Regulation of anterior insular cortex activity using real-time fMRI
Andrea Caria,a,b,⁎Ralf Veit,a,eRanganatha Sitaram,aMartin Lotze,a
Nikolaus Weiskopf,cWolfgang Grodd,dand Niels Birbaumera,f
aInstitute of Medical Psychology and Behavioral Neurobiology, Eberhard-Karls-University of Tübingen, Tübingen, Germany
bDepartment of Cognitive Science and Education, University of Trento, Trento, Italy
cWellcome Department of Imaging Neuroscience, Institute of Neurology, University College London, London, UK
dSection of Experimental MR of the CNS, Department of Neuroradiology, University of Tübingen, Tübingen, Germany
eMax Planck Institute for Biological Cybernetics, Tuebingen, Germany
fNational Institute of Health (NIH), NINDS, Human Cortical Physiology, Bethesda, MD, USA
Received 17 October 2006; revised 12 January 2007; accepted 12 January 2007
Available online 31 January 2007
Recent advances in functional magnetic resonance imaging (fMRI)
data acquisition and processing techniques have made real-time fMRI
(rtfMRI) of localized brain areas feasible, reliable and less susceptible
to artefacts. Previous studies have shown that healthy subjects learn to
control local brain activity with operant training by using rtfMRI-
based neurofeedback. In the present study, we investigated whether
healthy subjects could voluntarily gain control over right anterior
insular activity. Subjects were provided with continuously updated
information of the target ROI’s level of activation by visual feedback.
All participants were able to successfully regulate BOLD-magnitude in
the right anterior insular cortex within three sessions of 4 min each.
Training resulted in a significantly increased activation cluster in the
anterior portion of the right insula across sessions. An increased
activity was also found in the left anterior insula but the percent signal
change was lower than in the target ROI. Two different control
conditions intended to assess the effects of non-specific feedback and
mental imagery demonstrated that the training effect was not due to
unspecific activations or non feedback-related cognitive strategies.
Both control groups showed no enhanced activation across the sessions,
which confirmed our main hypothesis that rtfMRI feedback is area-
specific. The increased activity in the right anterior insula during
training demonstrates that the effects observed are anatomically
specific and self-regulation of right anterior insula only is achievable.
This is the first group study investigating the volitional control of
emotionally relevant brain region by using rtfMRI training and
confirms that self-regulation of local brain activity with rtfMRI is
© 2007 Elsevier Inc. All rights reserved.
Keywords: Self-regulation; Physiological regulation; Real-time fMRI;
Brain–computer interface; Neurofeedback; Blood oxygen level-dependent;
Studies on physiological self-regulation of brain activity,
mostly using electroencephalography (EEG) demonstrated that,
with appropriate training, individuals can learn to control brain
processes. Learned regulation of slow cortical potentials was used
to allow communication in severely paralyzed patients (Birbaumer
et al., 1999; Kübler et al., 2001) and to suppress epileptic activity
(Kotchoubey et al., 2001). By using self-regulation of oscillatory
EEG activity patients with motor impairments were able to control
a hand prosthesis (Pfurtscheller et al., 2000; Neuper et al., 2003).
The mechanisms of these changes in EEG (SCP, mu and alpha
rhythms) are neuroanatomically specific and reflect activity in
complex brain networks (Hinterberger et al., 2005). Due to poor
spatial resolution, highly localized and subcortical brain regions
are difficult to regulate with EEG-neurofeedback.
Recent advances in functional magnetic resonance imaging
(fMRI) data acquisition and processing techniques have made
rtfMRI of localized brain areas feasible, reliable and less
susceptible to artefacts. rtfMRI allows on-line analysis of func-
tional brain activity and feedback of the Blood Oxygen Level-
Dependent (BOLD) signal from a targeted region of interest. In
addition, fMRI-based techniques in comparison to all the other
human brain mapping techniques, represents the only non-invasive
method allowing feedback regulation of deep subcortical brain
regions such as the limbic and paralimbic areas.
Previous studies (for a complete review see Weiskopf et
al., 2004b) showed that healthy subjects can learn to control
local brain activity by operant training with rtfMRI-based
These studies focused on different cortical and subcortical
areas: the sensorimotor cortex (deCharms et al., 2004; Yoo and
Jolesz, 2002), the supplementary motor area (SMA; Weiskopf et
al., 2004a), the parahippocampal place area (PPA; Weiskopf et al.,
2004a), the anterior cingulate cortex (ACC) (Weiskopf et al.,
2003), and the amygdala (Posse et al., 2003). In a recent study
NeuroImage 35 (2007) 1238–1246
⁎Corresponding author. Institute of Medical Psychology and Beha-
vioural Neurobiology, Eberhard-Karls-University of Tübingen, Gartenstr.
29, D-72074 Tuebingen, Germany. Fax: +49 7071 295956.
E-mail address: email@example.com (A. Caria).
Available online on ScienceDirect (www.sciencedirect.com).
1053-8119/$ - see front matter © 2007 Elsevier Inc. All rights reserved.
deCharms et al. (2005) demonstrated that subjects were able to
learn to control activation in the rostral anterior cingulate cortex
(rACC), a region implicated in mediating the perception of pain.
Furthermore, this study showed that control of up- and down-
regulation of rACC activation was associated with changes in pain
perception induced by noxious thermal stimulation. Chronic pain
patients were also trained to control activation in rACC and
reported reduction of the level of chronic pain after training. All
these studies provided significant evidence that individuals can
learn to voluntarily self-regulate brain activity by using feedback
training based on rtfMRI and that changes in behaviour might
occur as a direct consequence.
Among the few studies conducted so far the study by deCharms
et al. (2004, 2005) and Weiskopf et al. (2003, 2004a,b) provided
visual rtfMRI feedback to the subjects. Only long delayed rtfMRI
was used in the pilot study from Yoo and Jolesz (2002):
information about fMRI data was delayed about 20 s. Posse et
al. (2003) provided verbal feedback of the BOLD-activity in the
amygdala with a delay of 60 s.
In addition, very few studies have explored the application of
rtfMRI for training subjects to self-regulate activity in emotionally
relevant areas. In the study made by Posse and colleagues (2003)
the feedback was based on the experimenter’s rating and it was not
possible to disentangle whether regulation was achieved by
feedback or mood induction. While Weiskopf et al. (2003) did
indeed use rtfMRI for training to self-regulate BOLD signal of the
affective division of the anterior cingulate cortex, only one subject
was tested and hence the results may not be readily generalized.
In the present group study we investigated whether healthy
subjects can voluntarily gain control over right anterior insular
activity by using rtfMRI.
Cortical representation of smell and taste (Francis et al., 1999;
Rolls, 1996, 2004), viscerosensation (Craig, 2002), and pain
perception (Davis et al., 1998; Coghill et al., 1999; Peyron et al.,
2000) converge in the insula and surrounding operculum. The
activity of the insulae correlates with the subjective perception of
emotional states (Craig, 2002, 2003). Studies on emotional
perception showed that insula activity is correlating with the
aversive valence of stimuli (Anders et al., 2004). A review of PET
and fMRI studies investigating the neuroanatomy of emotion (Phan
et al., 2002) revealed that the anterior cingulate and insula were
recruited during induction by emotional recall/imagery and during
emotional tasks with cognitive demand.
Awareness of salient emotionally stimuli increases right insula
cortex activity (Critchley et al., 2004) suggesting that this area is
critical for the representation of bodily responses and interoception
Therefore, the possibility of volitional modulation of insula
activity may be a valuable tool to study emotion regulation.
Modulation of the insular activity with rtfMRI training might be
relevant for the development of novel approaches for clinical
treatment of social phobia or antisocial behavior which have shown
overactivity and hypoactivity, respectively, in the insular region
(Veit et al., 2002; Birbaumer et al., 2005).
Fifteen healthy right-handed subjects (9 women and 6 men; age
range 22–38 years; mean age 25.13 years) participated in this
study. Nine of them were trained to voluntarily control the local
BOLD signal of the right anterior insular cortex using the rtfMRI
information. The remaining six subjects participated in two
different control conditions (see below). All participants were
students of the Medical School and had no history of neurological
or psychiatric disorders including substance abuse/dependence and
psychotropic medications. All were naive to neurofeedback and
fMRI experiments. Written instructions were provided to all
participants and informed written consent was obtained. Subjects
were carefully instructed not to move, relaxing and breathing
regularly in order to avoid potential BOLD artefacts due to
manipulation of internal state. This study was approved by the
local ethics committee of the Faculty of Medicine of the University
fMRI data acquisition
Functional images were acquired on 3.0 T whole body scanner,
with standard 12-channel head coil (Siemens Magnetom Trio Tim,
Siemens, Erlangen, Germany). A standard echo-planar imaging
sequence was used (EPI; TR=1.5 s, matrix size=64×64, effective
echo time TE=30 ms, flip angle α=70°, bandwidth=1.954 kHz/
pixel). Sixteen slices (voxel size=3.3×3.3×5.0 mm3, slice
gap=1 mm), AC/PC aligned in axial orientation were acquired.
For superposition of functional maps upon brain anatomy a
high-resolution T1-weighted structural scan of the whole brain was
collected from each subject (MPRAGE, matrix size=256×256,
160 partitions, 1 mm3isotropic voxels, TR=2300 ms, TE=
3.93 ms, TI=1100 ms, α=8°).
In order to reduce movements two foam cushions immobilized
the participant’s head.
Real-time fMRI data processing
The fMRI setup used for real-time data processing is based on
Turbo-BrainVoyager (Brain Innovation, Maastricht, The Nether-
lands; Goebel, 2001) as previously described by Weiskopf et al.
Data were analyzed in real-time with Turbo-BrainVoyager
software performing on-line incremental 3D motion detection and
correction, and drift removal. The software is capable of
incrementally computing statistical maps based on the General
Linear Model (GLM) and event-related averages.
The selection of ROI1 – the right anterior insula– was
anatomically based on the high resolution T1 structural scan.
This ROI was a rectangular area encompassing 4×5 voxels
(~15×20 mm) on a single slice (5 mm). The reference ROI2 was a
large background region of interest selected from a reference slice
positioned distant from ROI1 encompassing the whole brain with
the intent to cancel global effects and to average out any unspecific
activation. During training, the mean BOLD signal from the
regions of interest ROI1 and ROI2 was extracted. The first ten
volumes of each session were excluded from statistical analysis to
account for T1 equilibration effects. For the feedback presentation
the difference between the two ROI time-courses was calculated
and normalized to the baseline.
−(BOLDreg−BOLDrest)ROI2,BOLDregand BOLDrestconstitute the
respective BOLD-signal during regulation and rest period. The
subjects were provided visually with continuously updated
information of the ROI1 level of activation. The visual feedback
1239 A. Caria et al. / NeuroImage 35 (2007) 1238–1246
of brain activity was calculated using Matlab 6.5 (The MathWorks,
Natick, MA) software running on a separate personal computer
connected via LAN to the scanner and to the Turbo-BrainVoyager.
The feedback consisted of a graduated thermometer displaying
changes of BOLD-activity with increasing or decreasing number of
bars (see Fig. 1). The number of available bars was limited. Bars
above the baseline level of activation were colored in red while
those below the baseline level in blue. Thermometer bars were
constantly updated and new fMRI information was available with a
delay of about 1.5 s.
The training consisted of four feedback sessions followed by a
‘transfer’ session performed in 1 day. One feedback session
consisted of four regulation blocks (22.5 s each) during which the
subjects had to learn to increase insula activity alternating with five
baseline blocks (22.5 s each) during which they had to return the
activity to the baseline level. Each session lasted about 4 min and
was repeated five times including the transfer session. During the
feedback session the normalized average BOLD signal from the
right anterior insula was presented to the subjects by means of
thermometer bars. The thermometer display was present both
during regulation and during rest period. The regulation blocks
were cued with a red arrow at the thermometer display while
during rest blocks a cross hair was presented in the same position
(see Fig. 1).
During the transfer session, subjects were instructed to perform
the same task as during feedback but fMRI information was not
provided and bars were not shown. The transfer session was per-
formed to verify the efficacy of the feedback and to check whether
training effects might persist beyond the experimental situation.
Pilot experiments showed that learning without any guidelines
for mental strategies was not achievable in a short training period
and led to a drop of motivation especially in the uncomfortable
environment of an MRI scanner (Sitaram et al., 2005; deCharms et
al., 2005). For this purpose subjects were instructed to use
cognitive strategies that potentially would help to learn to control
the activity of the target ROI. Specifically, strategies for regulation
blocks were focused on emotion induction by recall of personal
and affectively relevant events. No cues for aversive or pleasant
imagery was given. During baseline subjects were required to
count back from 100. At the end of each session subjects reported
mental strategies they had used during regulation blocks. Subjects
were also informed of the data processing delay of about 1.5 s and
of the intrinsic physiological hemodynamic response delay of
about 6 s.
Additionally, two different control experiments were conducted
to verify that the effects of the self-regulation of the insular activity
were due to fMRI feedback. Three subjects participated to each of
the control experiments. The first control condition aimed to verify
the specificity of the feedback information; the control group
performed three sessions of the same experimental paradigm but
received sham feedback. This sham feedback was not specific to
any particular brain area but consisted of information from a large
background ROI from the same subjects not encompassing the
anterior insulae. Feedback was comparable in terms of signal
magnitude and variability. The second control condition assessed
the effects of repetitive use of mental imagery. Subjects were
provided with the same instructions and same general strategies as
before, the thermometer frame was present but no rtfMRI
information was available (no bars were shown). Subjects
performed three consecutive sessions during which they were
asked to recall and evoke memories and imagery of personally
relevant affective events.
Off-line data analysis
Off-line image post-processing and data analysis were
performed using SPM2 statistical parametric mapping software
package (Wellcome Department of Imaging Neuroscience,
London), while MarsBar toolbox (the Marseille region of interest
toolbox for SPM2) and BrainVoyager QX were used for ROI
Before whole brain statistical analysis, functional EPI volumes
were realigned spatially, normalized into Montreal Neurological
Institute (MNI) space, and smoothed spatially (9-mm Gaussian
kernel) and temporally (0.0088 Hz, 2.5 times the duration of the
activation and baseline block) to remove high-frequency artefacts.
Hemodynamic response amplitudes were estimated using standard
regressors, constructed by convolving a boxcar function, repre-
senting the block duration, with a canonical hemodynamic
response function using standard SPM2 parameters. Motion
parameters were also included into the general linear model
(GLM) as covariates to take into account artefacts caused by head
Signal increase during regulation with respect to the baseline
was evaluated by SPM2. Areas showing training related changes
Fig. 1. Real-time feedback video projected to the subjects. The feedback consisted of a graduated thermometer displaying percentage change of BOLD-activity
by showing an increasing or decreasing number of bars. The number of bars available was limited and fixed to a value of 20. Bars over the baseline level of
activation were colored in red while those under the baseline level in blue. Thermometer bars were constantly updated and new fMRI information was available
with a delay of about 1.5 s. The regulation blocks were cued with a red arrow (right) beside the thermometer display while during the rest blocks a cross hair (left)
was presented in the same position.
1240 A. Caria et al. / NeuroImage 35 (2007) 1238–1246
were analysed by performing t-test comparisons of increased
BOLD-effect magnitude over sessions. Group analysis was
performed session by session using a random effects analysis.
Statistical significance of activation maps were based on a t test
with a voxel-wise threshold of P<0.01.
Hypothesis-driven ROI analysis was performed using the ROI
previously selected for each subject during the training. ROI time
series underwent the same preprocessing and GLM used for whole
brain analysis. The percent signal change during regulation blocks
with respect to the baseline blocks was calculated for each session
separately and then averaged across subjects.
ROI analysis was also performed on a comparable contralateral
region of the same extension positioned at the left anterior insular
region. The training effect was evaluated by computing an
ANOVA for repeated measures on all subjects of percent signal
changes in the specific ROI session by session. Furthermore, all
significantly activated clusters from statistical maps other than the
left and right anterior insular region were checked with MarsBar
toolbox for potential increase across sessions. Activation maps
produced by offline analysis matched and validated activations
maps produced in real-time by Turbo-BrainVoyager. Additionally a
lateralization index (LI) was calculated based on the normalized
difference between percent signal change extracted from the target
ROI (%R) and from the contralateral ROI (%L) as follows: (%R−
%L)/(%R+%L). The LI calculation intended to assess laterality
effects during training and it was calculated for each subject and
then compared with control experiments.
All participants were able to successfully regulate BOLD-
magnitude in the right anterior insular cortex. Training resulted in a
significantly increased activation cluster in the anterior portion of
the right insula across sessions. Subjects reported the use of both
positive and negative mental imagery. Positive strategies were
focused on recalling themselves playing music, playing with
daughter, engaging in sport activities, recall of holidays; while
negative strategies were focused mostly on bringing back
themselves in dangerous situations, anger states and while taking
Linear regression across all sessions performed on the indi-
vidually selected region of interest showed significant increase of
activity in the target area [y=0.174+0.127, P<0.012] (see Fig. 2).
Three subjects did not complete the fourth training session hence
we conducted the group analysis considering the first three sessions.
The success of training is clearly visible by comparing the time
course of the selected area during the last session (see Fig. 3, lower
image) with the first session (see Fig. 3, upper image). Percent
signal change calculated in the ROI as difference between task and
rest for each subject and then averaged across all the participants
resulted in a clear monotonic increase across the first three sessions
[repeated measures ANOVA, F(2,7)=10.32, P=0.001] (see Fig. 4).
Fig. 2. BOLD percent single change computed on the individually selected
region of interest across feedback sessions on single subjects. A significant
BOLD increase in the target area was observed across sessions.
Fig. 3. Single subject statistical maps (left) and BOLD time-courses (right) of the right anterior insula in the first (top) and in the last session (bottom). The
selectedregion of interest is delineatedby the green box.Functional imagesare in the radiological convention and are not normalized. Statistical significancewas
based on t test comparing activation on each voxel during the regulation blocks with respect to the baseline blocks, with a threshold of P<0.05 false discovery
rate (FDR) corrected for multiple comparisons (Genovese et al., 2002). The time course of the BOLD activity (white line) is related to the ROI selected and is
showing the progress during the regulation blocks (green) and the baseline blocks (gray). Number of volumes is in the x axis and magnitude signal in the y axis;
these values are the raw output from the scanner.
1241A. Caria et al. / NeuroImage 35 (2007) 1238–1246
Percent signal change was also calculated in the corresponding
contralateral area of the target ROI (see Fig. 4). No significant
monotonic increase was found in the left anterior insula [repeated
measures ANOVA, F(2,7)=1.94, P=0.177] even though a small
significant increase was found between session3 and session1
[paired samples t test, t(8)=2.92, P=0.019]. Activation during
training was lateralized to the right with a mean lateralization index
of 0.38±0.16 in the last session [one-sample t test, t(8)=2.20,
P=0.029] (see Fig. 8).
Random effects analysis on the experimental group confirmed
an increased BOLD-magnitude in the right anterior insular cortex
over time (see Fig. 5). Analysis of the single sessions revealed no
significant activation during the first session in the target area; a
significant activation cluster [t=4.50; P=0.001] during the second
session (MNI coordinates: 39, 33, 0); and a highly significant
activation cluster [t=10.23; P<0.001] during the third session
(MNI coordinates: 36, 26, 6). Fixed effect analysis was also
performed on the six subjects who completed the fourth session
reporting a higher significant cluster [t=12.47, P<0.001, FWE
corrected, MNI coordinates: 36, 23, 5].
Fig. 4. Group analysis of percent signal change during training in the right
(target ROI) (top) and left (bottom) anterior insula. Percent signal change,
calculated in the ROI as difference between task and rest on each subject and
then averaged across all the participants, reported clear monotonic increase
across the three sessions. Percent signal change calculated in the
corresponding contralateral area of the target ROI did not show a significant
monotonic increase even though a significant increase was measured
between session3 and session1. Transfer session (no rtfMRI information
provided to the subjects) showed an increase of BOLD-magnitude in the
right and left anterior insula but only the left insula showed a significant
BOLD-magnitude increase in comparison with the first session.
Fig. 5. Random effects analysis on the experimental group confirmed an increased BOLD–magnitude in the right anterior insular cortex over time course. SPM2
of the single sessions showed no significant activation during the first session in the target area; a significant activation cluster (t=4.50; P=0.001, uncorrected)
during the second session (MNI coordinates: 39, 33, 0); and a highly significant activation cluster (t=10.23; P<0.001, uncorrected) during the third session
located (MNI coordinates: 36, 26, 6). Fixed effect analysis was also performed on the six subjects who completed the fourth session reporting a higher significant
cluster (t=12.47, P<0.001, FEW-corrected; MNI coordinates: 36, 23, 5). All activation maps are projected on a single-subject T1 template at the coordinate
Comparison of significant signal increase during activation blocks in the last
Brain regions Brodmann
(x, y, z)
R frontal inferior
R premotor cortex
36, 26, 6
49, 20, 5
BA 650 8.24
49, −3, 55
49, 0, 35
36, −3, 35
26, −63, 50
30 −49 50
30, 3, 65
26, −7, 55
30, 3, 55
46, −69, 0
43, −53, 5
−53, −3, 50
−40, −3, 45
10, 10, 60
R angular gyrus
R inferior parietal gyrus
R superior frontal gyrus
BA 6 28
R middle temporal gyrus BA 3724
L premotor cortexBA 613
R supplementary motor
BA 6 24
BA 47 62 5.63
−33, 20, 0
Clusters of significant signal increase during activation blocks from random
effects analysis in the last session. Bold numbers correspond to highest peak
of the cluster. Clusters exceeding the threshold of P<0.01 uncorrected and
witha spatialextentlarge than10voxelswereconsidered. Coordinates arein
MNI stereotaxic space (Collins et al., 1994) and labelled anatomically
according to Tzourio-Mazoyer et al. (2002).
1242A. Caria et al. / NeuroImage 35 (2007) 1238–1246
Group analysis showed additional brain activations during the
last session, specifically in the left and right premotor cortex, right
angular gyrus, right superior frontal gyrus, right middle temporal
gyrus, right supplementary motor area and left insula (see Table 1).
Further ROI analysis of all additional activation clusters in the last
session was also performed with MarsBar in order to estimate
increase of signal change across sessions. No significant increase
was observed in areas other than the target ROI during the
feedback training (see Fig. 6).
The results of the transfer session showed an increase of
BOLD-magnitude in the target ROI but in comparison to the first
session t test it was not significant [t(8)=4.86, P=0.06] because of
the high variance between subjects. A significant increase of
activity was found in the left anterior insula in comparison to the
first session [paired samples t test, t(8)=2.72, P=0.030]. Both
control conditions showed no increase in BOLD-magnitude in the
right anterior insula (see Fig. 7) and did not show any significant
With rtfMRI feedback a specific modulation of the right
anterior insula is possible. This was achieved after a short training
time. BOLD signal in the target ROI increased with the number of
feedback sessions, indicating training effects and learning.
Previous studies from deCharms et al. (2005) and Weiskopf et
al. (2004a,b) also showed that one single-day training with rtfMRI
feedback is enough to achieve learning.
Areas which showed activation during the last training session
did not show a training effect. They may, however promote the
self-control of the specific target ROI. The left anterior insula
showed increased activity in the last session with respect to the
first session but the percent signal change was much lower than in
the target ROI and no significant monotonic increase was found.
Furthermore, the lateralization index confirmed a stronger effect in
the right anterior insula during training (Fig. 8). This demonstrate
that, even though emotional tasks often involve both insulae,
specific regulation of the right anterior insula only is achievable.
deCharms et al. (2004) did not report a significant increase in
activation of the ROI placed in a comparable position to the target
ROI but ipsilateral to the motor task being performed; in this study,
subjects were instructed that during the task blocks they had to
Fig. 6. All significantly activated clusters from statistical maps, besides the
left and right anterior insular region, were checked with MarsBar toolbox for
potential increase across sessions correlated with the feedback task due to
effects of general arousal. No significant increase was observed in areas
other than the target ROI during the feedback training.
Fig. 7. Group analysis of percent signal change during control experiments
in the right anterior insula. (A) Percent signal change averaged over the
group during sham training. (B) Task-driven activation during mental
imagery performance. Both control groups showed no increase in BOLD-
magnitude in the right anterior insula.
Fig. 8. Lateralization Index (R=+÷L=−). It was calculated based on the
normalized difference between percent signal change extracted from the
target ROI and from the contralateral ROI on each single subject and then
averaged. Activation during training was lateralized to the right with a mean
lateralization indexof 0.38±0.16in the last session.Control experiments did
not show a similar result.
1243A. Caria et al. / NeuroImage 35 (2007) 1238–1246
imagine moving their dominant (right) hand so as to increase the
level of activation in the ROI of the contralateral somatomotor
The group analysis of the transfer session showed increase of
activity in both right and left anterior insula. This result further
underscores the specificity of the rtfMRI feedback in the target
ROI. During feedback, the activation in the right anterior insula is
larger than in the left anterior insula, but the activation are
comparable during the transfer session (no rtfMRI information
provided). One could speculate that the requirement for specific
regulation of the target ROI reduces activity in the contralateral
area. A previous study (deCharms et al., 2004) reported that after a
longer training it was possible to continue regulation of the BOLD
signal without feedback.
Two different control conditions demonstrated that the training
effect was not due to global and unspecific activations. Both control
conditions and activation of other brain areas showed no enhanced
activation across sessions confirming our hypothesis. These
findings are in line with the studies of deCharms et al. (2004,
2005) which reported no effect in several control conditions and
The subjects reported that the experiment was hard but
challenging. In the previous rtfMRI studies (Weiskopf et al.,
2003, 2004a,b; deCharms et al., 2004, 2005), subjects were
presented with continuously updated time-course plot of the target
ROI and they were also provided with the time-course of a
background ROI and the difference between the two (deCharms
et al., 2004), on-line motion correction (Weiskopf et al., 2003), or
video images (deCharms et al., 2005). In this study only
information of the target ROI activation was provided and no
graph was displayed to the subjects; we assert that that subject’s
attention was focused more on the task by this display.
To our knowledge, this is the first group study reporting
volitional control over emotionally relevant brain region with
rtfMRI training. This provides further confirmation of self-
regulation of local brain activity with rtfMRI and extends previous
findings to the area of the anterior insula.
An important point to mention in order to properly interpret the
reported results is the question whether insula cortex activity in
humans represents or controls visceral and cardiorespiratory
functions which still remains controversial. The experiments
conducted so far (Penfield and Faulk, 1955; Oppenheimer et al.,
1992; Sander and Klingelhöfer, 1995; Saper, 2002) indicate that the
human insular cortex does contain autonomic control sites, however
these results were obtained from patients and using poor anatomical
localization. Moreover a study from King et al. (1999) indicated
different sub-regions in the anterior insular cortex, a ventral part
responding to gustatory stimulation and a more superior part
responding to cardiopulmonary stimulation. Wager and Feldman
Barrett (2004) suggest that the ventral anterior agranular insula is
activated consistently by neuroimaging studies involved in aurally
and recall-generated emotion induction. Furthermore Rainville et al.
(2006) indicates that the feeling of basic emotions is inherently
associated with distinct patterns of cardiorespiratory activity.
However Nagai et al. (2004) reported no right anterior insula
activity during biofeedback regulation of skin conductance level
and in a rtfMRI study from Posse et al. (2003) subjects performing a
sad mood induction task did not show changes in respiration rate
and end-expiratory pco2(Petco2) strong enough to alter global
fMRI contrast. Yet, a study from Critchley et al. (2004) showed
right anterior insula activity enhanced by interoceptive awareness in
the absence of physiological changes indicating a major role of this
region for feelings perception.
The increase in the right insula activity during the training may
represent an increased attention to internal body sensations
generated during subjective affective experience as the right insula
contributes to subjective emotional responses (Critchley et al.,
2004). This increased activation may even result in an increased
sensitivity to emotionally relevant stimuli, a hypothesis not tested
by the present study.
Previous studies on cognitive control of emotions (Ochsner et
al., 2004; Ochsner and Gross, 2005) demonstrated that controlling
arousing stimuli depend upon interactions between cortical and
subcortical (e.g. insula, amygdala) emotional systems.
Furthermore, a recent hypothesis is that the anterior insula is
involved in representing one’s own and others’ affective states.
Activations in the insula bilaterally were reported both when
feeling ones own pain and when observing a loved one experience
pain (Singer et al., 2004).
In addition, fMRI studies investigating the hypothesis that
underactivity of the frontolimbic fear circuitry underlies psycho-
pathic behaviour revealed differential activation in the prefrontal–
limbic circuit (orbitofrontal cortex, insula, anterior cingulate,
amygdala) in the healthy subjects while psychopaths displayed
brief amygdala, but no further brain activation (Birbaumer et al.,
1998, 2005; Davidson et al., 2000; Raine, 2000; Blair, 2003; Veit et
al., 2002). This prefrontal–limbic circuit mediates anticipatory
avoidance and emotion regulation and adjustment, particularly in
social contexts. The modulation of insular activity by using rtfMRI-
based training may be particularly relevant for the development of
novel approaches for the treatment of anxiety disorders and
antisocial behaviour. Testing whether these patients are able to
modulate the insula activity and whether its learned modulation
may lead to behavioural changes is particularly intriguing.
Insula hyperactivity seems to be a common feature in persons
with elevated trait anxiety (Simmons et al., 2006; Stein et al.,
2007). Recently an interesting hypothesis to explain anxiety
proneness has been proposed by Paulus and Stein (2006)
indicating the anterior insula as the key region for integration of
affective and cognitive processing. Anxiety-prone individuals
would have altered interoception correlated with increased activity
in the anterior insula as a basis of the trigger of the internal anxiety
state which in turn would modify cognitive and behavioural
A potential and interesting application of the rtfMRI technique
in such a clinical setting aiming to assess the effects of down
regulation of anterior insular cortex in anxiety-prone subjects is of
great interest and it would also extend the knowledge of the role of
the anterior insular cortex in anxiety.
Finally, the functional interaction of different brain areas (e.g.
insula, ACC, amygdala) may play a significant role in the local
activation, the connectivity between brain areas may become an
important physiological target for rtfMRI feedback. Selection of
more than one brain region of interest or patterns of distributed
network brain activity might be a fascinating target for future
This work was supported by grants from Deutsche Forschungs-
gemeinschaft (SFB 437/F1). Andrea Caria is supported by a Marie
Curie Host Fellowship for Early Stage Researchers Training. We
1244 A. Caria et al. / NeuroImage 35 (2007) 1238–1246
are indebted to M. Erb for technical assistance in data acquisition
and helpful discussions.
Adam, G., 1998. Visceral Perception: Understanding Internal Cognition.
Plenum Press, New York.
Anders, S., Lotze, M., Erb, M., Grodd, W., Birbaumer, N., 2004. Brain
activity underlying emotional valence and arousal: a response-related
fMRI study. Hum. Brain Mapp. 23, 200–209.
Birbaumer, N., Grodd, W., Diedrich, O., Klose, U., Erb, M., Schneider, F.,
Weiss, U., Flor, H., 1998. fMRI reveals amygdala activation to human
faces in social phobics. NeuroReport 9, 1223–1226.
Birbaumer, N., Ghanayim, N., Hinterberger, T., Iversen, I., Kotchoubey, B.,
Kübler, A., Perelmouter, J., Taub, E., Flor, H., 1999. A spelling device
for the paralysed. Nature 398, 297–298.
Birbaumer, N., Veit, R., Lotze, M., Erb, M., Hermann, C., Grodd, W., Flor,
H., 2005. Deficient fear conditioning in psychopathy. Arch. Gen.
Psychiatry 62, 799–805.
Blair, R.J.R., 2003. Neurobiological basis of psychopathy. Br. J. Psychiatry
Coghill, R.C., Sang, C.N., Maisog, J.M., Iadarola, M.J., 1999. Pain intensity
processing within the human brain: a bilateral, distributed mechanism.
J. Neurophysiol. 82 (4), 1934–1943.
Collins, D.L., Neelin, P., Peters, T.M., Evans, A.C., 1994. Automatic 3D
intersubject registration of MR volumetric data in standardized Talairach
space. J. Comput. Assist. Tomogr. 18, 192–205.
Craig, A.D., 2002. How do you feel? Interoception: the sense of the
physiological condition of the body. Nat. Rev., Neurosci. 3 (8),
Craig, A.D., 2003. Interoception: the sense of the physiological condition of
the body. Curr. Opin. Neurobiol. 13 (4), 500–505.
Critchley, H.D., Wiens, S., Rotshtein, P., Ohman, A., Dolan, R.J., 2004.
Neuralsystems supporting interoceptive awareness. Nat. Neurosci. 7 (2),
Davidson, R.J., Putnam, K.M., Larson, C.L., 2000. Dysfunction in the
neural circuitry of emotion regulation – a possible prelude to violence.
Science 289, 591–594.
Davis, K.D., Kwan, C.L., Crawley, A.P., Mikulis, D.J., 1998. Functional
MRI study of thalamic and cortical activations evoked by cutaneous
heat, cold, and tactile stimuli. J. Neurophysiol. 80 (3), 1533–1546.
deCharms, R.C., Christoff, K., Glover, G.H., Pauly, J.M., Whitfield, S.,
Gabrieli, J.D.E., 2004. Learned regulation of spatially localized brain
activation using real-time fMRI. NeuroImage 21, 436–443.
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. 102 (51), 18626–18631.
Francis, S., Rolls, E.T., Bowtell, R., McGlone, F., O’Doherty, J., Browning,
A., 1999. The representation of pleasant touch in the brain and its
relationship with taste and olfactory areas. NeuroReport 10 (3),
Genovese, C.R., Lazar, N.A., Nichols, T., 2002. Thresholding of statistical
maps in functional neuroimaging using the false discovery rate.
NeuroImage 15 (4), 870–878.
Goebel, R., 2001. Cortex-based real-time fMRI. Neuroimage 13, S129.
Hinterberger, T., Veit, R., Wilhelm, B., Weiskopf, N., Vatine, J.J.,
Birbaumer, N., 2005. Neuronal mechanisms underlying control of a
brain–computer interface. Eur. J. Neurosci. 21 (11), 3169–3181.
King, A.B., Menon, R.S., Hachinski, V., Cechetto, D.F., 1999.
Human forebrain activation by visceral stimuli. J. Comp. Neurol. 413,
Kotchoubey, B., Strehl, U., Uhlmann, C., Holzapfel, S., Konig, M.,
Froscher, W., Blankenhorn, V., Birbaumer, N., 2001. Modification of
slow cortical potentials in patients with refractory epilepsy: a controlled
outcome study. Epilepsia 42, 406–416.
Kübler, A., Kotchoubey, B., Kaiser, J., Wolpaw, J.R., Birbaumer, N., 2001.
Brain–computer communication: unlocking the locked in. Psychol.
Bull. 127, 358–375.
Nagai, Y., Critchley, H.D., Featherstone, E., Trimble, M.R., Dolan, R.J.,
2004. Activity in ventromedial prefrontal cortex covaries with
sympathetic skin conductance level: a physiological account of a
“default mode” of brain function. NeuroImage 22, 243–251.
Neuper, C., Müller, G.R., Kübler, A., Birbaumer, N., Pfurtscheller, G., 2003.
Clinical application of an EEG based brain–computer interface: a case
study in a patient with severe motor impairment. Clin. Neurophysiol.
Ochsner, K.N., Gross, J.J., 2005. The cognitive control of emotion. Trends
Cogn. Sci. 9, 242–249.
Ochsner, K.N., Ray, R.D., Cooper, J.C., Robertson, E.R., Chopra, S.,
Gabrieli,J.D.E., Gross,J.J., 2004.Forbetteror for worse:neuralsystems
supporting the cognitive down- and up-regulation of negative emotion.
NeuroImage 23, 483–499.
Oppenheimer, S.M., Gelb, A., Girvin, J.P., Hachinski, V.C., 1992.
Cardiovascular effects of human insular cortex stimulation. Neurology
Paulus, M.P., Stein, M.B., 2006. An insular view of anxiety. Biol. Psychiatry
Penfield, W., Faulk Jr., M.E., 1955. The insula: further observations on its
function. Brain 78, 445–470.
Peyron, R., Laurent, B., Garcia-Larrea, L., 2000. Functional imaging of
Pfurtscheller, G., Guger, C., Müller, G., Krausz, G., Neuper, C., 2000. Brain
oscillations control hand orthosis in a tetraplegic. Neurosci. Lett. 292,
Phan, K.L., Wager, T., Taylor, S.F., Liberzon, I., 2002. Functional
neuroanatomy of emotion: a meta-analysis of emotion: activation
studies in PET and fMRI. NeuroImage 16, 331–348.
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. Neuro-
Image 18, 760–768.
Raine, A., 2000. Autonomic nervous system factors underlying disinhibited,
antisocial, and violent behavior. Ann. N. Y. Acad. Sci. 794, 46–59.
Rainville, P., Bechara, A., Naqvi, N., Damasio, A.R., 2006. Basic emotions
are associated with distinct patterns of cardiorespiratory activity. Int. J.
Psychophysiol. 61, 5–18.
Rolls, E.T., 1996. The orbitofrontal cortex. Philos. Trans. R. Soc. Lond., B
Biol. Sci. 351 (1346), 1433–1443.
Rolls, E.T., 2004. The functions of the orbitofrontal cortex. Brain Cogn. 55
Sander, D., Klingelhöfer, J., 1995. Changes of circadian blood pressure
patterns and cardiovascular parameters indicate lateralization of
sympathetic activationfollowing hemispheric brain infarction. J. Neurol.
Saper, C.B., 2002. The central autonomic nervous system: conscious
visceral perception and autonomic pattern generation. Annu. Rev.
Neurosci. 25, 433–469.
Simmons, A., Strigo, I., Matthews, S.C., Paulus, M.P., Stein, M.B., 2006.
Anticipationof aversivevisual stimuliis associatedwithincreased insula
activation in anxiety-prone subjects. Biol. Psychiatry 60, 402–409.
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 (5661), 1157–1162.
Sitaram, R., Caria, A., Veit, R., Kübler, A., Gaber, T., Birbaumer, N., 2005.
Real-time fMRI based brain–computer interface enhanced by interactive
virtual worlds. 45th Annual Meeting Society for Psychophysiological
Research. Lisbon, Portugal.
Stein, M.B., Simmons, A., Feinstein, J.S., Paulus, M.P., 2007. Increased
amygdala and insula activation during emotion processing in anxiety-
prone subjects. Am. J. Psychiatry 164 (2), 318–327.
Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard,
1245 A. Caria et al. / NeuroImage 35 (2007) 1238–1246
O., Delcroix, N., Mazoyer, B., Joliot, M., 2002. Automated anatomical Download full-text
labeling of activations in SPM using a macroscopic anatomical
parcellation of the MNI MRI single-subject brain. NeuroImage 15,
Veit, R., Flor, H., Erb, M., Hermann, C., Lotze, M., Grodd, W.,
Birbaumer, N., 2002. Brain circuits involved in emotional learning in
antisocial behavior and social phobia in humans. Neurosci. Lett. 328,
Wager, T.D., Feldman Barrett, L., 2004. From affect to control: functional
specialization of the insula in motivation and regulation. Published
online at PsycExtra.
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., Mathiak, K., Bock, S.W., Scharnowski, F., Veit, R.,
Grodd, W., Goebel, R., Birbaumer, N., 2004a. Principles of a brain–
computer interface (BCI) based on real-time functional magnetic
resonance imaging (fMRI). IEEE Trans. Biomed. Eng. 51 (6),
Weiskopf, N., Scharnowski, F., Veit, R., Goebel, R., Birbaumer, N.,
Mathiak, K., 2004b. Self-regulation of local brain activity using real-
time functional magnetic resonance imaging (fMRI). J. Physiol. (Paris
98 (4–6), 357–373.
Yoo, S.S., Jolesz, F.A., 2002. Functional MRI for neurofeedback: feasibility
study on a hand motor task. NeuroReport 13, 1377–1381.
1246 A. Caria et al. / NeuroImage 35 (2007) 1238–1246