r Human Brain Mapping 30:3865–3886 (2009) r

Functional Segmentation of the Brain Cortex

Using High Model Order Group PICA

Vesa Kiviniemi,1* Tuomo Starck,1 Jukka Remes,1 Xiangyu Long,1,2

Juha Nikkinen,1 Marianne Haapea,1,3 Juha Veijola,3,5 Irma Moilanen,4
Matti Isohanni,3 Yu-Feng Zang,2 and Osmo Tervonen1
1
Department of Diagnostic Radiology, Oulu University Hospital, Oulu, Finland
2
State Key Laboratory of Cognitive Neuroscience, Beijing Normal University, Beijing, China
3
Department of Psychiatry, Oulu University, Finland
4
Department of Child Psychiatry, Oulu University Hospital, Oulu, Finland
5
Academy of Finland, Helsinki, Finland

r r

Abstract: Baseline activity of resting state brain networks (RSN) in a resting subject has become one of
the fastest growing research topics in neuroimaging. It has been shown that up to 12 RSNs can be dif-
ferentiated using an independent component analysis (ICA) of the blood oxygen level dependent
(BOLD) resting state data. In this study, we investigate how many RSN signal sources can be separated
from the entire brain cortex using high dimension ICA analysis from a group dataset. Group data
from 55 subjects was analyzed using temporal concatenation and a probabilistic independent compo-
nent analysis algorithm. ICA repeatability testing verified that 60 of the 70 computed components
were robustly detectable. Forty-two independent signal sources were identifiable as RSN, and 28 were
related to artifacts or other noninterest sources (non-RSN). The depicted RSNs bore a closer match to
functional neuroanatomy than the previously reported RSN components. The non-RSN sources have
significantly lower temporal intersource connectivity than the RSN (P < 0.0003). We conclude that the
high model order ICA of the group BOLD data enables functional segmentation of the brain cortex.
The method enables new approaches to causality and connectivity analysis with more specific anatomi-
cal details. Hum Brain Mapp 30:3865–3886, 2009. VC 2009 Wiley-Liss, Inc.

Key words: ICA; fMRI; resting state; brain cortex

r r

INTRODUCTION
Additional Supporting Information may be found in the online
version of this article. Because the discovery of functionally connected low
frequency fluctuations of the blood oxygen dependent
Contract grant sponsor: Finnish Academy-Chinese NSFC
Collaboration NEURO-Program; Contract grant number: 117111; (BOLD) with a functional MRI by Biswal et al. [1995], the
Contract grant sponsor: Finnish Medical Foundation; Contract detection of baseline activity within functional brain net-
grant sponsor: Finnish Neurological Association. works has become a fast growing interest area in brain
*Correspondence to: Vesa Kiviniemi, Department of Diagnostic imaging research [Fox and Raichle, 2007; Vincent et al.,
Radiology, Oulu University Hospital, Oulu, Finland. 2007]. Intracortical local field potential and multiunit activ-
E-mail: vesa.kiviniemi@oulu.fi ity fluctuations partly correlate with the fluctuations of the
Received for publication 24 October 2008; Accepted 1 April 2009 BOLD signal with a six second lag in anesthetized condi-
DOI: 10.1002/hbm.20813 tions [Logothetis et al., 2001; Shmuel and Leopold, 2008].
Published online 8 June 2009 in Wiley InterScience (www. In addition to the electrophysiological activity, metabolic
interscience.wiley.com). and vasomotor effects partly explain the detected changes

C 2009 Wiley-Liss, Inc.

V
 r Kiviniemi et al. r

of BOLD fluctuation [Fukunaga et al., 2008; Kannurpatti MATERIALS AND METHODS

et al., 2008; Kiviniemi, 2008].
The combined effects of these fluctuations in neuronal The ethical committee of Oulu University Hospital has
networks have been shown to be differentiable from noise approved the studies for which the subjects have been
during normal, awake resting conditions. Independent recruited, and informed consent has been obtained from
component analysis (ICA) is an effective tool in separating each subject individually according the Helsinki declara-
statistically independent source signals of the BOLD data. tion. Fifty-five control subjects were chosen (age 24.96

ICA separates various sources of the fMRI signal by maxi- 5.25 years, 32 females, 23 males) from three resting state
mizing the non-Gaussianity of the source signals. Spatial studies: a 1986 birth cohort study of ADHD and psychosis;
domain ICA (sICA) can separate BOLD signal sources that brain tumor resting state study, total n ¼ 200. Subjects
represent reactions to externally cued task-activations, were imaged on a GE 1.5T HDX scanner equipped with
background activity within functional brain (i.e. resting an eight-channel head coil using parallel imaging with an
state) networks (RSN), and various physiological noise acceleration factor 2. The scanning was done during Janu-
and artifact sources [Beckmann and Smith, 2004; Beck- ary 2007–May 2008. All subjects received identical instruc-
mann et al., 2005; Calhoun et al., 2001; Kiviniemi et al., tions: to simply rest and focus on a cross on an fMRI
2003; McKeown et al., 1998; Van de Ven, 2004]. ICA meth- dedicated screen which they saw through the mirror sys-
odology yields results that are equally accurate to other tem of the head coil. Hearing was protected using ear
contemporary methods of detecting large scale temporally plugs, and motion was minimized using soft pads fitted
coherent networks from the BOLD signal data [Long et al., over the ears.
2008]. The functional scanning was performed using an EPI
It has been suggested that 10 to 12 RSNs can be GRE sequence. The TR used was 1,800 ms and the TE was
detected from the brain cortex from resting state BOLD 40 ms. The whole brain was covered, using 28 oblique
data, using ICA with a component dimensionality, i.e. a axial slices 4 mm thick with a 0.4 mm space between the
model order, around 25–40 components [Beckmann et slices. FOV was 25.6 cm  25.6 cm with a 64  64 matrix,
al., 2005; De Luca et al., 2006; Damoiseaux et al., 2006]. and a flip angle of 90 . The resting state scan consisted of
However, the most recent reports on topographic delin- 253 functional volumes. The first three images were
eation of cortex and functional connectivity nodes show excluded due to T1 equilibrium effects. In all three studies,
that there should be more than a dozen detectable func- the resting state scanning started the protocols, and lasted
tional networks on the brain cortex [Bartels and Zeki, 7 min and 36 s. In addition to resting-state fMRI, T1-
2005; Cohen et al., 2008; Di Martino et al., 2008; Malinen weighted scans were taken with 3D FSPGR BRAVO
et al., 2007; Mezer et al., 2009; Pawela et al., 2008]. sequence (FOV 24.0 cm, matrix 256  256, slice thickness
When the model order of the ICA estimation is 1.0 mm, TR 12.1 ms, TE 5.2 ms, and flip angle 20 ) to
increased, the separated BOLD signal sources have been obtain anatomical images for co-registration of the fMRI
shown to split into several functional nodes [Li et al., data to standard space coordinates.
2007; Ma et al., 2007; Malinen et al., 2007; Eichele et al.,
2008]. McKeown et al. [2002] have shown that the
detected ICA components of BOLD data actually repre- Preprocessing of Imaging Data
sent deterministic signal sources of the data, even in
very high model orders. Head motion in the fMRI data was corrected using mul-
In this study, we evaluate how many independent RSN tiresolution rigid body co-registration of volumes, as
signal sources of baseline brain activity can be detected implemented in FSL 3.3 MCFLIRT software [Jenkinson et
from the brain cortex using a fairly large group of subjects al., 2002]. The default settings used were as follows: mid-
and a relatively high order group ICA. We show that the dle volume as reference, a three-stage search (8 mm rough
previously detected RSN signal sources can be differenti- þ 4 mm, initialized with 8 mm results þ 4 mm fine grain,
ated into a more finely tuned functional anatomy by initialized with the previous 4 mm step results) with final
extending the use of temporal connectivity, the frequency trilinear interpolation of voxel values, and normalized spa-
power spectral and the anatomical clustering characteris- tial correlation as the optimization cost function. Brain
tics of the signal sources into group analysis data. We extraction was carried out for motion corrected BOLD vol-
show that over 60 of the detected signal sources can be umes with optimization of the deforming smooth surface
robustly detected with repeatability analysis (ICASSO-soft- model, as implemented in FSL 3.3 BET software [Smith,
ware package: http://www.cis.hut.fi/projects/ica/icasso/) 2002] using threshold parameters f ¼ 0.5 and g ¼ 0, and
from group data. We found supporting evidence for our for 3D FSPGR volumes, using parameters f ¼ 0.25 and g ¼
hypothesis that true RSN sources have increased temporal 0. After brain extraction, the BOLD volumes were spatially
intersource connectivity compared to noise sources. Spatial smoothed; 7 mm FWHM Gaussian kernel and voxel time
overlap of the RSN components and predefined anatomi- series were detrended using a Gaussian linear high-pass
cal structures are presented. The possible functional and filter with a 125 second cutoff. The FSL 4.0 fslmaths tool
clinical implications of the results are discussed. was used for these steps.

r 3866 r
 r Functional Segmentation of Brain using Group-PICA r

Multiresolution affine co-registration as implemented in estimation suggested 73 components in an initial test, we

the FSL 4.0 FLIRT software [Jenkinson et al., 2002] was reduced it to 70 components due to better extraction of
used to co-register mean nonsmoothed fMRI volumes to 3D extracranial voxels after using a more sensitive BET algo-
FSPGR volumes of corresponding subjects, and 3D FSPGR rithm outside the MELODIC framework.
volumes to the Montreal Neurological Institute (MNI) The jointly whitened and reduced individual fMRI data
standard structural space template (MNI152_T1_2mm_- sets are further whitened, and then temporally concaten-
brain template included in FSL). Trilinear interpolation was ated. Further dimensionality reduction is performed with
used, a correlation ratio was used as the optimization cost principal components analysis (PCA) on the concatenated
function, and regarding the rotation parameters a search data. The number of dimensions in the concatenated data
was done in the full [p p] range. The resulting transforma- (number of subjects times 70) is reduced to the number of
tions and the trilinear interpolation were used to spatially independent components to be computed, 70 in this study.
standardize smoothed and filtered BOLD volumes to the 2 Finally, FastICA [Hyvärinen, 1999] is performed using
mm MNI standard space. Because an sICA was run later voxel values as samples, and PCA reduced and whitened
on fMRI data concatenated from the 55 subjects, in practice temporal dimensions as variables. In this study, we used
the spatial resolution of spatially standardized BOLD the default settings in the MELODIC software: symmetric
volumes had to be lowered to 4 mm. orthogonalization and skewness (‘‘pow3’’) as the contrast
function.
Spatial Domain Analysis In a PICA framework, estimated intensity maps (inde-
pendent components) are converted to Z-scores by
Analysis was carried out using probabilistic independ- dividing intensity values by voxel-wise standard devia-
ent component analysis (PICA) [Beckmann et al., 2004] on tions of noise distribution. Standard deviations are esti-
preprocessed and spatially standardized BOLD data, tem- mated through calculation of corresponding voxel-wise
porally concatenated from data sets of individual subjects. variances of the original preprocessed data in the sub-
The implementation of PICA and temporal concatenation space left out by dimensionality reduction steps. Proba-
in FSL 4.0 MELODIC software was used in this study. The bility distributions of values in individual resulting
default processing provided in MELODIC, and used in score maps are modeled using a mixture of functions, a
this study, starts on the subject level. Intensities of voxel zero mean Gaussian function and two gamma functions.
time courses in preprocessed fMRI volumes are converted The Gaussian function corresponds to the effects due to
to percentages of change with respect to the mean inten- chance, and the gamma functions model outliers not
sity in those voxels. This is followed by group level joined observed by chance. For a given Z-score, the sum of
normalization of voxel time course variances. The mean gamma function values divided by the sum of all func-
4D fMRI data set is averaged from individual data sets, tion values can be used to calculate the probability of
and the variance estimate is computed from it. The var- actually observing an effect with that score. With regard
iance estimate used in normalization is computed so that to these probabilities, a probability threshold (P-value)
(1) data with respect to its time points (temporal dimen- is set to leave out (set to zero) those voxel values in
sions) is whitened, (2) in whitened space, large values cor- score maps that are presumed to be observed in the
responding to variance, possibly not originating from PICA results by chance. In this study, we used P ¼ 0.5,
Gaussian processes, are noted with some threshold value, attributing equal relevance to both false negatives and
and their influence excluded from data in the original false positives.
space, after which (3) variance is computed ordinarily on The Juelich histological atlas [Eickhoff et al., 2007] and
resulting residual data. Within MELODIC implementation the Harvard-Oxford cortical and subcortical atlases (Har-
and in its joined normalization scheme, this threshold vard Center for Morphometric Analysis) provided with
value was set to 3.1. Fourth, voxels with insufficient var- the FSL4 software package were used to quantify anatom-
iances (<0.0001) are set to zero because they contain nearly ical characteristics of thresholded Z-score maps. For this
constant time courses, and in that sense a bad signal-to- purpose, maps were upsampled back to 2 mm resolution
noise ratio that would (after variance normalization) con- MNI standard space. Individual anatomical regions in the
tribute too much to the analysis. Each individual subject probabilistic atlases were binarized and used as masks to
fMRI data set is divided voxel-wise with a standard devia- extract the corresponding regions in maps. An FSL4
tion corresponding to the voxel-wise variance estimate. A fslstats tool was used to calculate mean and maximum in-
whitening matrix corresponding to the whitening mean tensity in regions, as well as the location of maximum in-
4D fMRI data set is computed and used for group-level tensity. The number of voxels in the masked PICA maps
joined whitening and dimensionality reduction of each was also calculated, and divided by the number of voxels
individual subject fMRI data sets. The number of resulting in the whole anatomical masking region, to provide a
dimensions corresponds to the model order of ICA that is percentage of the coverage of different anatomical areas
intended to be used. In this study, the model order was by individual maps.
chosen to be 70, in correspondence to the high order sICA A neuroradiologist (VK) depicted the thresholded PICA
modeling of the resting state BOLD data. Because PPCA maps corresponding to the RSNs by choosing anatomically

r 3867 r
 r Kiviniemi et al. r

clustered sources in the cortical regions, in the vicinity of In this study, these subject specific 250 point time courses
functional brain regions. The sources were further classi- were used to assess pair-wise inter component connectiv-
fied into three regions for illustration purposes with a ity on the group level. The time courses were first
scheme modified from Salvador et al. [2005] (peri-sylvian detrended, and then correlated with each other on the sub-
¼ rolandic and temporal, frontal, occipito-parietal). In each ject level with the Pearson product–moment correlation
of the selected sources (1) sources presented clustered coefficient with a zero lag. The group level average and
voxel groups, (2) sources were focused on cortical struc- standard deviation of the results were calculated for each
tures, and (3) time courses in the ICA mixing matrix corre- correlation pair’s absolute values.
sponding to the maps showed elevated low frequency In addition, the mean of these group-level absolute cor-
(<0.1 Hz) power. Artifactual non-RSNs were identified relation coefficients was calculated per component within
based on their motion-related location at the borders of the group of components corresponding to RSNs, and
the brain, in cerebrospinal fluid (CSF), at the proximity of within the group corresponding to non-RSNs. These val-
large blood vessels, in white matter, and in the vicinity ues were tested for differences with a two-sample t-test
of areas shown to be susceptible to physiological pulsa- between RSNs and non-RSNs.
tions with relatively increased or mixed high frequency
power [Birn et al., 2006; Lund et al., 2006].
Frequency Domain Analysis
Repeatability Analysis The subject-wise time courses extracted from the mixing
matrix were (after detrending) also converted to power
The data from the final step prior to FastICA was also
spectra, to assess contribution of low frequency power in
subjected to ICA repeatability analysis (ICASSO) [Himberg
them. For each time course, a spectrogram was computed
et al., 2004]. Because ICA is sensitive to its initialization,
with 128-point rectangular windowing in Matlab (R2008a)
this framework runs FastICA multiple times on the same
with the similarly named function. Rectangular window-
data, with random initializations and clusters with a hier-
ing was sufficient, because no sinc-interpolation (e.g. no
archical clustering algorithm, on the results from individ-
zero padding) was used in the Discrete Fourier Transform.
ual runs. The number of clusters is the same as the
The absolute values of the resulting 123 spectra were
number of components and provides statistics about the
raised to the second power to approximate power spectra,
quality of the clusters, reflecting the stability of the ICA
and they were averaged to obtain a final power spectrum
results. In this study, we used the provided Matlab imple-
estimate for each time course.
mentation [Himberg et al., 2004] with 100 runs, and a low
For the group-level analysis, each power spectrum was
epsilon threshold (0.0000001) so that optimization would
normalized by dividing its values by its total power. Nor-
not converge too soon before reaching extrema. Symmetric
malized power spectra from individual subjects were aver-
orthogonalization and skewness (‘‘skew’’) as the contrast
aged for each component, to produce a group-level
function were also used in this setting. The conservative
representation of temporal power. In addition, Singular
cluster quality index Iq was used to estimate individual
value decomposition was used for each component to pro-
component repeatability [Himberg et al., 2004]. Iq is one
duce, on a group-level, a rank-1 approximation of subject-
for an ideal cluster, and decreases when component clus-
specific power spectra for that component. Rank-1 approx-
ters become less compact and isolated in repeated ICA
imation depicts frequencies with the most commonly
runs.
elevated power in subject-level time courses.

Time Domain Analysis

A mixing matrix estimated in an sICA contains time RESULTS
courses corresponding to estimated spatial maps (inde- Repeatability Measures
pendent components). The time courses represent the
behavior of nonartifact components in time more accu- An ICASSO algorithm was run on the pre-whitened
rately than the average BOLD time courses computed PCA-data matrix provided by the MELODIC. After 100
from the same spatial area as in the maps; this is because repeated runs of the group ICA, significantly clustered
the most dominant effects of cardiac, respiratory, CSF-pul- components were detected. Fifty-one of the 70 ICs had an
sation and motion, and other artifacts are eliminated into Iq over 0.8, and 60 ICs still had an Iq of at least 0.7. Figure
separate components. This offers preprocessed time 1 shows clustering of the centroid components and stabil-
courses for temporal correlation analysis of BOLD signal ity indexes in reducing the stability order. PICA estimation
sources [Jafri et al., 2008; Sorg et al., 2007]. of the group data yielded 73 components. We made seven
In a temporal concatenation scheme, the mixing matrix separate runs altering the basic spatial smoothening
can be divided into subject specific consecutive segments, parameters (3–7 mm FWHM), temporal high (100–420 s)
reflecting subject-wise temporal behavior of components. and low-pass filtering. In essence, the spatial correlation

r 3868 r
 r Functional Segmentation of Brain using Group-PICA r

Figure 1.
On the left, the clustering of group PICA-derived components in ICASSO after 100 repeated
runs. On the right, ICASSO run components in descending clustering index values, showing that
in 60/70 components the clusters are repeatedly detectable (Iq  0.7). Please note that the
ICASSO cluster numbers are not the same as the PICA numbers in the text and other images.

between the detected ICA components of separate runs Anatomical Parcellation

was always >0.7.
In practice, most of the cerebral cortex can be shown to
be segmented with the IC components (c.f. the upper rows
in Figs. 2–4). On average, the maximal spatial correlation
RSN-Identification between detected RSN components and other sources was
0.2 (SD 0.04). This can be seen in Figures 2–4 as a limited
The group PICA separated 42 components that were
spatial overlap of separate signal source components. Table
identified as RSN. These 42 components cover most of the
I lists the first three to five largest overlapping anatomical
brain cortex excluding areas having susceptibility artefacts
segments used in FSL4 (Juelich and Oxford atlases) in com-
due to air sinuses. The corresponding RSN components
parison with the selected RSN signal sources. The RSN
had an average Iq ¼ 0.83 (
0.13) in the ICASSO analysis.
components actually show more connectivity nodes to mul-
The RSN components had elevated low frequency fluctua-
tiple other anatomical structures, and the complete table is
tion power and, spatially, were clustered components
presented in the Supporting Information material (Support-
located in the cortex. Figures 2–4 show fused maps and
ing Information Table I.2). In the following, the individual
separate spatial distributions of 25 of the different signal
RSN sources are evaluated, based on spatial overlap with
sources on both sides of the Sylvian fissure and the central preexisting anatomical segments. Temporal intersource
sulcus (peri-sylvian), in the frontal lobe (frontal), and in synchrony of the sources and the highest spatial correlation
the occital and parietal regions (occipito-parietal). The with ICASSO components are also presented.
remaining signal sources are shown in Figure 5 in more
optimal 3D presentations. Rolandic Signal Sources
Twenty-eight ICs were related to non-RSN sources.
Eighteen non-RSN components were identified as motion/ Figure 2 shows six signal sources identified as peri-
realignment artifacts, and physiological noise sources such rolandic, situated at or in the vicinity of the central sulcus.
as arterial and CSF pulsation. Ten of the discarded compo- Although most of the sources have strong localized con-
nents were detected as large voxel clusters in the white nectivity, some have long range connectivity to other pri-
matter and were not further analyzed. Ten examples of mary and secondary sensory areas, such as to the visual
the non-RSN sources are shown in Figure 6. and auditory regions.

r 3869 r
 r Kiviniemi et al. r

Figure 2.
Signal sources located around the Sylvian fissure, both peri-rolandic and temporal sources are
presented. At the top, a fused image presenting overlays of the different sources. Below that, IC
sources are mapped with color labels provided by the fsl-view tool. On the right, the mean
power spectra and rank 1 approximate on the same row as the map of the given source. On
the left, the sources are numbered using the same numbers as referenced in the text.

The motor cortex was divided into two different signal posterior to the typical SMA seen during activation stud-
sources resembling the classical separation of upper corti- ies. Another motor cortex component, IC 24, was depicted,
cal structures representing the hand areas (IC 27) and mid- but it was not clear whether such parcellation is due to
line feet areas (IC 19). Similar to the original discovery of physiological factors, or if the results are split due to posi-
Biswal et al. [1995], there was a midline connectivity node tioning differences in the original BOLD axial slices.

r 3870 r
 r Functional Segmentation of Brain using Group-PICA r

Figure 3.
The occipital and parietal RSN signal
sources presented in the same way as in
Fig. 2 with power spectra and rank 1
approximates. The DMN posterior part is
depicted in IC 38.

Anterior to motor areas there were unilateral components bilateral post-somatosensory cluster (Fig. 5, IC 40). The
(ICs 53 and 60) with positive connectivity to the cerebel- somatosensory signal source also showed smaller clusters
lum. The right anterior motor IC 60 was spatially larger of connectivity to anatomical visual, auditory and Broca’s
than the left IC 53. areas (c.f. Supporting Information Table I.2). The signal
Primary and secondary somatosensory cortices were source IC 14 was situated below the somatosensory cortex
separated into signal sources of their own (IC 6 and, 43, in the white-gray matter boundary. The secondary somato-
respectively). Posterior to these, there was a symmetrical, sensory source (IC 43) also showed connectivity with

r 3871 r
 r Kiviniemi et al. r

Figure 4.
Examples of frontal signal sources in a similar order to Figures 2 and 3. Note the relative lack of
power peaks compared to the more posterior components. The default mode network pattern
can be seen to originate from several subnetworks in the fused images at the top.

structures in the cerebellum, in addition to secondary Temporal Lobe Signal Sources

somatosensory, auditory, motor, and premotor anatomical
structures. The supplementary motor area (SMA, i.e. juxta- In Figure 2, the primary auditory cortex was identified
positional lobule) area had bilateral connectivity clusters as a bilateral component of its own, with a right brain
in IC 49, primarily involving the insular cortices, and dominance (IC 25). As with the peri-rolandic sources, this
partly involving the secondary somatosensory area, resem- large network presented connectivity with other sources in
bling mirror neuron networks. Interestingly, both premo- distant regions, in both the motor and visual areas. A Bro-
tor areas BA6 also presented unilateral components, i.e., ca’s area on the left and an homologous right hemisphere
ICs 53 and 60 (see Fig. 5). There was a tendency that the Broca’s area were detected in a unique bilateral compo-
farther the source was from the central sulcus the longer nent, with a clear, functionally connected cluster in the
the distances were between detected connectivity clusters. right paracingulate areas, posterior cingulate gyrus, and

r 3872 r
 r Functional Segmentation of Brain using Group-PICA r

Figure 5.
Examples of the remaining 17 RSN in cor-
onal, axial, and sagital views on an MNI
template. The MNI coordinates are
shown in the image. Red-yellow positive
Z-scores, and blue negative Z-scores.

bilateral occipital poles (IC 30). The fronto-insular cortex present at the lower frontal pole and medial frontal cortex,
of the salience network was detected in a bilateral IC 15 and IC 8 was related to the upper frontal pole, and the
with a partial Broca’s area and inferior frontal gyrus middle and superior frontal and angular gyri. IC 18 was
involvement, in addition to paracingulate gyrus connectiv- connected to the superior frontal gyrus, paracingulate, and
ity [Shihadran et al., 2008]. IC 49 showing connectivity to cingulate gyri. IC 23 was connected to the angular and
SMA could be differentiated from the salience network. supramarginal gyri, along with the visual cortex. IC 37
The temporal poles (IC 13, Fig. 5) were also detected in a was related to the superior and medial frontal gyri and
unique signal source, with connectivity to midline struc- the paracingulate gyri.
tures to bilateral hippocampi.

Occipital and Parietal Signal Sources

Frontal Signal Sources
There were seven separated signal source clusters in the
The frontal lobe had 11 RSN signal sources, seven of occipital lobe, all spatially connected to visual areas (c.f.
which are presented in Figure 3. In five sources (ICs 7, 10, Fig. 4). IC 2 was situated at the lateral part of the primary
18, 31, and 37), connectivity to language-related areas, visual cortex (V1), as in the lateral visual component in
such as the auditory cortex, inferior frontal gyrus, and/or previous papers. Above and lateral to it was IC 4, with
Broca’s anatomical areas, was detected. In addition to midline connectivity in V5. IC 39 was also comprised of
these areas, IC 7 and 10 were also connected to the ante- bilateral V5 areas, yet different V5 areas from those in IC
rior cingulate and paracingulate gyri. IC 7 had connectiv- 4, with connectivity to V1 and 2 and the inferior and
ity to the angular gyri and the visual cortex. IC 10 was medial temporal gyrus. IC 59 typically presented the

r 3873 r
 r Kiviniemi et al. r

Figure 6.
Ten artefactual non-RSN sources, from top left: brain stem pulsation (IC 11), motion artefact (IC
18), two white matter sources (ICs 16 and 17), cerebral artery related (IC 28), sagital sinus (IC
46), two CSF-ventricle pulsation sources (ICs 51 and 56) and temporal motion-related source
(IC 65) and frontal sinus susceptibility artifacts (IC 66).

previously detected medial visual RSN, comprising almost Buckner et al., 2008]. IC 17 resembles the frontal areas cen-
completely of the intracalcarine and supracalcarine cortex, tral executive network [Shihadran et al., 2008]. Figure 7
the lingual cortex, and the major parts of V1 and 2 and the shows an example of PICA images calculated with 100
cuneal cortex. An anterior medial visual component IC 32 repeated ICASSO runs with model orders 10, 70, and 150.
in V2, with hippocampal and geniculate body connectivity, Note how the core areas of the default mode divide into
was also detected. A posterior component IC 21 was sepa- smaller subnetworks with increasing model order.
rated from the medial central IC 59. The cuneal and precu-
neal cortex was dominant in IC 63, along with V1&2, with
clear retrosplenial connectivity. In addition, there were Time Domain Synchrony
two anticorrelated components in the visual areas (ICs 52
and 61) showing positive Z-scores caudally on the other The intersource connectivity between all the calculated
visual cortex, and negative Z-scores on the contralateral ICs are presented in a matrix in Figure 8. The graph shows
side more cranially. how the individual sources are connected to the other
sources on average with zero time lag. The correlation
coefficients range from 0.64 to 0.66 (
0.21). The mean
Default Mode and Frontal Attention correlation coefficients of each component provide an
Control Sources overall assessment of general intersource connectivity of a
given IC. The overall connectivity of the non-RSN sources
Several ICA sources were detected in the areas known was statistically significantly lower than the connectivity
to be related to the default mode network (DMN). IC 10, of the RSN sources (t ¼ 4.6, P < 0.000024, c.f. Fig. 8 bot-
ventromedial prefrontal cortex (vmPFC); IC 38, precuneus tom). Eight out of the 10 lowest values of summed correla-
and parietal lobule (PCC). ICs 7, 10, 23, 18, and 37 were tion were depicted in non-RSN sources. There is, however,
closely related to anatomical structures of the DMN (c.f. some overlap in the correlation values between the RSN
Figs. 3 and 4). The regions usually detected as anti-corre- and non-RSN sources, and temporal zero-lagged correla-
lated to DMN in correlation analysis were also detected as tion alone cannot be used to separate the sources with
sources of their own, i.e. ICs 6, 40 and 49 [Fox et al., 2007; absolute certainty.

r 3874 r
 r Functional Segmentation of Brain using Group-PICA r

of the power spectra shows the most contributing frequen-

cies, mostly presenting the same proportional distribution
as the mean power spectra. In the majority of the compo-
nents, the dominant feature is the 1/f characteristic trend.
However, in ICs 7, 18, 19, 21, 25, 27, 30, 32, 38, 43, 59, and
63, there is a distinct peak in lower frequencies. That is,
the lowest frequency is not with the largest power. The
peaks were almost absent from frontal sources, with the
exception of DFM related IC 18. Primary auditory IC 25,

Figure 7.
Sagital and axial examples of the default mode network shown
with model order 10, 70, and 150 after 100 ICASSO runs
shown in a PICA setting in fsl4, with P ¼ 0.5 threshold. The
core default mode areas become separated into at least three Figure 8.
independent subnetworks, one in the precuneus and posterior Top: the individual correlation matrix of each signal source com-
parietal areas, and two in the ventromedial prefrontal areas. pared to each other enables assessment of the intersource con-
nectivity of basically the entire brain cortex. Mean correlation
coefficients with 0 time lag are color encoded at the top/right,
Frequency Domain and standard deviations of individual source correlation pairs in
gray scale at the bottom. Below: mean inter-component correla-
The mean power spectra of the individual ICA time tion coefficients between individual sources are analyzed at the
courses extracted from the group data mixing matrix had subject level. The RSN components show increased intersource
increased low frequency power with 1/f distribution char- connectivity between all the calculated sources compared to
acteristics (see also Figs. 2–4). The first rank approximate artifact related non-RSN sources.

r 3875 r
 r Kiviniemi et al. r

somatosensory IC 43, motor ICs 19 and 27, and visual be detectable with increased spatiotemporal accuracy in
sources 21, 32, and 59 showed peaks. In the visual cortex, the future [Kenet et al., 2003; Pelled and Goelman, 2004].
IC 59 in the medial areas had an exceptional 0.045 Hz fre- The previously detected 12 signal sources that were
quency peak at the most dominant approximate. Default found in smaller datasets are basically summations of
mode IC 38 and the Broca area source IC 30 also presented some of the sources presented here [Beckmann et al., 2005;
peaks in their power spectra. Damoiseaux et al., 2006; Sorg et al., 2007; Jafri et al., 2008].
An example of this can be seen in the Figure 7, where the
DMN is initially presented as one single source with the
DISCUSSION model order 10, and, after increasing the model order,
becomes divided into several sources. It seems that low
The results of this study represent functional segmenta- model orders provide information on large scale networks,
tion of the human brain cortex, using the statistical inde- whereas higher model orders, at least in group-data set-
pendence of spontaneous brain activity sources as a tings, provide subnetwork accuracy. We are currently ana-
delineating feature. We have shown that the intrinsic base- lyzing the effect of model order on the dividing of RSN’s
line activity of the brain cortex can be separated into 42 into subnetworks. We suggest that the 42 group-PICA sig-
independent sources, having RSN features, using a high nal sources depicted in this study represent a more finely-
model order PICA approach. These sources cover virtually tuned and segmented version of the functional neuro-anat-
the whole brain cortex, excluding the sources affected by omy of the RSN’s in the brain cortex, and therefore we
susceptibility artifacts near air-sinus interfaces. The domi- strongly support using high model order in large group
nant low frequency power characteristics of rank-1 ap- datasets.
proximate in the RSN support the idea that the sources The distinction of components into two categories of
represent low frequency background activity of the RSN. either RSN or non-RSN may be somewhat arbitrary
All of the RSN sources have increased LFF power in the although there have been some successful attempts
mean power spectra (c.f. Figs. 2–4). towards automated noise removal that basically also delin-
The utilization of continuous in vivo baseline activity eates ICs into neuronal and artifactual originated ones [De
information on brain activity offers a complementary Martino et al., 2007; Tohka et al., 2008]. One might justifi-
method to previous anatomical/histological dissection ably speculate in the absence of definite exclusion criteria
studies, as well as imaging data-based segmentation stud- that there exist a ‘‘borderline’’ between RSN sources and
ies, which currently form the basis of spatial localization non-RSN sources. Some ICs have features of cortical signal
schemes. The introduction of functional segmentation source overlapping with regions often presenting artifacts.
offers a more accurate tool for delineating functional activ- For example ICs 3, 24, and 40 all are situated within the
ity that traverses presently used tissue segmentation bor- cranial brain cortex and somewhat resemble sources seen
ders in connected neuronal networks. In addition, these IC in individual ICA results as motion artefact/partial vol-
source maps can be used as data driven, statistically inde- ume effects. However, at least ICs 40 seems to be a RSN
pendent, functional seed regions for further correlation based on further studies not shown here. On the other
analyses. hand, ICs 8, 52, and 61 are situated very close to sagital or
Using present 1.5-T MRI scanning methodology our transverse sinuses and one might speculate that the source
group has identified 42 signal sources over the whole has some contribution from pulsation from sagital sinus.
brain cortex. The results of this study are in excellent However, the sagital sinus at least was shown to have an
agreement with the recent results of Cohen et al. [2008], even more confined source of its own (IC 46; c.f. Fig. 6). It
who showed several small functionally connected areas would be beneficial to obtain a more quantitative criterion
within the same cortical regions. The work of Malinen than anatomical positioning and low frequency fluctuation
et al. [2007] convincingly shows that during natural stim- of the sources to differentiate true RSN sources from
uli there are all together at least 20 task- and nontask- artefacts.
related signal sources detectable in the brain. Compared to The signal sources depicted here accurately follow the
a higher resolution fMRI scan on a rat, the results of this known functional anatomy of the sensory and motor net-
study seem conservative, because rats were shown to have works. Sources IC 19 and 27 together cover 89% of the
some 20 connectivity hubs detectable in the somatosensory right and 86% of the left primary motor cortices when
and visual systems alone [Pawela et al., 2008]. Two groups compared to MNI atlases in fsl4. Primary sensorimotor
have been able to identify separate regions from the stria- areas were shown to have spatial connectivity with 14 ICs,
tum and thalamus with distinct functional connectivity to 12 of which were the same (ICs 3, 4, 6, 14, 19, 24, 27, 40,
the brain cortex in human subjects [DiMartino et al., 2008; 43, 53, 60, 66). The M1 anatomical area was uniquely con-
Mezer et al., 2009]. On a submillimeter scale, anatomical nected to ICs 41 and 45, while S1 was connected to IC 9
structures of the visual field columns and cortical layers and 35 (c.f. Supporting Information Table I.2). Somatosen-
can be identified based on the spontaneous electrophysio- sory S1 had a spatial overlap of 61% (IC 6) and S2 72% (IC
logical activity of subsecond time scales. This suggests that 43). The supplementary motor area (i.e. juxtapositional
an even greater number of spontaneous signal sources will lobule) activity was differentiated from the primary

r 3876 r
 r Functional Segmentation of Brain using Group-PICA r

sensorimotor components, although it was shown to have the areas anticorrelated to DMN in correlation analyses
connectivity to them. SMA was shown to be connected to are also present as independent signal sources and there-
12 RSNs, the largest share being detected in IC 49 (58%). fore may not be completely a by-product of global mean
Also IC 53 (24%), IC 60 (34%), IC 18 (30%), and IC 19 signal regression of BOLD signal.
(36%) had relatively high spatial overlaps with the SMA To maximize the chances of effective treatment, it is cru-
template. The unilateral and somewhat anterior pre-motor cially important to identify diseases at the earliest stage
sources (ICs 53 and 60: Broadman area BA6, 42% and 55%, possible. Functional alterations often present themselves as
respectively) could be related functional differences of the symptoms prior to irreversible tissue damage. For clinical
subjects premotor areas [Longcamp et al., 2005]. brain research, function-based tissue segmentation is of
Visual sources are separated into multiple sources, and pivotal importance in the early detection and characteriza-
not only into medial and lateral components as before. tion of diseases. Greicius et al. [2004, 2007] have already
The primary and secondary visual cortices V1 and V2 shown that minimally impaired memory deficiency or
were detected in 10 ICs (ICs 1, 2, 4, 21, 32, 38, 52, 59, 61, depression are related to the function of a brain network.
and 63), and the largest overlap for V1 was in ICs 21 and ICA analyses detecting multiple brain networks have
55, both having 55.5%. The secondary V2 had the largest shown that, while some networks show no changes, others
overlaps in ICs 21 (51%) and IC 59 (54%). The multiple may be more affected by diseases such as Alzheimer’s and
components have a strong resemblance to natural viewing schizophrenia [Calhoun et al., 2008; Jafri et al., 2008; Sorg
(V1–V5, lateral occipital complexes) and listening-related et al., 2007]. The high model order ICA segmentation
components in the occipital and temporal regions [Bartels increases functional and anatomical specificity in the con-
and Zeki, 2005; Malinen et al., 2007]. The positive/negative nectivity analysis of the networks and so may enhance our
ICs 52 and 61 resemble visual fields seen in visual hemi- ability to detect disease-related network alterations.
field activation study using the frequency based group The ICA methodology in this study is widely used by
ICA [Calhoun et al., 2003]. other researchers, and we suggest that the relatively high
IC 25 was the one most overlapping with the primary number of detected RSN components is due to increased
auditory cortex (90.5% with the right TE 1.0 and 89.7% signal-to-noise ratio related to the large dataset used in the
with TE 1.2 on the left) and is left-dominant. The primary group analysis. How inter-subject variance affects the
auditory cortex was connected to 10 ICs (ICs 1, 15, 18, 25, results is, however, an important question; are we seeing
29, 37, 39, 43, 47, 68). Broca’s area had connectivity with more fringe components that are only present on subpopu-
16 sources (ICs 3, 4, 7, 8, 9, 10, 15, 30, 31, 35, 37, 39, 43, 49, lations within the group [Schmidhorst and Holland, 2004]?
66, 68) with largest nodes in IC 30 (77% right side), IC 35 For example, bored, anxious, or enthusiastic subjects might
(97% left side). Several other frontal and temporal sources have different strengths of fluctuations in default mode
have interesting and complex connectivity with various subnetworks. The answer to this problem is also related to
functional regions related to attention, language, and other model order of a large group dataset and therefore this
cognitive functions. Their spatial overlapping with pres- issue may not be entirely handled within the scope of this
ently known anatomical regions are rich, just like those of manuscript.
the primary sensory and motor cortices, and they are The only minor alteration compared to general trends in
presented more precisely in the Supporting Information ICA analyses was the utilization of 70 components. The
Table I.2. high model order is supported by the works of McKeown
Recent findings show that the DMN nodes have sepa- et al. [2002] and some recent ones as well [De Martino
rate causality and connectivity characteristics [Sridharan et al., 2007; Malinen et al., 2007]. One can always speculate
et al., 2008; Uddin et al., 2009]. Also researchers using ICA on the effects of over-fitting in the detected components, if
have noticed the separation of the DMN into at least two there are any. It should be noted that model order estima-
subnetworks [Calhoun et al., 2008; Damoiseaux et al., tion with PPCA also yielded high model order numbers,
2006]. The results of this study show that the DMN is i.e. 73 in this dataset, which strongly speaks against major
functionally differentiated into anterior and posterior overfitting. Two of the original 73 components were
parts. There are also ‘‘DMN-type’’ components (ICs 7, 10, related to nonaccurate extra brain tissue removal, and this
23, 18, and 37) with clear spatial overlap with DMN was corrected by utilizing another bet-algorithm outside
related areas [Buckner et al., 2008]. The detection of these the MELODICA automated tool.
areas is particularly sensitive to model order of the ICA To validate the stability of the method, we employed
and we are currently investigating this phenomenon. ICASSO on the pre-processed MELODIC data matrix. In
The regions detected as anticorrelated with the DMN in doing so, 60 of the 70 components were shown to have a
correlation analyses using global mean regression were stability value of at least 0.7 in the group analysis, espe-
also detected as sources of their own (ICs 6, 40, 49). The cially when the ICASSO was computed using a low epsi-
use of ICA avoids the problematic regression of the global lon threshold. Higher epsilon thresholds led to less
mean signal, although the PCA mean subtraction step in clustered components in repeated runs due to less
preprocessing of PICA also alters the data to some extent demanding convergence of the ICA algorithm around local
[Murphy et al., 2009]. On the basis of the present results, maxima. It should be interesting to see how the repeated

r 3877 r
 TABLE 1. Overlap with FSL templates, maximum t-score MNI-coordinates & clustering index of group-PICA RSN-sources

Overlap % t-score MNI max-t Template

No. of Spatial
No. of IC Anatomical ROI area Area IC vs. ROI Mean Max X Y Z Source ICASSO correlation Iq

1 GM Visual cortex V5 R/L 60.0/55.0 6.7/5.6 14.0/11.0 27/67 31/26 28/31 J 15 0.86 0.964
Temporal Occipital Fusiform 66.5 7.3 20.1 30 29 26 HO
Cortex
Occipital Fusiform Gyrus 78.0 8.1 20.6 30 28 26 HO
Lateral Occipital Cortex, inferior 58.0 6.7 20.1 30 27 26 HO
division
2 GM Visual cortex V1 BA17 V1/V2 37.0/37.5 8.5/8.6 31.3/31.3 60/60 16/16 34/34 J 63 0.84 0.810
GM Visual cortex V2 BA18 37.5 8.6 31.3 60 16 34 J
Occipital Fusiform Gyrus 55.0 4.9 31.3 60 16 34 HO
3 GM Premotor cortex BA6 R/L 30.0/31.0 5.8/6.0 13.8/13.5 32/54 74/74 66/66 J
Superior Frontal Gyrus 43.6 6.9 14.6 30 74 64 HO
Middle Frontal Gyrus 42.0 6.8 14.6 30 74 64 HO
4 GM Visual cortex V5 R/L 73.0/62.0 5.9/5.4 11.2/9.9 24/64 22/21 42/41 J 61 0.56 0.627
Lateral Occipital Cortex, superior 44.9 6.8 17.0 52 18 52 HO
division
r

Cuneal Cortex 55.8 6.8 17.0 52 18 52 HO

6 GM Anterior intra-parietal sul- R/L 78.0/56.0 7.5/5.2 18.2/12.0 13/76 50/48 52/50 J 43 0.83 0.821

r
cus hIP2
GM Secondary somatosensory 61.5 9.0 19.0 12 52 54 J

3878
cortex/Parietal operculum

r
OP1
Kiviniemi et al.

Supramarginal Gyrus, anterior 68.8 7.3 19.0 12 52 54 HO

r

division
7 GM Anterior intra-parietal sul- 32.9 6.7 17.0 70 32 50 J 46 0.84 0.730
cus hIP1
GM Visual cortex V5 46.3 5.0 14.4 71 29 48 J
Middle Temporal Gyrus, tem- 38.8 5.2 16.4 73 32 48 HO
porooccipital part
Angular Gyrus 36.2 6.5 18.0 72 32 50
8 Frontal Pole 27.2 6.8 21.3 44 94 46 HO 9 0.90 0.950
Superior Frontal Gyrus 27.8 7.0 21.0 44 94 47 HO
Paracingulate Gyrus 28.6 6.3 18.8 43 92 48 HO
9 GM Anterior intra-parietal sul- 68.9 6.4 15.5 30 28 60 J 26 0.86 0.914
cus hIP1
GM Anterior intra-parietal sul- 56.7 5.0 10.2 24 36 62 J
cus hIP2
Lateral Occipital Cortex, superior 32.6 6.2 17.5 32 26 62 HO
division
10 GM Broca’s area BA45 20.6 1.9 4.1 24 82 36 J 4 0.72 0.982
WM Cingulum 20.0 3.0 9.0 50 85 36 J
Frontal Medial Cortex 65.3 8.7 22.2 46 92 32 HO
Paracingulate Gyrus 50.5 6.5 22.2 46 92 32 HO
 TABLE 1. (continued)
Overlap % t-score MNI max-t Template
No. of Spatial
No. of IC Anatomical ROI area Area IC vs. ROI Mean Max X Y Z Source ICASSO correlation Iq

13 GM Amygdala_laterobasal group R/L 37.0/37.0 5.6/5.1 11.0/9.5 27/67 63/68 17/20 J 8 0.94 0.947
GM Hippocampus entorhinal R/L 40.0/38.0 5.5/4.8 11.5/10.0 27/62 63/60 16/16 J
cortex
Inferior Temporal Gyrus, ante- 76.7 7.8 14.5 68 68 16 HO
rior division
r

Temporal Fusiform Cortex, ante- 73.8 6.8 13.6 22 64 16 HO

rior division
14 GM Primary somatosensory cor- R/L 19.8/33.6 5.0/5.5 10.8/12.8 42/51 41/44 62/60 J 11 0.78 0.939
tex BA3a
Cingulate Gyrus, posterior 40.7 6.6 12.9 50 44 60 HO
division
Precuneous Cortex 40.2 5.7 12.9 50 44 60 HO
Cuneal Cortex 21.4 3.9 6.4 38 32 58 HO
15 GM Broca’s area BA44 R/L 44.9/30.7 4.2/4.5 13.0/14.6 18/67 74/75 34/32 J 67 0.84 0.579
GM Broca’s area BA45 R/L 53.1/45.5 5.0/5.3 13.5/15.1 19/68 74/74 32/32 J

r
Inferior Frontal Gyrus, pars tria./operc. 51.0/42.6 5.5/5.3 15.1/14.6 68/69 74/74 32/32 HO
Paracingulate Gyrus 43.2 3.9 8.2 46 74 60 HO

3879
Frontal Operculum Cortex 66.4 7.6 15.1 68 74 32 HO

r
18 GM Primary auditory cortex R/L 44.0/64.0 3.3/4.1 4.9/6.3 24/66 48/50 44/40 J 60 0.72 0.746
TE1.1
Paracingulate Gyrus 62.7 6.4 19.0 44 76 48 HO
Cingulate Gyrus, anterior 68.5 7.0 19.0 44 76 48 HO
division
19 GM Primary motor cortex R/L 50.2/50.8 7.4/8.0 15.3/15.4 45/46 52/52 68/68 J 35 0.90 0.919
BA4a(þp)
GM Primary somatosensory cor- R 41.0/39.8 7.8/7.7 14.6/15.0 44/56 50/48 66/66 J
tex BA3a(þ1-3b)
Functional Segmentation of Brain using Group-PICA

Juxtapositional Lobule Cortex 35.6 7.2 15.4 46 52 68 HO

r

21 GM Visual cortex V1 BA17 V1/V2 55.5/51.2 8.7/7.7 26.8/26.8 42/42 18/18 36/36 J 66 0.91 0.645
Intracalcarine Cortex 51.5 10.3 26.8 42 18 36 HO
Occipital Pole 52.3 8.3 26.8 42 18 36 HO
23 GM Anterior intra-parietal sul- R/L 47.5/38.1 8.0/5.6 19.0/12.5 20/69 34/34 59/60 J 30 0.81 0.859
cus hIP1-2
Supramarginal Gyrus, posterior 35.2 6.8 19.8 18 36 58 HO
division
Angular Gyrus 50.4 7.8 19.8 18 36 58 HO
Lateral Occipital Cortex, superoir 29.3 5.7 19.8 18 36 58 HO
division
 TABLE 1. (continued)
Overlap % t-score MNI max-t Template
No. of Spatial
No. of IC Anatomical ROI area Area IC vs. ROI Mean Max X Y Z Source ICASSO correlation Iq

24 GM Primary motor cortex R/L 38.6/35.8 7.0/5.8 12.7/13.5 40/50 56/57 74/74 J 58 0.67 0.907
BA4aþp
GM Primary somatosensory cor- R/L 34.4/31.1 6.4/4.7 12.1/11.6 30/50 50/51 72/74 J
tex BA1
GM Premotor cortex BA6 R R/L 27.0/27.5 6.6/5.8 12.7/13.6 40/52 56/58 74/74 J
Juxtapositional Lobule Cortex 26.0 4.7 13.5 50 58 74 HO
(SMA)
25 GM Primary auditory cortex R/L 90.5/87.6 8.1/8.9 17.7/19.3 14/76 52/50 40/40 J 14 0.94 0.925
TE1.0 (-1.2)
GM Secondary somatosensory R/L 39.7/52.9 6.7/9.2 17.0/19.3 14/76 53/50 40/40 J
cortex/Parietal operculum
OP1
Heschl’s Gyrus (includes H1 and 90.6 8.7 19.4 76 50 40 HO
H2)
r

27 GM Anterior intra-parietal sul- R/L 49.3/48.9 7.0/7.9 17.9/18.7 22/70 50/48 63/61 J 69 0.88 0.578
cus hIP2

r
GM Primary motor cortex BA4p 35.3 7.9 18.3 22 51 64 J
(þa)
GM Primary somatosensory cor- R/L 55.6/59.8 7.6/9.9 18.3/19.8 22/66 52/52 64/66 J

3880
tex BA2

r
Supramarginal Gyrus, anterior 56.3 6.4 19.3 68 48 63 HO
Kiviniemi et al.

division
r

29 GM Amygdala_centromedial R/L 68.6/40.1 5.0/4.5 11.6/9.6 24/62 61/54 26/31 J 23 0.90 0.940
group
GM Hippocampus cornu R/L 41.0/32.6 5.2/5.3 12.9/13.7 24/65 57/54 29/30 J
ammonis
GM Lateral geniculate body R/L 54.4/42.7 3.5/3.4 5.8/5.7 30/60 55/53 32/30 J
Planum Polare 64.5 7.7 15.6 66 56 30 HO
30 GM Anterior intra-parietal sul- R/L 37.9/21.6 3.8/3.7 6.2/5.8 26/24 34/36 60/60 J 33 0.84 0.923
cus hIP1
GM Broca’s area BA44 (þBA45) R/L 77.5/33.9 6.7/4.4 13.1/8.3 20/68 70/72 48/48 J
GM Secondary somatosensory R 55.3 4.5 10.5 22 65 46 J
cortex/Parietal operculum
OP3 (þOP2,4)
Inferior Frontal Gyrus, pars 63.9 6.3 13.1 20 70 48 HO
opercularis
Right Caudate 20.0 3.1 4.2 34 61 46 HO
31 GM Broca’s area BA45 R/L 16.7/29.0 0.4/4.0 7.4/11.1 22/66 86/86 45/43 54 0.69 0.665
Frontal Pole 42.5 5.1 18.6 60 94 40
Inferior Frontal Gyrus, pars 22.3 1.5 8.8 66 84 45
triangularis
 TABLE 1. (continued)
Overlap % t-score MNI max-t Template
No. of Spatial
No. of IC Anatomical ROI area Area IC vs. ROI Mean Max X Y Z Source ICASSO correlation Iq

32 GM Hippocampus subiculum R/L 42.9/48.2 4.8/5.4 9.4/11.02 32/58 42/43 30/30 J 28 0.66 0.936
GM Visual cortex V2 BA18 26.0 3.4 9.9 52 34 38 J
GM Medial (þlateral) geniculate R/L 55.8/88.5 3.4/3.9 4.9/6.2 38/50 47/46 29/30 J
body
Parahippocampal Gyrus, poste- 83.3 5.7 12.0 58 42 28 HO
rior division
r

34 Cerebellum 13 0.91 0.935

35 GM Broca’s area BA45(þBA45) R/L 19.4/97.0 3.5/8.8 5.6/17.7 20/70 80/84 40/36 J 42 0.84 0.811
Inferior Frontal Gyrus, pars 59.6 8.0 17.7 70 84 36 J
triangularis
Inferior Frontal Gyrus, pars 51.5 6.7 16.5 70 78 41 J
opercularis
Frontal Operculum Cortex 39.7 5.8 15.6 70 77 41 J
37 Frontal Pole 32.2 4.5 15.7 28 84 54 HO 34 0.69 0.897
Superior Frontal Gyrus 27.5 3.9 15.7 28 84 54 HO
Middle Frontal Gyrus 36.9 5.5 15.7 28 84 54 HO

r
Paracingulate Gyrus 22.9 1.0 5.1 41 90 53 HO
38 Cingulate Gyrus, posterior 50.5 8.2 19.7 46 33 50 HO 65 0.83 0.883

3881
division

r
Precuneous Cortex 43.8 8.0 19.9 46 32 50 HO
Cuneal Cortex 40.2 8.4 19.7 46 31 50 HO
Supracalcarine Cortex 53.6 9.2 19.9 46 32 50 HO
39 GM Visual cortex V5 R/L 91.5/56.0 7.1/4.0 12.7/6.2 18/70 34/32 38/36 J 16 0.86 0.923
Middle Temporal Gyrus, tem- 69.4 6.4 14.6 20 38 38 J
porooccipital part
Inferior Temporal Gyrus, tem- 37.7 5.9 13.3 17 37 36 J
porooccipital part
Lateral Occipital Cortex, inferior 34.2 5.3 13.6 20 36 38 J
division
Functional Segmentation of Brain using Group-PICA
r

40 GM Primary somatosensory cor- R/L 36.8/39.0 5.6/6.4 15.0/12.7 32/58 36/36 70/70 J 20 0.82 0.908
tex BA2
Postcentral Gyrus 22.1 5.2 14.9 33 36 70 HO
Superior Parietal Lobule 52.1 6.1 15.0 32 36 70 HO
Lateral Occipital Cortex, superoir 20.3 5.6 15.0 32 36 70 HO
division
Precuneous Cortex 17.8 3.4 13.7 37 34 70 HO
41 GM Amygdala_centromedial R/L 80.1/76.3 6.9/6.2 18.4/14.6 31/58 64/63 32/32 J 12 0.93 0.942
group
Putamen R/L 92.7/93.4 9.4/7.9 21.0/18.4 58/32 66/66 34/32 HO
Pallidum 80.0/80.4 5.5/7.0 11.4/16.9 56/54 66/70 34/33 HO
Accumbens R/L 51.4/55.7 5.4/5.5 12.3/11.4 36/54 70/70 32/33 HO
 TABLE 1. (continued)
Overlap % t-score MNI max-t Template
No. of Spatial
No. of IC Anatomical ROI area Area IC vs. ROI Mean Max X Y Z Source ICASSO correlation Iq

43 GM Primary auditory cortex R/L 59.9/43.5 9.9/8.7 26.2/23.5 14/74 60/58 43/45 J 1 0.95 0.984
TE1.2
GM Primary somatosensory cor- R/L 37.2/37.2 14.0/13.4 34.1/33.7 16/72 60/58 48/50 J
tex BA3b
GM Secondary somatosensory R/L 72.9/71.9 12.2/12.3 34.1/33.7 16/72 60/58 48/50 J
cortex/Parietal operculum
OP4 (OP1-3)
Central Opercular Cortex 39.3 10.0 30.7 16 60 46 HO
45 GM Medial geniculate body R/L 24.7/16.4 3.0/3.0 3.5/3.5 38/49 52/48 33/30 J 17 0.90 0.901
Brain-Stem 22.3 4.2 11.1 42 40 24 HO
Right Thalamus 8.5 3.3 4.3 40 52 34 HO
47 GM Primary auditory cortex R/L 31.2/43.7 4.5/4.5 7.9/7.8 22 60 36 J 47 0.82 0.795
TE1.2
Subcallosal Cortex 29.0 6.6 16.4 44 78 38 HO
Paracingulate Gyrus 31.7 5.7 16.4 44 81 38 HO
r

Cingulate Gyrus, anterior 33.8 6.8 17.1 44 80 38 HO

division

r
Left Accumbens 31.4 4.4 8.6 48 74 36 HO
49 GM Broca’s area BA44 R/L 53.5/41.1 5.3/1.8 12.9/10.2 16/74 66/64 36/36 J 57 0.65 0.586

3882
GM Primary auditory cortex R/L 99.8/91.0 7.6/5.6 13.0/10.2 16/74 66/64 36/36 J

r
TE1.2
Insular Cortex 59.7 5.4 12.7 24 66 36 J
Kiviniemi et al.
r

Juxtapositional Lobule Cortex 57.9 4.9 11.4 44 66 58 HO

(SMA)
Frontal Operculum Cortex 62.0 6.0 13.0 16 66 36 HO
Central Opercular Cortex 65.2 6.1 13.0 16 66 36 HO
Planum Polare 69.8 6.3 13.0 16 66 36 HO
52 GM Visual cortex V1 BA17 L, V1/V2 19.7/19.4 1.7/1.6 16.6/15.8 54/52 18/18 24/25 J 32 0.73 0.888
Lateral Occipital Cortex, inferior L 24.5 3.6 17.1 56 20 24 J
division
Occipital Fusiform Gyrus L 34.7 5.9 17.1 56 20 24 HO
Occipital Pole L 28.1 1.7 17.0 52 18 24 HO
53 GM Primary motor cortex BA4a L 18.9 4.5 10.5 60 58 62 J 41 0.30 0.942
Superior Frontal Gyrus 19.7 5.5 12.2 58 60 64 HO
Middle Frontal Gyrus 14.2 5.5 12.2 58 60 64 HO
Juxtapositional Lobule Cortex 25.8 4.1 7.9 53 64 64 HO
(SMA)
59 GM Visual cortex V1 BA17 55.5 9.7 22.3 50 28 38 J 10 0.88 0.951
GM Visual cortex V2 BA18 R 54.4 9.3 22.3 50 28 38 J
Intracalcarine Cortex R 86.9 12.0 22.3 50 28 38 HO
Lingual Gyrus R 63.1 10.4 22.3 50 28 38 HO
Supracalcarine Cortex R 68.6 11.6 22.3 50 28 38 HO
 TABLE 1. (continued)
Overlap % t-score MNI max-t Template
No. of Spatial
No. of IC Anatomical ROI area Area IC vs. ROI Mean Max X Y Z Source ICASSO correlation Iq

60 GM Anterior intra-parietal sul- R 49.4 4.4 8.4 25 51 62 J 52 0.79 0.678

cus hIP2
r

GM Primary motor cortex BA4p R 47.9 5.5 11.0 33 56 64 J

GM Primary somatosensory cor- R 50.0 4.6 9.1 24 53 64 J
tex BA2
GM Premotor cortex BA6 R 55.1 6.1 13.8 32 60 64 J
61 GM Visual cortex V1 BA17 R 23.8 2.8 16.5 36 18 26 J 55 0.64 0.627
GM Visual cortex V2 BA18 22.4 2.7 16.6 32 20 26 J
Occipital Fusiform Gyrus 33.7 5.8 16.7 32 20 24 HO
Occipital Pole 32.0 3.9 16.6 32 20 25 HO
63 Lateral Occipital Cortex, superior 21.3 6.0 25.0 43 24 58 J 61 0.61 0.628
division

r
Precuneous Cortex 30.9 8.1 25.1 44 24 56 J
Cuneal Cortex 54.0 8.9 25.1 44 24 56 J

3883
Supracalcarine Cortex 29.7 6.7 23.2 42 25 56 J

r
66 Frontal Medial Cortex 33.9 6.6 15.2 42 94 28 4 0.70 0.971
Juxtapositional Lobule Cortex 25.7 3.4 5.2 48 64 58
(SMA)
Middle Temporal Gyrus, tem- 19.6 3.6 6.0 72 36 40
porooccipital part
68 GM Broca’s area BA44 20.7 4.6 24.5 29 68 56 J 68 0.70 0.602
GM Primary auditory cortex 24.4 3.1 74.7 78 53 42 J
TE1.2
GM Visual cortex V5 33.2 3.6 19.2 20 28 47 J
Superior Frontal Gyrus 21.8 5.6 33.6 32 70 56 HO
Functional Segmentation of Brain using Group-PICA
r

max 0.95 0.984

mean 0.79 0.829
std 0.13 0.132

No. of IC ¼ number of selected component, Anatomical ROI Area ¼ Name of anatomical template from fsl4, Area ¼ side or other anatomical specification R ¼ right, L¼
left, Overlap % ¼ anatomical overlap percentage of components vs. anatomical structure, t-score ¼ t-value (mean, maximum [Max]), MNI max-t ¼ maximum t-value X, Y
& Z-direction Montreal Neurologic Institute 152 coordinates, Template source ¼ anatomical ROI source in fsl4: J ¼ Juelich, HO ¼ Harvard-Oxford template, # ICASSO ¼
icasso run equivalent to selected IC, spatial correlation ¼ spatial correlation between ICASSO and selected IC, Iq ¼ ICASSO clustering index.
 r Kiviniemi et al. r

ICASSO analysis results differentiate from single subject dynamic response functions and correlation-based connec-
run PICA results. We are currently exploring the benefits tivity measurement inaccurate. After establishing the
of using both PICA and ICASSO in a combined analysis cortical oscillatory activity within each of the networks
setting. Another unclear issue is how the RSN’s detected with high model order ICA, one may be able to utilize
in the group data can be used in individual settings. On these baseline oscillations in a regression analysis in a
the basis of the results shown here, a familiar ICA-related more accurate manner, and obtain increased signal-to-
issue reemerges; what is the optimal model order of ICA noise ratios in stimulus related effective connectivity anal-
in datasets with increased subject numbers, and how are yses of brain networks. Also, due to the fact that ICA sep-
preprocessing schemes influencing the results? We are cur- arates noise sources, the mixing matrix signals seem to
rently investigating whether it is still possible to detect yield effective measures for connectivity and causality
more accurate RSN components with bigger data samples measurements, without biasing physiological or motion
and higher model orders. related correlations [Calhoun et al., 2008; Eichele et al.,
The temporal correlation matrix of the ICA enables the 2008; Sorg et al., 2007].
assessment of connectivity between all the detected signal Mantini et al. [2007] were able to distinguish EEG power
sources. It can be seen that some of the sources have lim- distribution fingerprints of different RSNs. Although the
ited correlation to practically all the other sources, which frequency resolution of fMRI alone is not as good as that
becomes visible as a dark row in the mean connectivity of EEG, one is able to depict differences between the
matrix (c.f. Fig. 5). These dark lines tend to be non-RSNs power distributions and first rank approximates of the
sources related to noise/artifacts and, indeed, their mean RSNs (c.f. Figs. 2–4). Primary sensory and motor sources,
temporal intersource connectivity is significantly lower than default mode, Broca’s area, and some occipital visual sour-
those of the RSN sources. The temporal correlation struc- ces present some prominent frequency peaks in addition
ture could perhaps be used as a feature in differentiating to the 1/f trend in the spectrum. The frontal sources pre-
the non-RSN and RSN sources with different lag periods. dominantly have 1/f distributions without clear peaks.
The non-RSN sources are thus temporally more independ- The functional significance of this difference in peak fre-
ent than the connected RSN sources, and therefore their quency distributions needs further elucidation.
automated separation seems feasible from this point of
view [De Martino et al., 2007; Tohka et al., 2008].
CONCLUSION
There have been concerns about the uncontrolled nature
of the resting state regarding attention and free thinking The brain cortex can be robustly segmented into 42
induced alterations. In this study, we used a semi-resting independent RSNs by utilizing the group-ICA of baseline
state by instructing the subject to fixate on a cross shown BOLD data. This method offers complementary infor-
on the screen. The subject is not completely at rest and mation on tissue characteristics of cortical structures
tries to sustain attention on the cross. On the other hand, compared to previous post-mortem histological and micro-
no restrictions on the fixation or mental imagery were anatomic dissection of structures. Utilization of a relatively
given either. The fixation reduces motion artifacts and still high model order and multi-subject group data enables
offers some cognitive baseline function to reduce some of increased accuracy of functional segmentation of brain tis-
the variability of free thinking. McAvoy et al. [2008] sue due to increased signal-to-noise ratio. The method
recently presented rather similar spatial distribution of reveals multiple interlinked connectivity subnetworks ena-
areas where BOLD fluctuation frequency is dependent on bling more fine-tuned characterization of seed regions for
whether the subject has eyes open, closed, or is fixating on further causality and connectivity analyses.
a cross. We found a similar distribution of areas, and in
addition differences in power spectral densities between
the visual area sources. The V1-related medial (IC 59) and
ACKNOWLEDGMENTS
dorsal (IC 21) components in this study have somewhat The authors cordially thank Dr. Christian Beckmann for
different rank-1 approximates from the other occipital his assistance in PICA mixing matrix and bet analyses,
sources. It would be interesting to see how these sources and, Dr Gordon Roberts for editorial assistance. The
differ in subjects with eyes closed in a large number of authors would also like to express their deep gratitude to
subjects. the reviewers and handling editor for their excellent
Importantly, it has been shown that the effective estima- suggestions and corrections on the manuscript. This study
tion and removal of underlying oscillations significantly was supported by Academy of Finland Grant #111711, Fin-
improves the signal-to-noise ratio in the task activation nish Medical Foundation and Finnish Neurological Associ-
studies [Fox et al., 2006]. In addition, baseline activity ation grants.
oscillations preceding stimuli explains the major share of
the subsequent stimulus response variability [Boly et al.,
2007; Fox et al., 2006; Eichele et al., 2008]. There may be
REFERENCES
stimulus-locked connectivity alterations that can be Bartels A, Zeki S (2005): The chronoarchitecture of the cerebral
masked by the baseline oscillations, rendering the hemo- cortex. Philos Trans R Soc Lond B Biol Sci 360:733–750.

r 3884 r
 r Functional Segmentation of Brain using Group-PICA r

Beckmann CF, Smith SM (2004): Probabilistic independent compo- Fukunaga M, Horovitz SG, de Zwart JA, van Gelderen P, Balkin
nent analysis for functional magnetic resonance imaging. IEEE TJ, Braun AR, Duyn JH (2008): Metabolic origin of BOLD sig-
Trans Med Imaging 23:137–152. nal fluctuations in the absence of stimuli. J Cereb Blood Flow
Beckmann CF, DeLuca M, Devlin JT, Smith SM (2005): Investiga- Metab 28:1377–1387.
tions into resting-state connectivity using independent compo- Greicius MD, Srivastava G, Reiss AL, Menon V (2004): Default-
nent analysis. Philos Trans R Soc Lond B Biol Sci 360:1001– mode network activity distinguishes Alzheimer’s disease from
1013. healthy aging: Evidence from functional MRI. Proc Natl Acad
Biswal BB, Yetkin FZ, Haughton VM, Hyde JS (1995): Functional Sci USA 101:4637–4642.
connectivity in the motor cortex of resting human brain using Greicius MD, Flores BH, Menon V, Glover GH, Solvason HB,
echo-planar MRI. Magn Reson Med 34:537–541. Kenna H, Reiss AL, Schatzberg AF (2007): Resting-state func-
Birn RM, Diamond JB, Smith MA, Bandettini PA. (2006): Separat- tional connectivity in major depression: Abnormally increased
ing respiratory-variation-related fluctuations from neuronal- contributions from subgenual cingulate cortex and thalamus.
activity-related fluctuations in fMRI. Neuroimage 31:1536– Biol Psychiatry 62:429–437.
1548. Himberg J, Hyvärinen A, Esposito F (2004): Validating the inde-
Boly M, Balteau E, Schnakers C, Degueldre C, Moonen G, Luxen pendent components of neuroimaging time series via cluster-
A, Phillips C, Peigneux P, Maquet P, Laureys S (2007): Baseline ing and visualization. Neuroimage 22:1214–1222.
brain activity fluctuations predict somatosensory perception in Hyvärinen A (1999): Fast and robust fixed-point algorithms for
humans. Proc Natl Acad Sci USA 104:12187–12192. independent component analysis. IEEE Trans Neural Networks
Calhoun VD, Adali T, Pearlson GD, Pekar JJ (2001): A method for 10:626–634.
making group inferences from functional MRI data using inde- Jafri MJ, Pearlson GD, Stevens M, Calhoun VD (2008): A method
pendent component analysis. Hum Brain Mapp 14:140–151. for functional network connectivity among spatially independ-
Calhoun VD, Adali T, Pekar JJ, Pearlson GD (2003): Latency ent resting-state components in schizophrenia. Neuroimage
(in)sensitive ICA. Group independent component analysis of 39:1666–1681.
fMRI data in the temporal frequency domain. Neuroimage Jenkinson M, Bannister PR, Brady JM, Smith SM (2002): Improved
20:1661–1669 optimisation for the robust and accurate linear registration and
Calhoun VD, Kiehl KA, Pearlson GD (2008): Modulation of tem- motion correction of brain images. NeuroImage 17:825–841.
porally coherent brain networks estimated using ICA at rest Kannurpatti SS, Biswal BB, Kim YR, Rosen BR (2008): Spatio tempo-
and during cognitive tasks. Hum Brain Mapp 29:828–838. ral characteristics of low frequency BOLD signal fluctuations in
Cohen AL, Fair DA, Dosenbach NU, Miezin FM, Dierker D, Van isoflurane anesthetized rat brain. NeuroImage 40:1738–1747.
Essen DC, Schlaggar BL, Petersen SE (2008): Defining func- Kenet T, Bibitchkov D, Tsodyks M, Grinvald A, Arieli A (2003):
tional areas in individual human brains using resting func- Spontaneously emerging cortical representations of visual
tional connectivity MRI. NeuroImage 41:45–57. attributes. Nature 425:954–956.
Damoiseaux JS, Rombouts SA, Barkhof F, Scheltens P, Stam CJ, Kiviniemi V (2008): Endogenous brain fluctuations and diagnostic
Smith SM, Beckmann CF (2006): Consistent resting-state net- imaging. Hum Brain Mapp 29:810–817 (Review).
works across healthy subjects. Proc Natl Acad Sci USA Kiviniemi V, Kantola J-H, Jauhiainen J, Hyvärinen A, Tervonen O
103:13848–13853. (2003): Independent component analysis of non-deterministic
De Martino F, Gentile F, Esposito F, Balsi M, Di Salle F, Goebel R, fMRI signal sources. NeuroImage 19:253–260.
Formisano E (2007): Classification of fMRI independent com- Li YO, Adali T, Calhoun VD (2007): Estimating the number of in-
ponents using IC-fingerprints and support vector machine dependent components for functional magnetic resonance
classifiers. NeuroImage 34:177–194. imaging data. Hum Brain Mapp 11:1251–1266.
De Luca M, Beckmann CF, De Stefano N, Matthews PM, Smith Logothetis NK, Pauls J, Augath M, Trinath T, Oelterman A (2001):
SM (2006): fMRI resting state networks define distinct modes Neurophysiological investigation of the basis of the fMRI sig-
of long-distance interactions in the human brain. Neuroimage nal. Nature 412:150–157.
29:1359–1367. Long XY, Zuo XN, Kiviniemi V, Yang Y, Zou QH, Zhu CZ, Jiang
Di Martino A, Scheres A, Margulies DS, Kelly AM, Uddin LQ, TZ, Yang H, Gong QY, Wang L, Li KC, Xie S, Zang YF (2008):
Shehzad Z, Biswal B, Walters JR, Castellanos FX, Milham MP Default mode network as revealed with multiple methods for
(2008): Functional connectivity of human striatum: A resting resting-state functional MRI analysis. J Neurosci Methods
state fMRI study. Cereb Cortex 18:2735–2747. 171:349–355.
Eichele T, Debener S, Calhoun VD, Specht K, Engel AK, Hugdahl Longcamp M, Anton JL, Roth M, Velay JL (2005): Premotor activa-
K, von Cramon DY, Ullsperger M (2008): Prediction of human tions in response to visually presented single letters depend
errors by maladaptive changes in event-related brain networks. on the hand used to write: A study on left-handers. Neuropsy-
Proc Natl Acad Sci 105:6173–6178. chologia 43:1801–1809.
Eickhoff SB, Paus T, Caspers S, Grosbras MH, Evans AC, Zilles K, Lund TE, Madsen KH, Sidaros K, Luo WL, Nichols TE (2006):
Amunts K (2007): Assignment of functional activations to Non-white noise in fMRI: Does modelling have an impact?-
probabilistic cytoarchitectonic areas revisited. NeuroImage Neuroimage 29:54–66.
36:511–521. Ma L, Wang B, Chen X, Xiong J (2007): Detecting functional con-
Fox MD, Raichle ME (2007): Spontaneous fluctuations in brain ac- nectivity in the resting brain: A comparison between ICA and
tivity observed with functional magnetic resonance imaging. CCA. Magn Reson Imaging 25:47–56.
Nat Rev Neurosci 8:700–711. Malinen S, Hlushchuk Y, Hari R (2007): Towards natural stimula-
Fox MD, Snyder AZ, Zacks JM, Raichle ME (2006): Coherent tion in fMRI––Issues of data analysis. NeuroImage 35:131–139.
spontaneous activity accounts for trial-to-trial variability in Mantini D, Perrucci MG, Del Gratta C, Romani GL, Corbetta M
human evoked brain responses. Nat Neurosci 9:23–25. (2007): Electrophysiological signatures of resting state

r 3885 r
 r Kiviniemi et al. r

networks in the human brain. Proc Natl Acad Sci USA 104: component analysis of functional magnetic resonance imaging
13170–13175. data. J Magn Reson Imaging 19:365–368.
McAvoy M, Larson-Prior L, Nolan TS, Vaishnavi SN, Raichle ME, Shmuel A, Leopold DA (2008): Neuronal correlates of spontane-
d’Avossa G (2008)Resting states affect spontaneous BOLD ous fluctuations in fMRI signals in monkey visual cortex:
oscillations in sensory and paralimbic cortex. J Neurophysiol Implications for functional connectivity at rest. Hum Brain
100:922–931. Mapp 29:751–761.
McKeown MJ, Jung TP, Makeig S, Brown G, Kindermann SS, Lee Smith SM (2002): Fast robust automated brain extraction. Hum
TW, Sejnowski TJ (1998): Spatially independent activity pat- Brain Mapp 17:143–155.
terns in functional MRI data during the stroop color-naming Sorg C, Riedl V, Mühlau M, Calhoun VD, Eichele T, Läer L,
task. Proc Natl Acad Sci USA 95:803–810. Drzezga A, Förstl H, Kurz A, Zimmer C, Wohlschläger AM
McKeown MJ, Varadarajan V, Huettel S, McCarthy G (2002): (2007): Selective changes of resting-state networks in individu-
Deterministic and stochastic features of fMRI data: Implica- als at risk for Alzheimer’s disease. Proc Natl Acad Sci USA
tions for analysis of event-related experiments. J Neurosci 104:18760–18765.
Methods 118:103–113. Sridharan D, Levitin DJ, Menon V (2008): A critical role for the
Mezer A, Yovel Y, Pasternak O, Gorfine T, Assaf Y (2009): Cluster anal- right fronto-insular cortex in switching between central-execu-
ysis of resting-state fMRI time series. Neuroimage 45:1117–1125. tive and default mode networks. Proc Natl Acad Sci USA
Murphy K, Birn RM, Handwerker DA, Jones TB, Bandettini PA 105:12569–12574.
(2009): The impact of global signal regression on resting state Tohka J, Foerde K, Aron AR, Tom SM, Toga AW, Poldrack RA
correlations: Are anti-correlated networks introduced?Neuro- (2008): Automatic independent component labeling for artifact
image 44:893–905. removal in fMRI. Neuroimage 39:1227–1245.
Pawela CP, Biswal BB, Cho YR, Kao DS, Li R, Jones SR, Schulte Uddin LQ, Clare Kelly AM, Biswal BB, Xavier Castellanos F, Mil-
ML, Matloub HS, Hudetz AG, Hyde JS (2008): Resting-state ham MP (2009): Functional connectivity of default mode net-
functional connectivity of the rat brain. Magn Reson Med work components: Correlation, anticorrelation, and causality.
59:1021–1029. Hum Brain Mapp 30:625–637.
Pelled G, Goelman G (2004): Different physiological MRI noise van de Ven VG, Formisano E, Prvulovic D, Roeder CH, Linden
between cortical layers. Magn Res Med 52:913–916 DE (2004): Functional connectivity as revealed by spatial inde-
Salvador R, Suckling J, Coleman MR, Pickard JD, Menon D, Bull- pendent component analysis of fMRI measurements during
more E (2005): Neurophysiological architecture of functional rest. Hum Brain Mapp 22:165–178.
magnetic resonance images of human brain. Cereb Cortex Vincent JL, Patel GH, Fox MD, Snyder AZ, Baker JT, Van Essen
15:1332–1342. DC, Zempel JM, Snyder LH, Corbetta M, Raichle ME (2007):
Schmidhorst VJ, Holland SK (2004): Comparison of three methods Intrinsic functional architecture in the anaesthetized monkey
for generating group statistical inferences from independent brain. Nature 447:83–86.

r 3886 r