI activation more effectively. Next, the gppi toolbox (McLaren et al., 2012) was used to model psychophysiological interactions on the individual level. To examine the LAI functional interaction with other regions which was specific for effective AI regulation, we compared the late successful training sessions with the early ones (contrast: (session 3 + session 4) N (session 1 + session 2)) in the experimental group. Next on the second level a one-sample t-test was conducted based on these contrast images to assess training-related changes within the experimental group. Given the exploratory nature of this analyses an uncorrected threshold of p b 0.001 was used (see also Veit et al., 2012). Resting state analysis The fMRI images collected during resting scan were analyzed using the Data Processing Assistant for Resting-state fMRI (DPARSF) (Yan and Zang, 2010; http://www.restfmri.net). The first 5 volumes were excluded from the analysis. The BOLD time series pre-processing included slice timing correction, realignment, and spatial normalization. The covariates including the 6 head-motion parameters (using the Friston 24-parameter model), global signal, white matter, and cerebrospinal fluid were regressed out to rule out potential confounding artifacts (Fox et al., 2005). The normalized functional images were resampled using a 3 mm × 3 mm × 3 mm resolution. Next these images were spatially smoothed (6 mm FWHM), linearly detrended and filtered using a band pass filter of 0.01–0.08. A functional connectivity analysis between the LAI and other regions was conducted for the three resting scans to examine RSFC changes caused by the NF training. To avoid bias, the seed region was defined as a sphere with a 6 mm radius centered on the maximally activated voxel within the LAI during the localizing task across groups. The time series of each subject were obtained from the seed ROI separately for each resting scan. The functional connectivity map was generated

234


based on a voxel-wise correlation analysis between the ROI and other voxels in the whole brain, and then converted into a Z-map using Fisher's z transformation. Changes in the functional connectivity between the ROI and other regions associated with NF training were measured by subtracting the functional connectivity in Rest 2 before NF training from the functional connectivity after training in Rest 3. In comparison with Rest 1, subtracting Rest 2 from Rest 3 could better rule out the potential contamination caused by the localizing task. Comparisons between Rest 3 on day 1 and Rest 1 on day 3 in the experimental group were used to assess whether altered resting state connections on day 1 were maintained on day 3.

Results Personality trait and mood questionnaires Independent t-tests on questionnaire scores revealed no significant differences between the experimental and control groups on BIS/BAS, AQ, BDI, STAI, and EQ (Table 1). However, there was a marginally significant difference between the 2 groups in the PANAS-negative subscale (p = 0.053).

Effects of NF training ROI analysis Analysis of the training effect using a repeated-measures ANOVA on the percent BOLD signal change in the LAI with session as a withinsubject factor and group as a between-subject factor revealed a significant main effect of group (F(1, 30) = 4.54, p = 0.041), suggesting a better regulation effect across the 4 sessions on LAI activity in the experimental group relative to the control group. There was no significant main effect of session (F(3, 90) = 2.24, p = 0.107) or group × session interaction (F(3, 90) = 1.56, p = 0.215). To examine potential changes over sessions, an exploratory simple effect test on the interaction was conducted and revealed a significant increase of the LAI activity from session 1 to session 4 (p = 0.011 and p = 0.024 for session 1 and session 2 compared with session 4) in the experimental group, but not in the control group (all ps N 0.799; Fig. 2a). The linear regression analysis revealed a significant progressive increase from session 1 to session 4 in the experimental group (y = 0.022x + 0.023, t = 1.99, p = 0.05), but not in the control group (y = 0.002x + 0.021, t = 0.29, p = 0.774). Overall this indicates that subjects in the experimental group successfully learned to regulate their LAI activity over sessions. Similar but less robust results were obtained for the right AI (see Fig. S2 and Supplementary information).

Table 1 Ages and questionnaire scores in the experimental and control group (mean ± sem). Measurements

Experimental

Control

t-Value

p-Value

Age (years) Autism Spectrum Quotient Beck Depression Inventory Empathy Quotient Positive and Negative Affective Scale (PANAS) Positive Negative Behavioural Inhibition System Behavioural Activation System State–Trait Anxiety Inventory — State State–Trait Anxiety Inventory — Trait

22.2 ± 0.3 19.8 ± 1.5 6.2 ± 1.2 36.7 ± 2.0

21.7 ± 0.5 17.4 ± 1.2 7.9 ± 1.5 36.6 ± 2.4

0.99 1.23 −0.91 0.05

0.329 0.228 0.369 0.962

28.9 ± 1.8 17.1 ± 1.7 19.9 ± 0.6 38.9 ± 1.4

30.1 ± 2.0 22.3 ± 1.9 20.8 ± 0.8 41.5 ± 0.8

−0.47 −2.02 −0.84 −1.43

0.644 0.053 0.408 0.163

39.3 ± 1.6

43.0 ± 3.0

−1.15

0.261

40.7 ± 1.2

41.8 ± 1.9

−0.50

0.623

95% Confidence interval.

Whole brain analysis Group analysis with the contrast ‘regulation N baseline’ for each session confirmed the results of the ROI analysis. Subjects in the experimental group could learn to regulate LAI activity over sessions (sessions 1, 2 and 3: no LAI activation; session 4: Inferior Frontal Gyrus spreading into LAI, MNI = − 34, 30, − 2; t = 5.11; p b 0.01 FDR-corrected; Fig. 2b). Additional regions activated during training are displayed in Supplementary Table S1. In the control group there were no LAI or other brain region activations (p b 0.01 FDR-corrected). Behavioral effect associated with NF training To examine the functional relevance of changes caused by AI modulation, the mean rating difference to painful stimuli after regulation relative to baseline was calculated over sessions for each subject. To control for biasing effects of outliers, trials with rating scores extending mean ± 2 SD were excluded from the analysis. For pain empathy rating a repeated-measures ANOVA on rating differences with session as a within-subjects factor and group as a between-subjects factor revealed a significant main effect of group (F(1, 30) = 4.35, p = 0.046), indicating higher pain empathy rating changes in the experimental group compared with control group (see Fig. 3a). There was no significant main effect of session and no significant interaction between these two factors, all ps N 0.553. There was no significant effect on arousal ratings, all ps N 0.427. Since there was a marginally significant group difference in the PANAS-negative subscale we assessed the possibility of a potential confounding effect of negative mood on training success in terms of the neural regulation effect and enhanced empathic responses in the experimental group using correlation analyses. Correlations between the PANAS-negative subscale scores and the BOLD signal change difference between session 4 and session 1 and mean pain empathy rating difference across sessions were calculated but were not significant for either the mean BOLD signal change difference (Pearson r = −0.095, df = 32, p = 0.605) or the mean pain empathy rating scores (Pearson r = −0.279, df = 32, p = 0.122). These findings argue against strong confounding effects of negative mood on training-associated AI regulation and empathic ratings. Brain behavior associations To further explore the association between training-induced modulation of LAI activity and pain empathy ratings, a correlation analysis between the averaged difference of BOLD signal change and the averaged difference of pain ratings (regulation vs. baseline) was conducted. The correlation between the averaged BOLD signal change difference and pain rating score difference was significant only in the experimental group (Pearson r = 0.548, df = 18, p = 0.018), but not in the control group (Pearson r = −0.080, df = 14, p = 0.786). The correlation difference between groups was tested using the Fisher z-transformation test and revealed a marginally significant difference between the experimental group and the control group (Fishers z-score = − 1.753, p = 0.080). These results indicate a trend towards increased AI activity being associated with increased pain empathy ratings in the experimental group (Fig. 3b). Maintenance of NF training effects ROI analysis An independent samples t-test on BOLD signal change of LAI showed a significant difference between the experimental and control groups (t(30) = 2.95, p = 0.006), indicating that subjects in the experimental group maintained their ability to regulate the AI (Fig. 4a). Furthermore, a comparison between the mean BOLD signal changes of session 3 and session 4 on day 1 and the mean BOLD signal changes on day 3 for the experimental group was conducted to assess any attenuation of the


235

Fig. 2. (a) LAI BOLD signal changes during regulation relative to baseline conditions across the 4 training sessions. EX for the experimental group and SH for the sham control group. (b) LAI activation in a whole brain analysis across 4 training sessions in the experimental group.

training effect. A paired samples t-test revealed no significant changes from day 1 to day 3 in the LAI (t(17) = 1.28, p = 0.218; Fig. 4b). Thus there was no evidence for a reduction in the subjects' ability to regulate their AI on day 3. Whole brain analysis Group analysis was conducted across the two sessions separately for the experimental and control groups to examine the training effect of AI regulation. In the experimental group the contrast between regulation and baseline yielded activations in thalamus (MNI = − 12, − 12, 4; t = 6.10; p b 0.005, FDR-corrected) spreading into LAI (see Fig. 4c), indicating that subjects in the experimental group maintained their ability to regulate the LAI. Additional regions activated in the experimental group are displayed in Supplementary Table S2. In contrast, there was no LAI or other regional activation in the control group.

empathy would also be maintained during the follow-up assessment. An independent sample t-test on rating differences between regulation and control sessions showed no significant effect either on pain empathy rating or arousal rating, all ps N 0.528, indicating an absence of a behavioral effect on day 3. Furthermore, to examine rating changes from day 1 to day 3, we conducted a repeated ANOVA on averaged rating scores across sessions with task (regulation vs. baseline) and time (day 1 vs. day 3) as within-subject factors and group (experimental vs. control) as a between-subject factor. There was only a significant main effect of time (F(1, 30) = 26.50, p b 0.001), but no other significant main or interaction effect (all ps N 0.175), suggesting an unspecific pain empathy increase across groups from day 1 to day 3 (day 1: 5.56 ± 0.87 vs. day 3: 5.88 ± 0.84). Changes in functional connectivity associated with the acquisition of AI regulation

Maintenance effects on pain empathy The mean rating difference to painful stimuli was calculated across sessions for each subject to test whether the training effects on pain

PPI analysis To clarify the mechanisms underlying AI regulation during training we conducted an exploratory PPI analysis on data obtained

Fig. 3. (a) Pain empathy rating scores averaged across the 4 training sessions for the experimental and control group on day 1. To further clarify that the empathy rating differences between groups were mainly driven by enhancement in experimental group rather than decrease in the control group, we conducted a repeated ANOVA on averaged rating scores across sessions with task (regulation vs. baseline) as a within-subject factor and group (experimental vs. control) as a between-subject factor. Results showed no significant main effect of group (F(1, 30) = 0.81, p = 0.374), implying similar rating scores between groups. There was a significant interaction between these two factors (F(1, 30) = 4.35, p = 0.046). The simple effect test revealed a significant difference between regulation and baseline task only in the experimental group (5.81 ± 0.62 vs. 5.36 ± 1.24; p = 0.028) but not in the control group (5.56 ± 0.69 vs. 5.45 ± 1.06; p = 0.466). These results suggested that the behavioral effect found in the present study was mainly driven by the empathic enhancement due to NF training rather empathic decrease in the control group. (b) A significant positive correlation between the LAI activity and the pain empathy rating differences in the experimental group only (Pearson r = 0.548, df = 18, p = 0.018).

236


Fig. 4. (a) The LAI BOLD signal changes during regulation relative to baseline conditions between the experimental and control groups in the follow-up assessment (session 1 + session 2). (b) The LAI BOLD signal changes comparison between effective regulation sessions on day 1 (session 3 + session 4) and follow-up assessment sessions on day 3 (session 1 + session 2). (c) LAI activation in a whole brain analysis during the follow-up assessment in the experimental group. The left panel is the experimental group and the right penal is the control group.

during day 1. In the experimental group, PPI analysis for ‘regulation vs. baseline’ was conducted for each session separately. A paired ttest comparing the effective regulation sessions (session 3 and 4) and the ineffective sessions (sessions 1 and 2) revealed

strengthened connectivity between LAI and bilateral middle frontal gyrus (MFG), inferior parietal lobule (IPL), left dorsal medial prefrontal cortex (DMPFC), and right superior frontal gyrus (SFG, Fig. 5a and Table 2).

Fig. 5. (a) Regions showed increased functional connectivity with LAI during effective regulation (contrast: ‘session 3 + session 4’ N ‘session 1 + session 2’) revealed by a PPI analysis. (b) Regions showed increased RSFC with LAI after training on day 1. (c) Regions showed decreased RSFC with LAI after training on day 1.

S. Yao et al. / NeuroImage 130 (2016) 230–240 Table 2 Regions which showed increased functional connectivity with LAI during effective regulation (MNI coordinates). Brain region

BA

No. voxels

Peak t-value

x

y

z

L. medial frontal gyrus Middle frontal gyrus R. middle frontal gyrus Superior frontal gyrus R. Middle frontal gyrus Superior frontal gyrus R. middle cingulate gyrus R. lentiform nucleus L. inferior parietal lobule R. inferior parietal lobule Superior temporal gyrus Supramarginal gyrus R. superior parietal lobule L. precentral gyrus Postcentral gyrus L. precentral gyrus R. fusiform gyrus R. thalamus

6

352

9/8

603

5.93 5.46 7.15 4.62 5.91 4.27 4.47 4.45 4.39 5.26 5.22 4.06 5.10 4.96 4.36 4.96 5.64 5.21

−12 −28 46 28 24 12 10 26 −34 52 54 46 36 −44 −48 −34 30 8

−2 −8 36 46 2 4 −20 0 −66 −46 −48 −46 −66 −12 −18 −26 −68 −22

58 58 32 34 68 70 26 6 42 42 20 30 48 52 56 58 −10 6

6

81

23 39 40

69 21 41 254

7 4

221 111

4 19

119 316 55

All with an uncorrected p b 0.001 threshold. L indicates left; R indicates right.

Resting state analysis RSFC changes after NF training on day 1 were measured using a twosample t-test between the experimental and control groups. In the experimental group increased connectivity was found between LAI and bilateral IPL, right AI, bilateral superior temporal gyrus (STG), left middle temporal gyrus and bilateral postcentral gyrus (Fig. 5b and Table 3). Decreased functional connectivity after NF training was found with the bilateral SFG, right DMPFC, bilateral posterior cingulate cortex (PCC) and left IPL (p b 0.005, FDR corrected, Fig. 5c and Table 3).

Table 3 Regions which showed increased and decreased RSFC after training on day 1 (MNI coordinates). Brain region Experimental N Control L. superior frontal gyrus R. superior frontal gyrus L. middle frontal gyrus R. paracentral lobule L. postcentral gyrus Inferior parietal lobule R. postcentral gyrus R. inferior parietal lobule L. posterior cingulate cortex L. superior temporal gyrus L. superior temporal gyrus R. superior temporal gyrus Anterior insula L. middle temporal gyrus Middle occipital gyrus R. transverse temporal gyrus L. cuneus Control N Experimental L. superior frontal gyrus R. superior frontal gyrus Middle frontal gyrus Medial frontal gyrus R. anterior insula L. inferior parietal lobule L. posterior cingulate cortex Posterior cingulate cortex R. superior parietal lobule Follow-up assessment L. precentral gyrus

Peak t-value

x

y

z

6.00 5.97 5.75 8.44 5.65 5.36 7.11 5.67 6.39 6.05 5.75 6.93 5.41 5.72 4.75 5.70 6.34

−9 6 −45 9 −30 −36 30 66 −15 −51 −57 51 45 −45 −39 66 −15

12 15 3 −18 −48 −39 −42 −36 −27 12 −48 18 6 −69 −81 −12 −81

60 66 51 45 63 51 66 39 42 −9 15 −9 −6 6 3 12 24

213

4.99 7.13 5.26 4.51 5.89 5.54 8.01 7.86 7.88

−21 21 24 3 33 −42 −9 9 36

66 36 21 42 18 −72 −48 −48 −69

6 45 42 36 15 42 15 27 45

33

9.01

−24

−24

69

BA

No. voxels

6 6 6 6/31 2/5

30 20 25 126 65

5/40 12 31 38 22/47

109 40 19 15 22 56

19

58

42 7/18

18 54

10 8

10 209

13 39 23/30

42 46 436

7/40

4

All with a FDR p b 0.005 corrected threshold. L indicates left; R indicates right.

237

For resting state connectivity changes during the follow-up assessment, comparisons between the final third resting state scan on day 1 and the first on day 3 in the experimental group revealed no significant differences using a p b 0.005 FDR correction. Thus it would appear that alterations in resting state functional connections following NF training on day 1 were maintained at the beginning of day 3. Comparisons between the second and third resting state scans on day 3 revealed that the voluntary modulation of LAI in the two transfer sessions led to strengthened connectivity of LAI only with the left precentral gyrus (p b 0.005, FDR corrected; Table 3). Discussion The present study combined a validated rtfMRI-based NF training for the acquisition of volitional AI regulation with a sham-controlled between-subject design to evaluate effects on emotional behavior and the maintenance of the neural and behavioral training effects. The successful acquisition of AI up-regulation was accompanied by increased empathic responses to painful stimuli without effects on arousal. Two days after the training session subjects were able to maintain AI control without NF although no further increments in pain empathy ratings were seen. On the other hand, AI RSFC changes which were found to occur after successful NF training on day 1 were also maintained during follow-up assessment two days later. In line with previous studies, participants in the present study successfully acquired volitional control over the AI during four NF training sessions (Berman et al., 2013; Caria et al., 2007, 2010; Veit et al., 2012). The linear increase of LAI activity across the four consecutive training sessions and the specificity of the AI changes to the regulation condition in the experimental group indicated that the changes in AI activity were contingent upon accurate NF provided to the experimental group. In line with previous studies (Caria et al., 2007, 2010), a similar but less strong increase was found in the contralateral (right) AI of the experimental group (results see Supplementary information). However, changes in the right AI did not differ significantly between groups, suggesting that changes in the right AI were less robust. This might be since the right AI was not the source of NF and thus there was no direct association between subjects' regulation effort and the right AI activity during training. On the behavioral level, successful AI regulation was accompanied by stronger empathic responses, as evidenced by higher pain empathy ratings in the experimental relative to the sham control group and a significant brain–behavior association within the experimental group. These findings suggested that successful AI regulation might have an impact on empathic responses. However, the marginally significant brain–behavior correlation difference between groups and the absence of a significant increase of pain empathy ratings across the training sessions suggests that NF-specific effects on the behavioral level might be less robust than those at the neural one, possibly due to habituation effects on experienced pain empathy over the course of repeated training sessions. Inferences regarding the effects of LAI regulation on pain empathic processing should therefore be made with caution. In contrast no effects of successful AI regulation on behavioral indices of arousal were observed. The specificity of the behavioral effects accompanying AI up-regulation on pain empathy corroborates previous findings that suggest a selective and critical role of the AI in pain empathy (Saarela et al., 2007; Berntson et al., 2011; Lamm et al., 2011), but a more selective role of the amygdala in affective arousal (e.g., Berntson et al., 2011). The absence of significant effect on arousal was also found in previous studies (Caria et al., 2010; Lawrence et al., 2014). Future investigations are necessary to further examine specific associations between NF training associated volitional regulation of brain activity and effects on the behavioral level. Two days after the initial training subjects in the experimental group demonstrated significant AI increases without NF information, indicating a successful transfer of regulation from the feedback-aided training. Differences between volitional up-regulation and baseline at follow-up assessment were comparable to those observed during successful

238


regulation sessions (sessions 3 and 4) of the NF training, suggesting a robust consolidation of the learned self-regulation. Previous studies targeting emotional brain regions, including the AI, either only assessed regulation success while subjects still received feedback (Caria et al., 2010; Veit et al., 2012), or did not observe successful AI self-regulation immediately after the NF training when subjects no longer received feedback (Caria et al., 2007; Berman et al., 2013). The present findings thus provide the first evidence that AI regulation acquired during a single NF training can be maintained for a period up to two days without further intervention. While this successful regulation of AI activity after two days produced no significant effects on pain empathy ratings, this might have been the result of a general learning effect, as suggested by the finding that pain empathy increased in both groups between NF training and follow-up assessment. However, given the relatively short period of initial NF training in the current study it is possible that after longer periods of training more robust behavioral changes in the absence of further NF training might occur. The implementation of optimized training protocols, repeated feedback training, or the implementation of post-scanner training schedules in future studies might therefore help to maintain behavioral effects. At the network level, the acquisition of successful AI regulation across training sessions was accompanied by increased coupling of the LAI with prefrontal regions, particularly the bilateral dorsal lateral prefrontal cortex (DLPFC) and DMPFC. These AI-prefrontal circuits play a central role in emotional processing, particularly salience detection (Uddin, 2015) and emotional awareness (Gu et al., 2013), and promote flexible behavioral adaption, particularly via emotion regulation (Ochsner et al., 2002; Ochsner and Gross, 2004, 2005) and switching central processing resources to salient external stimuli (Uddin, 2015). Additional regions showing increased coupling with AI were also found in bilateral IPL and thalamus. The functional interplay of these regions has been specifically related to pain empathy processing (Engen and Singer, 2013; Lamm et al., 2011) and fine tuning of emotional regulation (Bush et al., 2000; Ochsner and Gross, 2004, 2005), suggesting that the feedback training not only affected regional processing in the AI but also changed the interaction of the central hubs in the relevant task networks. Accumulating evidence suggests that dysfunctional interactions within these networks is associated with a broad range of emotional processing deficits across psychiatric categories, including autism, depression, and schizophrenia (Hamilton et al., 2012; Uddin, 2015). These findings are therefore of particular clinical relevance in view of the implication that NF training can modify the interplay within the emotional brain networks. In the training group, modifications of resting state functional networks were observed immediately after the NF training and were maintained at the follow-up assessment after 2 days, suggesting that the NF training induced lasting reconfigurations of brain network interactions. Immediately after the training, subjects in the experimental group demonstrated increased functional connectivity between the LAI and regions within the networks subserving empathic processing (Engen and Singer, 2013; Lamm et al., 2011; Singer et al., 2009; Preston and de Waal, 2002), specifically the right AI, bilateral IPL relative to the control group. In line with the proposed central function of the AI in switching between task-oriented and default mode networks (Uddin, 2015), the AI showed stronger negative coupling with core regions of the default mode networks, including the bilateral PCC, DMPFC and dorsolateral SFG. Given that impaired empathy represents a core feature of ASD (Lai et al., 2014) and the strengths of these impairments has been related to the degree of aberrant intrinsic connectivity (Jung et al., 2014), the present findings suggest that NF training might help to normalize emotional processing in ASD. Emotional dysfunctions and aberrant AI functioning have been frequently observed across several clinical samples and hypothesized to contribute to the development and maintenance of the pathological states, including anxiety disorders, autism and depression (Shah et al., 2009; Surguladze et al., 2010; Silani et al., 2008). The present findings that both AI-activity and emotional behavior could be modified during

a single NF-aided training suggest that similar training approaches could successfully facilitate the normalization of emotional processing in these disorders. In addition, the maintenance of training success and modifications of resting state functional networks for two days after the NF feedback training seems promising with respect to the clinical implementation and costs of fMRI-aided feedback. Findings on dynamic network changes suggest that future training protocols that enable the regulation on the network level might be particularly promising. Furthermore, previous studies suggest that behavioral training approaches, such as cognitive and emotion regulation, can be employed to successfully modulate emotional processing on the neural and behavioral levels (Butler et al., 2006; Ochsner and Gross, 2005; Smith et al., 2009). Future studies might consider combining training strategies from evaluated behavioral training approaches with rtfMRI-based NF to enhance therapeutic effects. There were several limitations in the present study. First, there was a marginally significant group difference on the negative mood subscale that might confound the current findings. The lack of significant associations between negative mood and training success in terms of neural regulation and pain empathic responses argues against strong confounding effects. Nevertheless, it is possible that the more negative mood in the controls might have facilitated, rather than impaired, the application of the instruction to recall negative events during training to acquire LAI control and so we cannot completely rule out confounding effects of between-group mood differences. Furthermore, similar to some previous rtfMRI studies addressing the behavioral effects of AI regulation (Berman et al., 2013; Lawrence et al., 2014) our behavioral effect of NF training in the present study was less robust than at the neural level. Although, the observed between group differences in empathy ratings across training sessions suggest a general enhancement of empathy response in the training group, no linear increase across training sessions was observed. This might reflect less robust effects on the behavioral level and might be related to habituation effects of repeated training on perceived pain empathy or that optimized training protocols are needed to reveal a robust linear increase across training sessions. Finally, although the present study used sample sizes that were comparable to other recent studies evaluating rtfMRI NF training approaches (Emmert et al., 2016), larger sample sizes would have allowed evaluation of training effects with a higher statistical power. This limitation particularly refers to the smaller number of subjects in the control group which might have contributed to less robust effects of training observed in analyses incorporating between-group comparisons. Despite these limitations, findings from the present proof-of-concept study provide further support for the neurotherapeutic potential of fMRI-aided self-regulation on emotional brain function. Of particular importance for the clinical potential are: (1) behavioral effects that were specifically observed under conditions of contingent feedback and successful AI-up regulation, suggesting a functional relevance of AI selfregulation, (2) the maintenance of neural self-regulation two days after the NF feedback training, suggesting a successful transfer and consolidation of the training success and (3) lasting effects at the network level, suggesting that NF training is capable of reconfiguring interactions in emotional brain circuitry. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.neuroimage.2016.02.035.

Acknowledgments This study was supported by the National Natural Science Foundation of China (NSFC) grant (grant number 31530032). We thank Yong Zhang and Jianfu Li for their technical support. Competing financial interests statement The authors declare no competing financial interests.


References Adolphs, R., 2003. Cognitive neuroscience of human social behaviour. Nat. Rev. Neurosci. 4, 165–178. Baron-Cohen, S., Wheelwright, S., 2004. The empathy quotient: an investigation of adults with Asperger syndrome or high functioning autism, and normal sex differences. J. Autism Dev. Disord. 34, 163–175. Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., Clubley, E., 2001. The autism-spectrum quotient (AQ): evidence from Asperger syndrome/highfunctioning autism, males and females, scientists and mathematicians. J. Autism Dev. Disord. 31, 5–17. Beck, A.T., Steer, R.A., Brown, G.K., 1996. Manual for the Beck Depression Inventory-II. Psychological Corporation, San Antonio. Berman, B.D., Horovitz, S.G., Hallett, M., 2013. Modulation of functionally localized right insular cortex activity using real-time fMRI-based neurofeedback. Front. Hum. Neurosci. 7, 638. Bernhardt, B.C., Singer, T., 2012. The neural basis of empathy. Annu. Rev. Neurosci. 35, 1–23. Berntson, G.G., Norman, G.J., Bechara, A., Bruss, J., Tranel, D., Cacioppo, J.T., 2011. The insula and evaluative processes. Psychol. Sci. 22, 80–86. Botteron, K.C., Carter, C., Castellanos, F.X., Dickstein, D.P., Drevets, W., Kim, K.L., Pescosolido, M.F., Rausch, S., Seymour, K.E., Sheline, Y., Zubieta, J.K., 2012. Consensus Report of the APA Work Group on Neuroimaging Markers of Psychiatric Disorders. American Psychiatric Association. Bradley, M.M., Lang, P.J., 1994. Measuring emotion: the self-assessment manikin and the semantic differential. J. Behav. Ther. Exp. Psychiatry 25, 49–59. Brett, M., Anton, J.L., Valabregue, R., Poline, J.B., 2002. Region of interest analysis using the MarsBar toolbox for SPM 99. NeuroImage 16, S497. Brühl, A.B., Scherpiet, S., Sulzer, J., Stämpfli, P., Seifritz, E., Herwig, U., 2014. Real-time neurofeedback using functional MRI could improve down-regulation of amygdala activity during emotional stimulation: a proof-of-concept study. Brain Topogr. 27, 138–148. Bush, G., Luu, P., Posner, M.I., 2000. Cognitive and emotional influences in anterior cingulate cortex. Trends Cogn. Sci. 4, 215–222. Butler, A.C., Chapman, J.E., Forman, E.M., Beck, A.T., 2006. The empirical status of cognitive–behavioral therapy: a review of meta-analyses. Clin. Psychol. Rev. 26, 17–31. Caria, A., Veit, R., Sitaram, R., Lotze, M., Weiskopf, N., Grodd, W., Birbaumer, N., 2007. Regulation of anterior insular cortex activity using real-time fMRI. NeuroImage 35, 1238–1246. Caria, A., Sitaram, R., Veit, R., Begliomini, C., Birbaumer, N., 2010. Volitional control of insula activity modulates the response to aversive stimuli. A real-time functional magnetic resonance study. Biol. Psychiatry 68, 425–432. Carver, C.S., White, T.L., 1994. Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: the BIS/BAS scales. J. Pers. Soc. Psychol. 67, 319–333. Decety, J., Jackson, P.L., 2006. A social–neuroscience perspective on empathy. Curr. Dir. Psychol. Sci. 15, 54–58. deCharms, R.C., Maeda, F., Glover, G.H., Ludlow, D., Pauly, J.M., Soneji, D., Gabrieli, J.D.E., Mackey, S.C., 2005. Control over brain activation and pain learned by using real-time functional MRI. Proc. Natl. Acad. Sci. U. S. A. 102, 18626–18631. Derntl, B., Finkelmeyer, A., Toygar, T.K., Hülsmann, A., Schneider, F., Falkenberg, D.I., Habel, U., 2009. Generalized deficit in all core components of empathy in schizophrenia. Schizophr. Res. 108, 197–206. Di Martino, A., Ross, K., Uddin, L.Q., Sklar, A.B., Castellanos, F.X., Milham, M.P., 2009. Functional brain correlates of social and nonsocial processes in autism spectrum disorders: an activation likelihood estimation meta-analysis. Biol. Psychiatry 65, 63–74. Emmert, K., Kopel, R., Sulzer, J., Brühl, A.B., Berman, B.D., Linden, D.E., ... Johnston, S., 2016. Meta-analysis of real-time fMRI neurofeedback studies using individual participant data: how is brain regulation mediated? NeuroImage 124, 806–812. Engen, H.G., Singer, T., 2013. Empathy circuits. Curr. Opin. Neurobiol. 23, 275–285. Etkin, A., Wager, T.D., 2007. Functional neuroimaging of anxiety: a meta-analysis of emotional processing in PTSD, social anxiety disorder, and specific phobia. Am. J. Psychiatry 164, 1476–1488. Fox, M.D., Snyder, A.Z., Vincent, J.L., Corbetta, M., Van Essen, D.C., Raichle, M.E., 2005. The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proc. Natl. Acad. Sci. U. S. A. 102, 9673. Friston, K.J., Holmes, A.P., Worsley, K.J., Poline, J.P., Frith, C.D., Frackowiak, R.S., 1994. Statistical parametric maps in functional imaging: a general linear approach. Hum. Brain Mapp. 2, 189–210. Goebel, R., 2001. Cortex-based real-time fMRI. NeuroImage 13, S129. Gu, X., Liu, X., Guise, K.G., Naidich, T.P., Hof, P.R., Fan, J., 2010. Functional dissociation of the frontoinsular and anterior cingulate cortices in empathy for pain. J. Neurosci. 30, 3739–3744. Gu, X., Gao, Z., Wang, X., Liu, X., Knight, R.T., Hof, P.R., Fan, J., 2012. Anterior insular cortex is necessary for empathetic pain perception. Brain 135, 2726–2735. Gu, X., Hof, P.R., Friston, K.J., Fan, J., 2013. Anterior insular cortex and emotional awareness. J. Comp. Neurol. 521, 3371–3388. Hamilton, J.P., Etkin, A., Furman, D.J., Lemus, M.G., Johnson, R.F., Gotlib, I.H., 2012. Functional neuroimaging of major depressive disorder: a meta-analysis and new integration of base line activation and neural response data. Am. J. Psychiatry 169, 693–703. Hampson, M., Scheinost, D., Qiu, M., Bhawnani, J., Lacadie, C., Leckman, J.F., 2011. Biofeedback from the supplementary motor area reduces functional connectivity to subcortical regions. Brain Connect. 1, 91–98.

239

Hawkinson, J.E., Ross, A.J., Parthasarathy, S., Scott, D.J., Laramee, E.A., Posecion, L.J., Rekshan, W.R., Sheau, K.E., Njaka, N.D., Bayley, P.J., deCharms, R.C., 2012. Quantification of adverse events associated with functional MRI scanning and with real-time fMRI-based training. Int. J. Behav. Med. 19, 372–381. Jackson, P.L., Meltzoff, A.N., Decety, J., 2005. How do we perceive the pain of others? A window into the neural processes involved in empathy. NeuroImage 24, 771–779. Jackson, P.L., Rainville, P., Decety, J., 2006. To what extent do we share the pain of others? Insight from the neural bases of pain empathy. Pain 125, 5–9. Johnston, S.J., Boehm, S.G., Healy, D., Goebel, R., Linden, D.E.J., 2010. Neurofeedback: a promising tool for the self-regulation of emotion networks. NeuroImage 49, 1066–1072. Jones, A.P., Happé, F.G., Gilbert, F., Burnett, S., Viding, E., 2010. Feeling, caring, knowing: different types of empathy deficit in boys with psychopathic tendencies and autism spectrum disorder. J. Child Psychol. Psychiatry 51, 1188–1197. Jung, M., Kosaka, H., Saito, D.N., Ishitobi, M., Morita, T., Inohara, K., Asano, M., Arai, S., Munesue, T., Tomoda, A., Wada, Y., Sadato, N., Okazawa, H., Iidaka, T., 2014. Default mode network in young male adults with autism spectrum disorder: relationship with autism spectrum traits. Mol. Autism 5, 35. Koban, L., Pourtois, G., 2014. Brain systems underlying the affective and social monitoring of actions: an integrative review. Neurosci. Biobehav. Rev. 46, 71–84. Lai, M.C., Lombardo, M.V., Baron-Cohen, S., 2014. Autism. Lancet 383, 896–910. Lamm, C., Batson, C.D., Decety, J., 2007. The neural substrate of human empathy: effects of perspective-taking and cognitive appraisal. J. Cogn. Neurosci. 19, 42–58. Lamm, C., Decety, J., Singer, T., 2011. Meta-analytic evidence for common and distinct neural networks associated with directly experienced pain and empathy for pain. NeuroImage 54, 2492–2502. Lang, P.J., Bradley, M.M., Cuthbert, B.N., 2005. International Affective Picture System (IAPS): affective ratings of pictures and instruction manual. Technical Report A-6. Center for Research in Psychophysiology. University of Florida, Gainesville. Lawrence, E.J., Su, L., Barker, G.J., Medford, N., Dalton, J., Williams, S.C., Birbaumer, N., Veit, R., Ranganatha, S., Bodurka, J., Brammer, M., Giampietro, V., David, A.S., 2014. Selfregulation of the anterior insula: reinforcement learning using real-time fMRI neurofeedback. NeuroImage 88, 113–124. Linden, D.E., 2014. Neurofeedback and networks of depression. Dialogues Clin. Neurosci. 16, 103–112. Linden, D., Habes, I., Johnston, S., Linden, S., Tatineni, R., Subramanian, L., Sorger, B., Healy, D., Goebel, R., 2012. Real-time self-regulation of emotion networks in patients with depression. PLoS One 7, e38115. McLaren, D.G., Ries, M.L., Xu, G., Johnson, S.C., 2012. A generalized form of contextdependent psychophysiological interactions (gPPI): a comparison to standard approaches. NeuroImage 61, 1277–1286. Meng, J., Hu, L., Shen, L., Yang, Z., Chen, H., Huang, X., Jackson, T., 2012. Emotional primes modulate the responses to others' pain: an ERP study. Exp. Brain Res. 220, 277–286. Ochsner, K.N., Gross, J.J., 2004. Thinking makes it so: a social cognitive neuroscience approach to emotion regulation. In: Baumeister, R.F., Vohs, K.D. (Eds.), Handbook of Self-regulation: Research, Theory, and Applications. Guilford Press, New York, pp. 229–255. Ochsner, K.N., Gross, J.J., 2005. The cognitive control of emotion. Trends Cogn. Sci. 9, 242–249. Ochsner, K.N., Bunge, S.A., Gross, J.J., Gabrieli, J.D., 2002. Rethinking feelings: an FMRI study of the cognitive regulation of emotion. J. Cogn. Neurosci. 14, 1215–1229. Paret, C., Kluetsch, R., Ruf, M., Demirakca, T., Hoesterey, S., Ende, G., Schmahl, C., 2014. Down-regulation of amygdala activation with real-time fMRI neurofeedback in a healthy female sample. Front. Behav. Neurosci. 8, 299. Posse, S., Fitzgerald, D., Gao, K., Habel, U., Rosenberg, D., Moore, G.J., Schneider, F., 2003. Real-time fMRI of temporolimbic regions detects amygdala activation during single-trial self-induced sadness. NeuroImage 18, 760–768. Preston, S.D., de Waal, F.B., 2002. Empathy: its ultimate and proximate bases. Behav. Brain Sci. 25, 1–20. Rossi, S., Hallett, M., Rossini, P.M., Pascual-Leone, A., 2009. Safety, ethical considerations, and application guidelines for the use of transcranial magnetic stimulation in clinical practice and research. Clin. Neurophysiol. 120, 2008–2039. Rota, G., Sitaram, R., Veit, R., Erb, M., Weiskopf, N., Dogil, G., Birbaumer, N., 2008. Selfregulation of regional cortical activity using realtime fMRI: the right inferior frontal gyrus and linguistic processing. Hum. Brain Mapp. 30, 1605–1614. Ruiz, S., Lee, S., Soekadar, S.R., Caria, A., Veit, R., Kircher, T., Birbaumer, N., Sitaram, R., 2013. Acquired self-control of insula cortex modulates emotion recognition and brain network connectivity in schizophrenia. Hum. Brain Mapp. 34, 200–212. Saarela, M.V., Hlushchuk, Y., Williams, A.C., Schurmann, M., Kalso, E., Hari, R., 2007. The compassionate brain: humans detect intensity of pain from another's face. Cereb. Cortex 17, 230–237. Scheinost, D., Stoica, T., Saksa, J., Papademetris, X., Constable, R.T., Pittenger, C., Hampson, M., 2013. Orbitofrontal cortex neurofeedback produces lasting changes in contamination anxiety and resting-state connectivity. Transl. Psychiatry 3, e250. Shah, S.G., Klumpp, H., Angstadt, M., Nathan, P.J., Phan, K.L., 2009. Amygdala and insula response to emotional images in patients with generalized social anxiety disorder. J. Psychiatry Neurosci. 34, 296–302. Silani, G., Bird, G., Brindley, R., Singer, T., Frith, C., Frith, U., 2008. Levels of emotional awareness and autism: an fMRI study. Soc. Neurosci. 3, 97–112. Simmons, W.K., Avery, J.A., Barcalow, J.C., Bodurka, J., Drevets, W.C., Bellgowan, P., 2013. Keeping the body in mind: insula functional organization and functional connectivity integrate interoceptive, exteroceptive, and emotional awareness. Hum. Brain Mapp. 34, 2944–2958. Singer, T., 2006. The neuronal basis and ontogeny of empathy and mind reading: review of literature and implications for future research. Neurosci. Biobehav. Rev. 30, 855–863.

240


Singer, T., Seymour, B., O'Doherty, J., Kaube, H., Dolan, R.J., Frith, C.D., 2004. Empathy for pain involves the affective but not sensory components of pain. Science 303, 1157–1162. Singer, T., Critchley, H.D., Preuschoff, K., 2009. A common role of insula in feelings, empathy and uncertainty. Trends Cogn. Sci. 13, 334–340. Smith, G.E., Housen, P., Yaffe, K., Ruff, R., Kennison, R.F., Mahncke, H.W., Zelinski, E.M., 2009. A cognitive training program based on principles of brain plasticity: results from the Improvement in Memory with Plasticity-based Adaptive Cognitive Training (IMPACT) Study. J. Am. Geriatr. Soc. 57, 594–603. Spielberger, C.D., Gorsuch, R.L., Lushene, R., Vagg, P.R., Jacobs, G.A., 1983. Manual for the State–Trait Anxiety Inventory. Consulting Psychologists Press, Palo Alto (CA). Stoeckel, L.E., Garrison, K.A., Ghosh, S., Wighton, P., Hanlon, C.A., Gilman, J.M., Greer, S., Turk-Browne, N.B., deBettencourt, M.T., Scheinost, D., Craddock, C., Thompson, T., Calderon, V., Bauer, C.C., George, M., Breiter, H.C., Whitfield-Gabrieli, S., Gabrieli, J.D., LaConte, S.M., Hirshberg, L., Brewer, J.A., Hampson, M., Van Der, K.A., Mackey, S., Evins, A.E., 2014. Optimizing real time fMRI neurofeedback for therapeutic discovery and development. Neuroimage Clin. 5, 245–255. Sulzer, J., Sitaram, R., Blefari, M.L., Kollias, S., Birbaumer, N., Stephan, K.E., Luft, A., Gassert, R., 2013. Neurofeedback-mediated self-regulation of the dopaminergic midbrain. NeuroImage 75C, 176–184. Surguladze, S.A., El-Hage, W., Dalgleish, T., Radua, J., Gohier, B., Phillips, M.L., 2010. Depression is associated with increased sensitivity to signals of disgust: a functional magnetic resonance imaging study. J. Psychiatr. Res. 44, 894–902. Uddin, L.Q., 2015. Salience processing and insular cortical function and dysfunction. Nat. Rev. Neurosci. 16, 55–61. Veit, R., Singh, V., Sitaram, R., Caria, A., Rauss, K., Birbaumer, N., 2012. Using real-time fMRI to learn voluntary regulation of the anterior insula in the presence of threat-related stimuli. Soc. Cogn. Affect. Neurosci. 7, 623–634.

Wager, T.D., Phan, K.L., Liberzon, I., Taylor, S.F., 2003. Valence, gender, and lateralization of functional brain anatomy in emotion: a meta-analysis of findings from neuroimaging. NeuroImage 19, 513–531. Watson, D., Clark, L., Tellegen, A., 1988. Development and validation of brief measures of positive and negative affect — the PANAS scales. J. Pers. Soc. Psychol. 54, 1063–1070. Weiskopf, N., 2012. Real-time fMRI and its application to neurofeedback. NeuroImage 62, 682–692. Weiskopf, N., Veit, R., Erb, M., Mathiak, K., Grodd, W., Goebel, R., Birbaumer, N., 2003. Physiological self-regulation of regional brain activity using real-time functional magnetic resonance imaging (fMRI): methodology and exemplary data. NeuroImage 19, 577–586. Weiskopf, N., Scharnowski, F., Veit, R., Goebel, R., Birbaumer, N., Mathiak, K., 2004. Selfregulation of local brain activity using real-time functional magnetic resonance imaging. J. Physiol. Paris 98, 357–373. Weiskopf, N., Sitaram, R., Josephs, O., Veit, R., Scharnowski, F., Goebel, R., Birbaumer, N., Deichmann, R., Mathiak, K., 2007. Real-time functional magnetic resonance imaging: methods and applications. Magn. Reson. Imaging 25, 989–1003. Yan, C., Zang, Y., 2010. DPARSF: a MATLAB toolbox for “pipeline” data analysis of restingstate fMRI. Front. Syst. Neurosci. 4, 13. Young, K.D., Zotev, V., Phillips, R., Misaki, M., Yuan, H., Drevets, W.C., Bodurka, J., 2014. Real-time fMRI neurofeedback training of amygdala activity in patients with major depressive disorder. PLoS One 9, e88785. Zotev, V., Krueger, F., Phillips, R., Alvarez, R.P., Simmons, W.K., Bellgowan, P., Wayne, C., Drevets, W.C., Bodurka, J., 2011. Self-regulation of amygdala activation using realtime FMRI neurofeedback. PLoS One 6, e24522.