Default mode network activity in schizophrenia studied at resting state using probabilistic ICA.

Abstract

Alterations in brain function in schizophrenia and other neuropsychiatric disorders are evident not only during specific cognitive challenges, but also from functional MRI data obtained during a resting state. Here we apply probabilistic independent component analysis (pICA) to resting state fMRI series in 25 schizophrenia patients and 25 matched healthy controls. We use an automated algorithm to extract the ICA component representing the default mode network (DMN) as defined by a DMN-specific set of 14 brain regions, resulting in z-scores for each voxel of the (whole-brain) statistical map. While goodness of fit was found to be similar between the groups, the region of interest (ROI) as well as voxel-wise analysis of the DMN showed significant differences between groups. Healthy controls revealed stronger effects of pICA-derived connectivity measures in right and left dorsolateral prefrontal cortices, bilateral medial frontal cortex, left precuneus and left posterior lateral parietal cortex, while stronger effects in schizophrenia patients were found in the right amygdala, left orbitofrontal cortex, right anterior cingulate and bilateral inferior temporal cortices. In patients, we also found an inverse correlation of negative symptoms with right anterior prefrontal cortex activity at rest and negative symptoms. These findings suggest that aberrant default mode network connectivity contributes to regional functional pathology in schizophrenia and bears significance for core symptoms.

Full text

Page 1

Default mode network activity in schizophrenia studied at resting state using
probabilistic ICA
Gianluca Mingoiaa,⁎, Gerd Wagnera, Kerstin Langbeina, Raka Maitraa, Stefan Smesnya, Maren Dietzeka,
Hartmut P. Burmeisterb, Jürgen R. Reichenbachc, Ralf G.M. Schlössera, Christian Gasera,
Heinrich Sauera, Igor Nenadica,⁎⁎
aDepartment of Psychiatry and Psychotherapy, Jena University Hospital, Jena, Germany
bInstitute for Diagnostic and Interventional Radiology I (IDIR I), Jena University Hospital, Jena, Germany
cMedical Physics Group, Institute for Diagnostic and Interventional Radiology I (IDIR I), Jena University Hospital, Jena, Germany
a b s t r a c ta r t i c l ei n f o
Article history:
Received 28 July 2011
Received in revised form 11 January 2012
Accepted 27 January 2012
Available online xxxx
Keywords:
Amygdala
Default mode network (DMN)
Functional magnetic resonance imaging
(fMRI)
Independent component analysis (ICA)
Prefrontal cortex (PFC)
Resting state
Schizophrenia
Alterations in brain function in schizophrenia and other neuropsychiatric disorders are evident not only during
specific cognitive challenges, but also from functional MRI data obtained during a resting state. Here we apply
probabilistic independent component analysis (pICA) to resting state fMRI series in 25 schizophrenia patients
and 25 matched healthy controls. We use an automated algorithm to extract the ICA component representing
the default mode network (DMN) as defined by a DMN-specific set of 14 brain regions, resulting in z-scores
for each voxel of the (whole-brain) statistical map. While goodness of fit was found to be similar between
the groups, the region of interest (ROI) as well as voxel-wise analysis of the DMN showed significant differences
between groups. Healthy controls revealed stronger effects of pICA-derived connectivity measures in right and
left dorsolateral prefrontal cortices, bilateral medial frontal cortex, left precuneus and left posterior lateral
parietal cortex, while stronger effects in schizophrenia patients were found in the right amygdala, left
orbitofrontal cortex, right anterior cingulate and bilateral inferior temporal cortices. In patients, we also found
an inverse correlation of negative symptoms with right anterior prefrontal cortex activity at rest and negative
symptoms. These findings suggest that aberrant default mode network connectivity contributes to regional
functional pathology in schizophrenia and bears significance for core symptoms.
© 2012 Elsevier B.V. All rights reserved.
1. Introduction
Functional magnetic resonance imaging (fMRI) has mostly been
used to assess deficits of task-induced activation, e.g. during working
memory tasks (Minzenberg et al., 2009). More recent studies have
provided evidence for alterations detectable already under “resting
state” conditions, i.e. without performing a specific cognitive task.
While such resting-state abnormalities have been observed in several
neuropsychiatric disorders (Greicius, 2008; Broyd et al., 2009), these
findings bear particular significance for schizophrenia, as they might
be related to cognitive impairment and clinical symptoms.
The approaches to analyse resting-state fMRI data in schizophrenia
allexploitthefactthattheBOLDsignalshowslow-frequencyfluctuations
(Auer, 2008), which are assumed to be linked to resting-state networks
(Damoiseaux et al., 2006). While some studies have demonstrated
changes on regional amplitude of low-frequency fluctuations during
rest (Huang et al., 2009; Hoptman et al., 2010), others have used either
correlation of seed-regions such as the posterior cingulate cortex (PCC)
with other brain areas (Bluhm et al., 2007), or have used independent
component analysis (ICA) to extract sets of regions following a similar
time course (Garrity et al., 2007). Despite different methodologies,
these studies appear to overlap in alteration of nodes of the default
mode network (DMN), esp. the medial prefrontal cortex.
The default mode network (DMN) is a concept based on an inter-
connected set of areas showing higher activity during rest than task-
related activity (Raichle et al., 2001; Raichle and Snyder, 2007). This
network has been defined by initial studies of Shulman et al. based
on changes of cerebral blood flow during visual tasks (Shulman et al.,
1997). Since then it has been studied extensively with both seed-ROI
based correlations and ICA methods (Raichle and Snyder, 2007; van
den Heuvel and Hulshoff Pol, 2010), and linked to electrophysiological
activity in the beta and gamma band (Mantini et al., 2007).
Recent studies on DMN activity in schizophrenia have suggested me-
dial prefrontal cortical areas of the DMN network to show aberrant con-
nectivity or activity, although the evidence is not completely converging
on this area and direction of effects differ across studies (Zhou et al.,
2007; Kim et al., 2009; Whitfield-Gabrieli et al., 2009; Ongur et al.,
Schizophrenia Research xxx (2012) xxx–xxx
⁎ Correspondence to: G. Mingoia, Interdisziplinäres Zentrum für Klinische Forschung
(IZKF), RWTH Aachen University, Aachen, Germany.
⁎⁎Correspondence to: I. Nenadic, Department of Psychiatry and Psychotherapy, Jena
University Hospital, Philosophenweg 3, D-07743 Jena, Germany. Tel.: +49 3641
9390127; fax: +49 3641 935410.
E-mail address: igor.nenadic@uni-jena.de (I. Nenadic).
SCHRES-04902; No of Pages 7
0920-9964/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.schres.2012.01.036
Contents lists available at SciVerse ScienceDirect
Schizophrenia Research
journal homepage: www.elsevier.com/locate/schres
Please cite this article as: Mingoia, G., et al., Default mode network activity in schizophrenia studied at resting state using probabilistic ICA,
Schizophr. Res. (2012), doi:10.1016/j.schres.2012.01.036

Page 2

2010;Woodwardetal.,2011).However,thesestudiesshowlinkstoboth
cognitive deficits and to symptoms (Rotarska-Jagiela et al., 2010), which
links them to the relevant pathophysiology of schizophrenia. Moreover,
other findings suggest a relative specificity of certain abnormalities for
schizophrenia (Calhoun et al., 2008).
In the present study, we aimed to analyse resting state data,
devoid of any directed cognitive task, using probabilistic independent
component analysis (pICA) in a cohort of chronic schizophrenia
patients in order to test the hypothesis of activity differences across
the nodes of the DMN. More specifically, we aimed to test that
prefrontal differences in resting state DMN activity are evident in
resting-state conditions in the absence of cognitive stimulation, and
in patients with remission of psychotic episode. We applied an algo-
rithm using an overall approach similar to Greicius (Greicius et al.,
2004, 2007), focussing on an automated detection/extraction of the
DMN component and group comparison, in order to eliminate the ne-
cessity for observer-dependent interventions such as placement of
seed regions. We tested the hypothesis of impaired frontal cortical
connectivity by studying remitted patients, i.e. not during a psychotic
episode, and correlated data with negative symptoms.
2. Methods
2.1. Study participants
We studied 25 patients with DSM-IV schizophrenia (8 female;
mean age 30 years, SD 7.3; age range 21–49 years) and 25 healthy
controls (10 female; mean age 29.1 years, SD 8.6; age range 22–
55 years), of which three patients were left-handed as determined
by the Edinburgh Handedness Scale (Oldfield, 1971). Proportion of
left-handers did not differ between groups (Fisher Exact Probability
Test p>0.23). All participants gave written informed consent to par-
ticipation in this study, which was conducted as part of the EUTwinsS
project (European Twin Study Network on Schizophrenia) and ap-
proved by the Ethics Committee of the Friedrich-Schiller-University
Medical School, Jena. None of the participants had a neurological con-
dition of history of traumatic brain injury or learning disability. We
carefully interviewed all participants to exclude candidates with a
present or previous neurological CNS condition, history of traumatic
brain injury, or learning disability. In addition, patient records were
reviewed, where available, to ensure patients did not meet any of
these exclusion criteria. According to chart review, none of the pa-
tients had a concurrent axis II disorder. In addition, all participants
were screened using the MWT-B, a standardised and widely-used
German test assessing (pre-morbid) IQ, to ensure that none of the
participants had an IQ below 80.
Patients were recruited from the in-patient and out-patient ser-
vices of the Department of Psychiatry and Psychotherapy in Jena. All
patients had a diagnosis of schizophrenia established using DSM-IV
criteria (American Psychiatric Association). A board-certified psychia-
trist (I.N.) assessed each patient, conducting an additional chart re-
view where necessary, and also rated current psychopathology
using the Scale for Assessment of Negative Symptoms (SANS), the
Scale for Assessment of Positive Symptoms (SAPS), and the Brief
Psychiatric Rating Scale (BPRS). All patients were remitted, i.e. none
of them was experiencing an acute psychotic episode at the time of
the study, and hence they showed mostly residual negative psychopa-
thology and only little positive symptoms.
Healthy control subjects were recruited from the local community
and matched to the patients with regard to age, gender, and handed-
ness (details of demographics and comparison are given in Table 1; T-
Test for age difference: p>0.66, two-tailed; Chi-square test for gen-
der: p>0.53; Fisher Exact Test for handedness: P>0.11, one-tailed).
They underwent a semi-structured interview to exclude personal
history or any current psychiatric disorder.
2.2. Data acquisition and pre-processing
We obtained resting-state fMRI series on a 3 T Siemens Tim Trio
system (Siemens, Erlangen, Germany) using the 12-channel head ma-
trix coil. Subjects were instructed to relax and keep their eyes closed
(without falling asleep, which was confirmed immediately after the
scanning session). Foam pads were used for positioning and immobi-
lisation of subjects' heads during scanning. We obtained a series of
210 T2*-weighted whole-brain volumes over approx. 9 min, using a
standard BOLD-sensitive EPI sequence (TR 2550 ms; TE 30 ms; flip
angle 90°; 45 contiguous axial slices with 3 mm thickness, no gap,
matrix 64×64; in-plane resolution of 3×3 mm; field-of-view
192 mm×192 mm). In addition, we acquired a high-resolution struc-
tural scan for co-registration using a 3D MPRAGE sequence with 192
contiguous sagittal slices of 1 mm thickness (TR 2300 ms; TE 3 ms; TI
900 ms; echo time 8.9 ms; flip angle 9°; matrix size 256×256; isotro-
pic voxel dimensions of 1×1×1 mm).
Both functional and structural images series underwent a quality
assurance protocol, including visual inspection, and none of the par-
ticipants showed such artefacts.
Data analysis was performed using SPM5 (Institute of Neurology,
London, UK; www.fil.ion.ucl.ac.uk/spm) for pre-processing as well
as later voxel-wise statistics, and FSL MELODIC for independent com-
ponent analysis (FMRIB, University of Oxford, UK; www.fmrib.ox.ac.
uk/fsl/melodic2/index.html).
We first discarded the first three images of each functional series to
avoid T1 saturation effects. In order to remove movement artefact, im-
ages were realigned using a least-squares approach and a 6-parameter
rigid body spatial transformation. A two-pass procedure was used to
registerthe images to themeanof theimagesafterthefirst realignment.
We applied smoothing with a 4 mm FWHM Gaussian kernel before es-
timating therealignment parameters. None of the participants exceeded
the pre-defined movement limits (3 mm translation on x, y, or z axis or
3° rotation), which were also part of the quality assurance protocol.
Within-subject registration was performed between functional im-
ages (used asreference image) and the anatomical image. Then, the co-
registered anatomical images was segmented using tissue probability
maps of the ICBM template (International Consortium for Brain Map-
ping; based on T1 scans of 452 subjects; http://www.loni.ucla.edu/
ICBM/ ICBM_TissueProb.html), which were aligned with an atlas
space, corrected for scan inhomogeneities, and classified into grey mat-
ter, white matter, and CSF. These data were then registered with affine
transformation to MNI space and down-sampled to 2 mm resolution.
Functional images were then spatially normalised to Talairach and
Tournoux space using spatial normalisation parameters estimated in
the segmentation process. Images were re-sampled to 2 mm using
sinc interpolation, and then smoothed with an 8 mm FWHM Gaussian
kernel to account for residual inter-subject anatomical differences.
2.3. Probabilistic ICA and automated extraction of DMN component
We applied independent component analysis (ICA) using FSL soft-
ware. For each subject, pre-processed functional images were
Table 1
Demographical data of subject samples.
Patients Controls
Number
Gender
Age (mean±SD)
SANS total score
(mean±SD)
SAPS total score
(mean±SD)
BPRS total score
(mean±SD)
25
8 females, 17males
30±7.3 years
40.3 (14.5)
25
10 females, 15 males
29.1±8.6 years
N/A
21.8 (11.7)N/A
38.9 (7.3)N/A
2
G. Mingoia et al. / Schizophrenia Research xxx (2012) xxx–xxx
Please cite this article as: Mingoia, G., et al., Default mode network activity in schizophrenia studied at resting state using probabilistic ICA,
Schizophr. Res. (2012), doi:10.1016/j.schres.2012.01.036

Page 3

concatenated across time in a single 4D image. The MELODIC algo-
rithm of FSL applied uses probabilistic ICA and is suited to auto-
matically estimate the number of relevant noise and signal sources
in the data. Due to this “noise model” it is possible to assign a signif-
icance value (p value) to the output spatial maps (Beckmann and
Smith, 2004). The data is decomposed into a set of spatially indepen-
dent maps, each with an internally consistent temporal dynamic
characterised by a time course (McKeown et al., 1998). Probabilistic
ICA (pICA) provides intensity values (z scores) and thus a measure
of the contribution of the time course of a component to the signal
in a given voxel. The individual spatial maps are therefore not mere
correlations between the time course in a given voxel and the (select-
ed) independent component, but rather the result of a multiple re-
gression model. MELODIC therefore allows not only segregation of
functional networks but also a voxel-wise map of quantitative mea-
sures of functional connectivity.
We analysed each subject's 4D image using the “FSL single-session
ICA” model in MELODIC. We first applied to the time series a high-
pass filter (f>0.009 Hz) to remove low frequency drifts which can
significantly contribute to the overall variance of an individual voxel's
time course (Beckmann and Smith, 2004), and a low-pass filter
(fb0.18 Hz) to remove cardiac and breathing artefacts. We chose to
set the ICA analysis to deliver 30 components, based on the fact that
this approximates one seventh of the number of time-points in the
scans.
The presence of various artefacts as Nyquist ghosting, head motion
and large blood vessels strongly impacts on standard GLM analysis,
which could induce additional error variance (e.g. in the case of
movement-related artefacts) or show false positives (e.g. in the case
of susceptibility-related artefact). Based on the Laplace approxima-
tion of the Bayesian model evidence, the pICA approach estimates
components, including artefacts. The implemented MELODIC algo-
rithm can then pick out different activation and artefactual compo-
nents without any explicit time series model being specified
(Beckmann and Smith, 2005). Hence, the pICA approach makes it
feasible to separate uninteresting physiological noise from other ef-
fects such as resting-state maps even in cases where the physiolog-
ical noise fluctuations become aliased in the temporal domain
(Beckmann et al., 2005). The components relating to artefacts
were removed and only the artefact not related components was in-
cluded for further analysis.
We then used an automated algorithm to select the component
reflecting the DMN best. For this purpose, we used a modification of
the protocol devised by Greicius and colleagues applied in previous
studies (Greicius et al., 2004, 2007, 2008). This was done using an
in-house script developed in MATLAB. First, we anatomically defined
the DMN based on the initial studies of Shulman and colleagues and
subsequent studies by Raichle et al., including 14 brain areas for
which we created a template defining 6 mm-diameter spheres to
the centroids of each area (Shulman et al., 1997; Raichle et al.,
2001). We then applied this mask image to all components (spatial
maps) from the pICA analysis and masked the images with an inverse
template of the DMN. We then extracted the mean z score of the
masked ICA components and computed the difference between the
average z score of voxels falling within the DMN template mask
minus the average z score of voxels outside the template. The ICA
component with the highest difference was then selected as reflect-
ing the individual subject's DMN component. The advantage of this
algorithm is that the ICA component, which corresponds to the
DMN, is selected automatically, and that the computed difference
score also provides an index of goodness of fit. Hence, this analysis
serves to select and validate one ICA component, for which we can as-
sume that it reflects the DMN component of the resting state signal
fluctuations based on the anatomical definition of ROIs; yet, p values
are given in each voxel of the (whole-brain) statistical map of this
component (not only the DMN template mask).
2.4. Group statistical analyses
First, and preceding the main analyses, we compared the index of
goodness of fit between the two groups. This was aimed to assess
whether there is a systematic difference in the selection of the DMN
component, by which the variation in goodness of fit might indicate
fundamental differences of the DMN between the two groups. Group
differences in goodness-of-fit would have implications for both the
DMN group comparison itself and the overall integrity of the DMN.
Also, in order to assess the DMN in each group, we performed a
voxel-wise random effects analysis using a one-sample t-test
(pb0.05 FDR corrected) in SPM, separately for each group. This was
done to demonstrate the configuration and extent of the DMN com-
ponent derived from pICA for patients and controls separately.
Second and third, our main analyses were the group comparison
of the DMN component with an ROI-based and voxel-wise compari-
son, respectively. For this purpose, we first pooled the pICA-derived
DMN component for all subjects into a second level analysis at
pb0.05 (uncorrected), and used the resulting statistical map as an in-
clusive mask to limit subsequent comparisons to those areas/voxels,
which could be assumed to be significantly involved in DMN activity
(based on the total cohort of subjects). We then performed an ROI
analysis based on the 14 pre-defined regions derived from previous
studies of the DMN (Shulman et al., 1997; Raichle et al., 2001).
pICA-derived values were averaged across the voxels of each ROI
and compared using SPSS 18 applying Bonferroni correction for 14
comparisons. Subsequently, we then performed a voxel-wise com-
parison of the two diagnostic groups in SPM using a two-sample t-
test (pb0.05, FDR corrected).
Finally, we performed a correlation analysis to test associations of
altered DMN activity with psychopathology. For this, we focussed on
the negative symptoms, which were prevalent across the patient
sample, given that our patients showed rather low levels of positive
symptoms (not being in a psychotic episode). We performed correla-
tions of SANS total scores with DMN activity, restricting the analysis
to ROI and voxel-wise analyses for those areas that showed differ-
ences in the above group comparisons. For this purpose, we extracted
r values in individual pICA-derived DMN components of patients, av-
eraged across the cluster and then correlated values to SANS total
score using bivariate correlation analysis (Spearman's rho, imple-
mented in SPSS 18), thresholded at pb0.05 with FDR correction,
using a MATLAB script (www.sph.umich.edu/~nichols/FDR/).
3. Results
3.1. Comparison of goodness of fit for DMN component
In all subjects, we found a component with spatial features consis-
tent with the DMN template provided in the literature (Shulman et
al., 1997; Raichle and Snyder, 2007). Comparison of goodness of fit
index (as described above) between patients and controls did not
show significant differences (t-test, one tailed: p=0.293). The distri-
bution of the individual scores for goodness of fit to standard default-
mode network is shown in Supplementary Material 1. The voxel-wise
analysis of the DMN component for each group separately is shown in
Supplementary Material 2.
3.2. ROI-based and voxel-wise group comparison of DMN activity
Region of interest analysis showed significant differences in the ana-
tomical resting state connectivity pattern of areas (Fig. 1) with healthy
controls showing larger effects in the network's prefrontal and temporal
areas, including the left middle frontal gyrus (BA8; close to the superior
frontal junction), bilateral medial frontal gyri (BA9), bilateral superior
frontal gyri (BA6), left ACC (BA32) and superior temporal sulcus (BA21/
22). We found larger effects for schizophrenia patients in the right
3
G. Mingoia et al. / Schizophrenia Research xxx (2012) xxx–xxx
Please cite this article as: Mingoia, G., et al., Default mode network activity in schizophrenia studied at resting state using probabilistic ICA,
Schizophr. Res. (2012), doi:10.1016/j.schres.2012.01.036

Page 4

amygdala, middle frontal polar cortex (BA10), left gyrus rectus (BA11),
bilateral inferior frontal gyri (BA47) and finally the bilateral inferior tem-
poralgyri(BA20).TheresultsoftheROIbasedanalysis(includingtheco-
ordinates of the centroid of each area) are summarised in Table 2A.
In the voxel-wise analyses, we found significant increases in func-
tional connectivity in controls vs. patients in a big cluster including
bilateral medial frontal gyri (BA9; MNI: 5, 47, 30), left superior frontal
gyrus (BA6; MNI: −9, 27, 62) and left superior temporal gyrus (BA
21/22, MNI: −53, −28, −4), and in patients vs. controls in the
middle portion of the gyrus rectus (BA11; MNI: −9, 35, −17) and
the frontal polar cortex (BA10; MNI: −4, 72, −4). Results of the
voxel-wise analysis are summarised in Table 2B and cortical areas
are shown in Figs. 2 and 3, respectively.
3.3. Correlation of DMN and psychopathology
We restricted testing for correlations with psychopathology to
those areas where patients had stronger DMN activity effects than
Fig. 1. ROI-based group comparison of schizophrenia patients vs. healthy control subjects of the DMN nodes; ROIs with higher z scores in healthy controls in the upper panel, those
with higher z scores in schizophrenia patients in the lower panels (co-ordinates are given in Table 2A).
4
G. Mingoia et al. / Schizophrenia Research xxx (2012) xxx–xxx
Please cite this article as: Mingoia, G., et al., Default mode network activity in schizophrenia studied at resting state using probabilistic ICA,
Schizophr. Res. (2012), doi:10.1016/j.schres.2012.01.036

Page 5

healthy controls, and found no significant correlation with total SANS
scores in patients. We then performed an additional analysis of correla-
tionswiththefiveSANSsubscales.Applyingathresholdofpb0.05(FDR
corrected), we found significant negative correlations between the
frontal polar cortex DMN activity and SANS subscales “affective flatten-
ing or blunting” (ρ=52) and “alogia” (ρ=463), as well as a negative
correlation between the right inferior temporal gyrus and the “alogia”
subscale (ρ=546). There was no significant correlation between
SANS scores and DMN areas with lower effects/connectivity in patients.
4. Discussion
In this study we assessed the default mode network (DMN) at rest
in remitted patients with schizophrenia and healthy controls using a
completely automated algorithm for extraction of the DMN compo-
nent. Our findings can be summarised as revealing regional differ-
ences in the strength of effects, a relation to clinical symptoms, but
no general breakdown in the overall anatomical composition of the
network in schizophrenia.
Preceding our main analyses, the group comparison of goodness-
of-fit indicates consistent detection of the DMN both in patients and
controls without significant group differences. It therefore suggests
that there is no systematic group difference in extracting the DMN
and it also is an indicator of preserved overall architecture of the
DMN in schizophrenia. This is contrary to other neuropsychiatric dis-
orders such as Alzheimer's disease, where resting state analyses have
goodness of fit for DMN analyses to actually distinguish patients from
controls (Greicius et al., 2004). One previous report using ICA to
compare schizophrenia patients and healthy controls did find a stron-
ger correlation of the healthy subjects' network with their default
mode network template (Garrity et al., 2007). Comparing our findings
to other studies, however, one should note that many other DMN
studies have used functional MRI series from cognitive experiments
(rather than resting-state data); so far, it is not clear whether DMN
extraction and/or further analysis might be affected by this methodo-
logical difference and how spontaneous fluctuations might interact
with amplitude changes caused by specific cognitive demands of a
task. Of course, it is unclear whether such effects might be detected
in larger sample sizes.
Our main analysis, on the other hand, provides evidence of differ-
ences in the regional contribution or strength of connectivity be-
tween groups, and hence DMN activity itself. Three findings are of
particular interest. Firstly, we find several differences in prefrontal
cortices, areas that have repeatedly been associated with schizophre-
nia pathology and its resulting cognitive impairments (Minzenberg
et al., 2009). It is of interest to note that these differences include op-
posite effects for two sub-components of the DMN: while healthy
controls show higher effects in dorsal prefrontal and temporal areas,
patients show higher effects in some ventral areas of the prefrontal
cortex, including the orbitofrontal cortex. This is an interesting pat-
tern as it segregates different dysfunctions in schizophrenia: those as-
sociated with superior and middle frontal gyrus pathology (mostly
cognitive), and those of the orbitofrontal cortex, often linked to affec-
tive flattening, impulsivity, and other clinical features of the disorder
Table 2
Group comparison of default mode network (DMN) activity at rest for A) region-of-
interest (ROI) and B) voxel-wise analyses. Co-ordinates indicate the cluster centroid
in MNI space.
Brodmann
area
Hemi-
sphere
Voxelsp Valuexyz
ROI-based analysis Bonferroni
HC>SZ
Superior frontal
gyrus
Superior frontal
gyrus
Medial frontal Gyrus
Superior temporal
sulcus
ACC
Middle frontal gyrus
BA 6L 3250.000
−9 2762
BA 6R 1060.000 1321 59
BA 9
BA 21/22
bilateral
L
976
221
0.001
0.016
5
−53
47
−28
30
−4
BA 32
BA 8
L
L
53
60
0.005
0.002
−16
−28
40
16
11
34
SZ>HC
Gyrus rectus
Frontal polar cortex
Inferior frontal
gyrus
Inferior frontal
gyrus
Inferior temporal
gyrus
Inferior temporal
gyrus
Amygdala
BA 11
BA 10
BA 47
Bilateral
Bilateral
L
287
166
51
0.001
0.038
0.017
−9
−4
−30
35
72
32
−17
−4
−8
BA 47R680.010 3436
−8
BA 20L 340.024
−62
−22
−28
BA 20R490.00042
−14
−28
R380.00814
−10
−24
Voxel-wise analysisFDR
HC>SZ
Superior frontal
gyrus
Medial frontal gyrus
Superior temporal
sulcus
BA 6L3250.000
−927 62
BA 9
BA 21/22
Bilateral
L
976
221
0.000
0.029
5
−53
47
−28
30
−4
SZ>HC
Gyrus rectus
Frontal polar cortex
BA 11
BA 10
Bilateral
Bilateral
287
166
0.001
0.010
−9
−4
35
72
−17
−4
Fig. 2. Voxel-wise comparison of DMN maps contrasting healthy control subjects vs.
schizophrenia patients at a threshold of pb0.005 uncorrected.
5
G. Mingoia et al. / Schizophrenia Research xxx (2012) xxx–xxx
Please cite this article as: Mingoia, G., et al., Default mode network activity in schizophrenia studied at resting state using probabilistic ICA,
Schizophr. Res. (2012), doi:10.1016/j.schres.2012.01.036

Page 6

(Bellani et al., 2010). A similar spatial pattern of “hyperconnectivity”
in orbitofrontal areas in patients (also extending to anterior frontopo-
lar areas) has recently been identified with another methodological
approach (Salvador et al., 2010). Given that some of these ventral
areas, including the frontopolar node are correlated with negative
psychopathology in our sample, this lends further support to the no-
tion that this part of the DMN shows aberrant connectivity in schizo-
phrenia and is relevant to its core pathophysiology. In contrast, our
dorsal and medial prefrontal frontal findings corroborate findings
from previous studies, which have also linked these abnormalities
to cognitive deficits (Zhou et al., 2007; Huang et al., 2009; Kim
et al., 2009), although with different methodologies. In this frame-
work of a ventral–dorsal dissociation of effects, single nodes might
be involved in different aspects related to either cognitive deficits or
clinical symptoms (Kim et al., 2009). Secondly, we find a significant
difference with patients showing higher effects in the amygdala. Dys-
function of this brain area has often been linked to both positive and
negative symptoms in schizophrenia, including effects on emotional
pathology (Aleman and Kahn, 2005). While we find higher effects in
DMN connectivity, it is interesting to note that several fMRI studies
using emotional and/or cognitive tasks have reported diminished
amygdala activation in schizophrenia. For example, a recent study in-
vestigating amygdala reactivity to negative facial stimuli revealed a
dissociation of activation difference related to the emotional task
(Rasetti et al., 2009). This effect appears to be related to clinical
state rather than genetic risk. Although our method does not provide
a direct assessment of baseline amygdala functioning (as measured
by perfusion or metabolic scanning), but rather yields a measure of
connectivity derived from pICA, it still appears conceivable that an in-
creased baseline activity might explain the diminished effect during
activation (i.e. smaller difference between baseline and activation
state). This interpretation is also consistent with previous PET studies
detecting elevated amygdala baseline activity in schizophrenia
(Taylor et al., 2005). Post-hoc testing revealed that the amygdala ef-
fect was also present in the left amygdala, but failed to reach statisti-
cal significance. Thirdly, it is interesting to note that our correlations
with psychopathology are restricted to those areas where patients
show higher DMN connectivity effects, especially in the ventral
areas of the frontal pole. As discussed above, this might be related
to the fact that the group differences in the dorsal frontal areas are
rather related to cognitive deficits of the disorder rather than (nega-
tive) symptoms.
Finally, we need to consider a few limitations of the study. Al-
though our automated algorithm for extraction of the DMN compo-
nentrelieson well-established
(Shulman et al., 1997), thus making it susceptible to refinement of
the DMN anatomical definition, this definition served well to reliably
extract the DMN component in all subjects. Also, our voxel-wise anal-
ysis of all subjects included allows consideration of the potential var-
iability of the DMN in this cohort studied. Secondly, we need to
consider the fact that patients were on antipsychotic medication, sus-
pension of which would not have been justified on ethical grounds. A
recent study reported effects of olanzapine on resting state activity
(Sambataro et al., 2010), however, data were acquired during a cogni-
tive fMRI experiment (rather than resting state).
In conclusion, our study demonstrates the successful use of a fully
automated algorithm for the detection and reliable extraction of DMN
activity from resting-state fMRI data in schizophrenia. Our findings
suggest that schizophrenia is associated with regional difference
(both hyper- and hypoconnectivity) of single nodes of the default
mode network, which appear to be related to specific core symptoms
of the disorder.
Supplementary materials related to this article can be found on-
line at doi:10.1016/j.schres.2012.01.036.
findingsof previous studies
Role of funding source
Funding for this study was provided by grants from the EU (FP6, RTN: EUTwinsS
network), and TMWFK. None of these institutions had further roles in study design;
in the collection, analysis and interpretation of data; in the writing of the report; and
in the decision to submit the paper for publication.
Contributors
I.N., G.M., C.G., and H.S. designed the study.
I.N., St.S., and H.S., contributed to patient recruitment and scanning.
I.N., K.L., M.D., H.P.B., C.G., J.R.R., R.G.M.S., and H.S. contributed to the data collec-
tion, processing, and pre-processing.
G.M., G.W., C.G., R.M., and I.N. contributed to implementation of the image proces-
sing pipeline and imaging data analysis.
G.M. and I.N. wrote the first drafts of the manuscript and all authors commented
on and approved the final version.
Conflict of interest
The authors declare that they have no relevant conflicts of interest that might in-
fluence the study design, data acquisition, interpretation, or other parts of this work.
Acknowledgements
This work was supported by grants from the EU (FP6, RTN: EUTwinsS; MRTN-
CT-2006-035987) and the Thuringian Government (TKM; B514-07004).
References
Aleman, A., Kahn, R.S., 2005. Strange feelings: do amygdala abnormalities dysregulate
the emotional brain in schizophrenia? Prog. Neurobiol. 77, 283–298.
Auer, D.P., 2008. Spontaneous low-frequency blood oxygenation level-dependent
fluctuations and functional connectivity analysis of the ‘resting’ brain. Magn.
Reson. Imaging 26, 1055–1064.
Fig. 3. Voxel-wise comparison of DMN maps contrasting schizophrenia patients vs.
healthy control subjects at a threshold of pb0.005 uncorrected.
6
G. Mingoia et al. / Schizophrenia Research xxx (2012) xxx–xxx
Please cite this article as: Mingoia, G., et al., Default mode network activity in schizophrenia studied at resting state using probabilistic ICA,
Schizophr. Res. (2012), doi:10.1016/j.schres.2012.01.036

Page 7

Beckmann, C.F., Smith, S.M., 2004. Probabilistic independent component analysis for
functional magnetic resonance imaging. IEEE Trans. Med. Imaging 23, 137–152.
Beckmann, C.F., Smith, S.M., 2005. Tensorial extensions of independent component
analysis for multisubject FMRI analysis. NeuroImage 25, 294–311.
Beckmann, C.F., DeLuca, M., Devlin, J.T., Smith, S.M., 2005. Investigations into resting-state
connectivity using independent component analysis. Philos. Trans. R Soc. Lond. B
Biol. Sci. 360, 1001–1013.
Bellani, M., Cerruti, S., Brambilla, P., 2010. Orbitofrontal cortex abnormalities in
schizophrenia. Epidemiol. Psichiatr. Soc. 19, 23–25.
Bluhm, R.L., Miller, J., Lanius, R.A., Osuch, E.A., Boksman, K., Neufeld, R.W., Theberge, J.,
Schaefer, B., Williamson, P., 2007. Spontaneous low-frequency fluctuations in the
BOLD signal in schizophrenic patients: anomalies in the default network.
Schizophr. Bull. 33, 1004–1012.
Broyd, S.J., Demanuele, C., Debener, S., Helps, S.K., James, C.J., Sonuga-Barke, E.J., 2009.
Default-mode brain dysfunction in mental disorders: a systematic review.
Neurosci. Biobehav. Rev. 33, 279–296.
Calhoun, V.D., Maciejewski, P.K., Pearlson, G.D., Kiehl, K.A., 2008. Temporal lobe and
“default” hemodynamic brain modes discriminate between schizophrenia and
bipolar disorder. Hum. Brain Mapp. 29, 1265–1275.
Damoiseaux, J.S., Rombouts, S.A., Barkhof, F., Scheltens, P., Stam, C.J., Smith, S.M.,
Beckmann, C.F., 2006. Consistent resting-state networks across healthy subjects.
Proc. Natl. Acad. Sci. U. S. A. 103, 13848–13853.
Garrity, A.G., Pearlson, G.D., McKiernan, K., Lloyd, D., Kiehl, K.A., Calhoun, V.D., 2007.
Aberrant “default mode” functional connectivity in schizophrenia. Am. J. Psychiatry
164, 450–457.
Greicius, M., 2008. Resting-state functional connectivity in neuropsychiatric disorders.
Curr. Opin. Neurol. 21, 424–430.
Greicius, M.D., Srivastava, G., Reiss, A.L., Menon, V., 2004. Default-mode network activity
distinguishes Alzheimer's disease from healthy aging: evidence from functional MRI.
Proc. Natl. Acad. Sci. U. S. A. 101, 4637–4642.
Greicius, M.D., Flores, B.H., Menon, V., Glover, G.H., Solvason, H.B., Kenna, H., Reiss, A.L.,
Schatzberg, A.F., 2007. Resting-state functional connectivity in major depression:
abnormally increased contributions from subgenual cingulate cortex and
thalamus. Biol. Psychiatry 62, 429–437.
Greicius, M.D., Kiviniemi, V., Tervonen, O., Vainionpaa, V., Alahuhta, S., Reiss, A.L.,
Menon,V., 2008. Persistentdefault-mode network connectivity duringlight sedation.
Hum. Brain Mapp. 29, 839–847.
Hoptman, M.J., Zuo, X.N., Butler, P.D., Javitt, D.C., D'Angelo, D., Mauro, C.J., Milham, M.P.,
2010. Amplitude of low-frequency oscillations in schizophrenia: a resting state
fMRI study. Schizophr. Res. 117, 13–20.
Huang, X.Q., Lui, S., Deng, W., Chan, R.C., Wu, Q.Z., Jiang, L.J., Zhang, J.R., Jia, Z.Y., Li, X.L.,
Li, F., Chen, L., Li, T., Gong, Q.Y., 2009. Localization of cerebral functional deficits in
treatment-naive, first-episode schizophrenia using resting-state fMRI. NeuroImage
49, 2901–2906.
Kim, D.I., Manoach, D.S., Mathalon, D.H., Turner, J.A., Mannell, M., Brown, G.G., Ford,
J.M., Gollub, R.L., White, T., Wible, C., Belger, A., Bockholt, H.J., Clark, V.P.,
Lauriello, J., O'Leary, D., Mueller, B.A., Lim, K.O., Andreasen, N., Potkin, S.G.,
Calhoun, V.D., 2009. Dysregulation of working memory and default-mode net-
works in schizophrenia using independent component analysis, an fBIRN and
MCIC study. Hum. Brain Mapp. 30, 3795–3811.
Mantini, D., Perrucci, M.G., Del Gratta, C., Romani, G.L., Corbetta, M., 2007. Electrophysiolog-
ical signatures of resting state networks in the human brain. Proc. Natl. Acad. Sci. U. S. A.
104, 13170–13175.
McKeown, M.J., Jung, T.P., Makeig, S., Brown, G., Kindermann, S.S., Lee, T.W., Sejnowski,
T.J., 1998. Spatially independent activity patterns in functional MRI data during the
Stroop color-naming task. Proc. Natl. Acad. Sci. U. S. A. 95, 803–810.
Minzenberg, M.J., Laird, A.R., Thelen, S., Carter, C.S., Glahn, D.C., 2009. Meta-analysis of
41 functional neuroimaging studies of executive function in schizophrenia. Arch.
Gen. Psychiatry 66, 811–822.
Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburgh inventory.
Neuropsychologia 9, 97–113.
Ongur, D., Lundy, M., Greenhouse, I., Shinn, A.K., Menon, V., Cohen, B.M., Renshaw, P.F.,
2010. Default mode network abnormalities in bipolar disorder and schizophrenia.
Psychiatry Res. 183, 59–68.
Raichle, M.E., Snyder, A.Z., 2007. A default mode of brain function: a brief history of an
evolving idea. NeuroImage 37, 1083–1090 (discussion 1097–1089).
Raichle, M.E., MacLeod, A.M., Snyder, A.Z., Powers, W.J., Gusnard, D.A., Shulman, G.L.,
2001. A default mode of brain function. Proc. Natl. Acad. Sci. U. S. A. 98, 676–682.
Rasetti, R., Mattay, V.S., Wiedholz, L.M., Kolachana, B.S., Hariri, A.R., Callicott, J.H.,
Meyer-Lindenberg, A., Weinberger, D.R., 2009. Evidence that altered amygdala
activity in schizophrenia is related to clinical state and not genetic risk. Am. J. Psychiatry
166, 216–225.
Rotarska-Jagiela, A., van de Ven, V., Oertel-Knochel, V., Uhlhaas, P.J., Vogeley, K., Linden,
D.E., 2010. Resting-state functional network correlates of psychotic symptoms in
schizophrenia. Schizophr. Res. 117, 21–30.
Salvador, R., Sarro, S., Gomar, J.J., Ortiz-Gil, J., Vila, F., Capdevila, A., Bullmore, E.,
McKenna, P.J., Pomarol-Clotet, E., 2010. Overall brain connectivity maps show
cortico-subcortical abnormalities inschizophrenia.Hum. Brain Mapp. 31, 2003–2014.
Sambataro, F., Blasi, G., Fazio, L., Caforio, G., Taurisano, P., Romano, R., Di Giorgio, A., Gelao,
B., Lo Bianco, L., Papazacharias, A., Popolizio, T., Nardini, M., Bertolino, A., 2010. Treat-
ment with olanzapine is associated with modulation of the default mode network in
patients with schizophrenia. Neuropsychopharmacology 35, 904–912.
Shulman, G.L., Fiez, J.A., Corbetta, M., Buckner, R.L., Miezin, F.M., Raichle, M.E., Petersen,
S.E., 1997. Common blood flow changes across visual tasks: II. Decreases in
cerebral cortex. J. Cogn. Neurosci. 9, 648–663.
Taylor, S.F., Phan, K.L., Britton, J.C., Liberzon, I., 2005. Neural response to emotional
salience in schizophrenia. Neuropsychopharmacology 30, 984–995.
van den Heuvel, M.P., Hulshoff Pol, H.E., 2010. Exploring the brain network: a review
on resting-state fMRI functional connectivity. Eur. Neuropsychopharmacol. 20,
519–534.
Whitfield-Gabrieli, S., Thermenos, H.W., Milanovic, S., Tsuang, M.T., Faraone, S.V.,
McCarley, R.W., Shenton, M.E., Green, A.I., Nieto-Castanon, A., LaViolette, P.,
Wojcik, J., Gabrieli, J.D., Seidman, L.J., 2009. Hyperactivity and hyperconnectivity
of the default network in schizophrenia and in first-degree relatives of persons
with schizophrenia. Proc. Natl. Acad. Sci. U. S. A. 106, 1279–1284.
Woodward, N.D., Rogers, B., Heckers, S., 2011. Functional resting-state networks are
differentially affected in schizophrenia. Schizophr. Res. 130, 86–93.
Zhou, Y., Liang, M., Tian, L., Wang, K., Hao, Y., Liu, H., Liu, Z., Jiang, T., 2007. Functional
disintegration in paranoid schizophrenia using resting-state fMRI. Schizophr. Res.
97, 194–205.
7
G. Mingoia et al. / Schizophrenia Research xxx (2012) xxx–xxx
Please cite this article as: Mingoia, G., et al., Default mode network activity in schizophrenia studied at resting state using probabilistic ICA,
Schizophr. Res. (2012), doi:10.1016/j.schres.2012.01.036