Abnormal dynamic functional connectivity of amygdalar...
Transcript of Abnormal dynamic functional connectivity of amygdalar...
1
Abnormal dynamic functional connectivity of amygdalar subregions in untreated patients
with first-episode major depressive disorder
Lihua Qiu;† Mingrui Xia;
† Bochao Cheng; Lin Yuan; Weihong KuangD; Feng Bi; Hua Ai; Zhongwei Gu;
Su Lui; Xiaoqi Huang; Yong He;* and Qiyong Gong
*
Validation analysis
Dynamic functional connectivity of amygdalar subregions in female MDD
Given that many MDD patients were female in the current study, we conducted the primary analyses
again with the females only. The demographic information for only female participants can be found in
Table 1 of the main manuscript. Here, we present our preliminary findings of altered dFC of amygdalar
subregions in 22 untreated, first-episode female patients with major depressive disorder compared with
33 age-matched, healthy female controls.
The image preprocessing, definition of amygdalar subregions, resting-state dFC analysis and
statistical analysis were the same as those in the main text. Statistical group difference maps were
constructed using a general linear model (GLM) with age as covariates. The statistical significance
threshold was set as P < 0.001 at the voxel level with a familywise error (FWE)-corrected P-value < 0.05
at the cluster level using SPM.
Similar to the results from all patients, the female MDD patients also showed a decreased positive
mean dFC between the left CM and brainstem and between the left SF and left thalamus. Further, a
decreased positive mean dFC between the right SF region and brainstem and thalamus was found in
female MDD patients but not in all patients. There was no significant decreased/increased negative mean
dFC in female MDD patients. Both female and all MDD patients exhibited a decreased positive mean
dFC between the CM/SF and brainstem and between the SF and thalamus, indicating that the brainstem
and thalamus play an important role in untreated patients with first-episode MDD (P < 0.05, FWE
corrected, Figure S1 and Figure S2). The detailed location and size are listed in Table S1. The present
findings of altered dFC in female MDD are consistent with the results from all patients, indicating that
the results from females play a major role in the main outcome. Including a large sample of male MDD
patients is helpful in understanding the differences in brain functional connectivity in male patients with
MDD.
2
Table S1: Regions showing significant between-group differences
Hemisphe
re
Subregio
n
dFC
metri
c
dFC
type
Region B
A
Cluste
r size
MNI
coordinat
es
Directio
n
T
valu
e
Cluste
r
pFWE-co
rr
Left CM mean positiv
e
Brainste
m
N
A
34 -9 -18 -15 Dep<N
C
-5.0
7
0.041
Left CM mean positiv
e
Brainste
m
N
A
93 -6 -33 -45 Dep<N
C
-4.9
4
0.001
Right CM mean positiv
e
Brainste
m
N
A
32 -12 -21 -3 Dep<N
C
-4.6
9
0.064
(0.021a)
Left SF mean positiv
e
L.
Thalamu
s
N
A
39 -12 -27 3 Dep<N
C
-4.9
1
0.017
Right SF mean positiv
e
Brainste
m
35 56 15 -15
-24
Dep<N
C
-4.7
0
0.009
Right SF mean positiv
e
L.
Thalamu
s
N
A
35 -18 -24
12
Dep<N
C
-4.3
0
0.043
dFC, dynamic functional connectivity; CM, centromedial; SF, superficial; BA, Brodmann area; L, left;
R, right; MNI, Montreal Neurological Institute; FWE, familywise error. a Uncorrected cluster P-value.
3
Figure S1: Compared with female controls, female MDD patients showed a decreased positive mean
dFC between the left CM and brainstem. Colder colors represent decreased dFC in female MDD patients.
MNI coordinates: z=-15, y=-18, x=-9 (top row); z=-45, y=-33, x=-6 (bottom row).
4
Figure S2: Compared with female controls, female MDD patients showed a decreased positive mean
dFC between the right SF and brainstem (top row) and between the bilateral SF region and left thalamus
(middle and bottom row). MNI coordinates: z=-24, y=-15, x=15 (top row); z=3, y=-27, x=-12 (middle
row); z=12, y=-24, x=-18 (bottom row).
5
Dynamic functional connectivity of amygdalar subregions without GSR
Performing global signal regression in data preprocessing is still an open question in resting-state fMRI
studies. Previous studies have suggested that global signal regression (GSR) could reduce the effects of
non-neuronal activity, such as respiration,1-3
while simultaneously introducing widespread negative
functional connectivity with ambiguous biological interpretations.4,5
Thus, we repeated our analysis
without global signal regression in the data preprocessing.
A total of 30 first-episode, drug-naive patients with MDD (18-60 years old) and 62 age- and
sex-matched NCs (16-81 years old) were included in the analysis. The image preprocessing was similar
to that of the main manuscript, except the global mean signals as a regressor. The definitions of
amygdalar subregions, resting-state dFC analysis and statistical analysis were the same as those in the
main text. Two sample t-test was used to compare the between-group difference, and computations for
positive and negative FC were performed separately. The significance threshold was set to P < 0.001 at
the voxel level with FWE correction at the cluster level.
Compared with the controls, MDD patients demonstrated impaired dFC between the left insula and
bilateral LB and left SF, between the left LB and left posterior orbital frontal cortex, and between the left
SF and brainstem (Figure S3 and S4). The detailed location and size are listed in Table S2. No
significantly decreased/increased negative dFC was found in MDD patients without the global signal
regression.
The results without global mean regression were quite different from the results with global signal
regression. This result might indicate mixed factors in the difference in dFC between patients with MDD
and NCs. Interestingly, Yang et al.6 found that performing GSR could lead to different findings in
schizophrenic patients, indicating potential disease-related information in the GS. Thus, future studies
combining electrophysiological imaging and fMRI might provide further evidence to probe the
biological significance of the GS and its effect on dFC between psychiatric patients and NCs.
6
Table S2: Regions showing significant between-group dFC differences in amygdalar subregions
without GSR
Hemisphe
re
Subregio
n
dFC
metri
c
dFC
type
Region B
A
Cluste
r size
MNI
coordinat
es
Directio
n
T
valu
e
Cluste
r
pFWE-co
rr
Left LB mean positiv
e
L.OFC
post
11 89 -21 27
-27
Dep<N
C
-5.3
1
0.033
Left LB mean positiv
e
L. Insula 13 110 -36 -18
18
Dep<N
C
-4.7
2
0.014
Right LB mean positiv
e
L. Insula 13 178 -27 12
-15
Dep<N
C
-4.6
7
0.001
Left SF mean positiv
e
Brainste
m
35 76 -9 -18 -30 Dep<N
C
-4.5
1
0.045
Left SF mean positiv
e
L. Insula 13 74 -33 -18
18
Dep<N
C
-4.2
3
0.049
dFC, dynamic functional connectivity; LB, laterobasal; SF, superficial; BA, Brodmann area; OFC,
orbital frontal cortex; L, left; R, right; MNI, Montreal Neurological Institute; Dep: depressive patients;
NC: normal controls; FWE, familywise error.
7
Figure S3: Compared with the controls, MDD patients demonstrated impaired dFC between the left LB
and left posterior orbital frontal cortex (top row), the left LB and insula (middle row), and the right LB
and insula, which extended to the basilar ganglia (bottom row). MNI coordinates: z=-27, y=27, x=-21
(top row); z=18, y=-18, x=-36 (middle row); z=-15, y=12, x=-27 (bottom row).
8
Figure S4: Compared with the controls, MDD patients demonstrated impaired dFC between the left SF
and brainstem (top row) and the left SF and left insula (bottom row). MNI coordinates: z=-30, y=-18,
x=-9 (top row); z=18, y=-18, x=-33 (bottom row).
Static functional connectivity of amygdalar subregions in MDD patients
To facilitate comparison with earlier studies of traditional static RSFC analysis, we performed traditional
static functional connectivity analysis for each amygdalar subregion by using the entire time series and
compared the different findings between the two methods.
A total of 30 first-episode drug-naive adult patients with MDD and 62 age- and sex-matched NCs
were recruited in the present study. The image preprocessing and definition of amygdalar subregions
were the same as those in the main text. Resting-state functional connectivity analysis for each
amygdalar subregion was performed by using the entire time series. Two sample t-test was used to
compare the FC difference between the MDD and NC group. The significance threshold was set to P <
0.001 at the voxel level with FWE correction at the cluster level.
Compared with the controls, MDD patients exhibited decreased positive FC between the left CM
and brainstem, the right SF and left thalamus, and the left SF and left insula (Figure S5). Compared with
the controls, MDD patients also showed increased negative static FC between the left CM and right
superior medial frontal gyrus and the right LB and right superior frontal gyrus (Figure S6). The detailed
location and size between-group differences are listed in Table S3.
Static analysis showed some similar results to the mean dFC, such as the connection of the left CM
with brainstem and superior frontal gyrus. However, dynamic functional connectivity analysis revealed
additional differences in the connectivity between the left SF and brainstem, as well as dynamic
fluctuation between the left LB and supplementary motor area (SMA), which could not be identified
9
using static RSFC analysis. The present dFC results provided additional results that are complementary
to the static FC, confirming that the newly developed dFC analysis strategies might provide novel
understandings of the pathology of MDD.
Table S3: Regions showing significant between-group static FC differences in amygdalar
subregions
Hemisphe
re
Subregi
on
FC
type
Region B
A
Clust
er
size
MNI
coordinat
es
Directio
n
T
valu
e
Cluste
r
pFWE-co
rr
Left CM positiv
e
Brainstem N
A
103 12 -30
-45
Dep<N
C
-4.8
8
0.001
Left CM negativ
e
R.
Frontal_Sup_Me
dial
10 90 15 66 12 Dep>N
C
4.97 0.006
Right LB negativ
e
R. Frontal_Sup 10 57 18 72 9 Dep>N
C
4.17 0.037
Left SF positiv
e
L. Insula 13 52 -27 -27
15
Dep<N
C
-4.5
5
0.016
Right SF positiv
e
L. Thalamus N
A
45 -18 -24
12
Dep<N
C
-4.1
2
0.041
FC, functional connectivity; CM, centromedial; LB, laterobasal; SF, superficial; BA, Brodmann area; L,
left; R, right; MNI, Montreal Neurological Institute; Dep, depressive patients; NC, normal controls;
FWE, familywise error.
10
Figure S5: Compared with the controls, MDD patients showed decreased positive FC between the left
CM and brainstem (top row), the right SF and left thalamus (middle row), and the left SF and left insula
(bottom row). MNI coordinates: z=-45, y=-30, x=12 (top row); z=12, y=-24, x=-18 (middle row); z=15,
y=-27, x=-27 (bottom row).
11
Figure S6: Compared with the controls, MDD patients showed increased negative static FC between the
left CM and right superior medial frontal gyrus (top row) and the right LB and right superior frontal
gyrus (bottom row). MNI coordinates: z=12, y=66, x=15 (top row); z=9, y=-72, x=18 (bottom row).
12
Dynamic functional connectivity of the whole amygdala
To assess whether our findings derived by using the subregions of the amygdala as seeds could be
detected by using the whole amygdala as ROIs, we reperformed our analysis by combining the three
amygdalar subregions as one seed.
A total of 30 first-episode, drug-naive adult patients with MDD and 62 age- and sex-matched NCs
were recruited in the present study. The image preprocessing was the same as that in the main text;
however, the seeds were defined as the whole amygdala.
Resting-state dFC analysis and statistical analysis for the whole amygdala were the same as those
in the main text. Two sample t-test was used to compare the dFC difference between the MDD and NC
group. Computations for positive and negative FC were performed separately. The significance
threshold was set to P < 0.001 at the voxel level with FWE correction at the cluster level.
We did not find any significant between-group differences in the mean or variability of the dFC
when using the whole amygdala as seeds, suggesting the necessity of dividing the amygdala into
subregions.
13
Validation analyses of sliding-window widths
Given that the choice of window length remains controversial, two additional window lengths (160 s and
240 s) were used to validate the main results.
A total of 30 first-episode, drug-naive patients with MDD (18-60 years old) and 62 age- and
sex-matched NCs (16-81 years old) were included for further analysis. The image preprocessing,
definition of amygdalar subregions, and statistical analysis were the same as those in the main text,
except that the resting-state dFC analysis of the window lengths was performed by using 160 s and 240 s.
Statistical group difference maps were constructed using a GLM with age and sex as covariates. Two
sample t-test was used to compare the between-group difference. The statistical significance threshold
was set at P < 0.001 at the voxel level with a FWE-corrected P-value < 0.05 at the cluster level using
SPM.
Compared with the controls, MDD patients showed increased variability of positive dFC between
the right CM and SMA and the left LB and bilateral SMA (Figure S7) by using a sliding-window length
of 80 time points. Compared with the controls, MDD patients also showed decreased negative dFC
between the left CM and right superior medial frontal gyrus; moreover, the patients exhibited decreased
positive dFC between the left CM and brainstem, left SF and brainstem, and left SF and left thalamus
(Figure S8). The detailed location and size between-group differences using a sliding-window length of
80 time points are listed in Table S4. The location and size of between-group dFC differences in
amygdalar subregions by using a sliding-window length of 120 time points were similar to the results
using 80 time points, except there was no significant variability difference in the dFC by using a
sliding-window length of 120 time points (Table S5 and Figure S9).
The similar between-group dFC differences in amygdalar subregions by using different
sliding-window lengths (window length of 80, 100, and 120 time points) indicate that our main result is
reliable and repeatable. In addition, only 80 time points could detect the group differences in the
variability in the dFC, suggesting that short-interval dynamic FC may be more sensitive to
between-group dFC differences. Similarly, 100 time points also detect a non-significant trend toward an
increase of the variability of the positive dFC between the left LB region and right SMA in MDD
patients compared with that in NCs (P = 0.010, uncorrected, Figure S10). Consistent with our result,
Wilson et al.7 also found that short temporal windows could obtain dynamic information and that FC
variability increased with short epoch length. Since there is an increased likelihood of spurious
correlations when using shorter intervals,8 we chose 100 time points as our main analytic results.
14
Table S4: Regions showing significant between-group dFC differences in amygdalar subregions by
using a sliding-window length of 80 time points
Hemisph
ere
Subregi
on
dFC
metric
dFC
type
Region B
A
Clust
er
size
MNI
coordina
tes
Directi
on
T
val
ue
Clust
er
pFWE-c
orr
Left CM mean negati
ve
R.
Frontal_Sup_M
edial
10 46 15 66 12 Dep<
NC
-4.4
5
0.047
Left CM mean positi
ve
Brainstem N
A
93 12 -30
-45
Dep<
NC
-4.6
3
0.002
Right CM mean negati
ve
Brainstem N
A
39 -12 -51
-33
Dep<
NC
-3.9
7
0.071
(0.02
1a)
Right CM variabil
ity
positi
ve
L. SMA 6 25 -9 3 48 Dep>
NC
4.0
5
0.013
Left LB variabil
ity
positi
ve
R. SMA 6 16 3 -12 63 Dep>
NC
4.0
7
0.043
Left SF mean positi
ve
Brainstem 35 52 -9 -15
-27
Dep<
NC
-4.6
4
0.013
Left SF mean positi
ve
L. Thalamus N
A
38 -12 -27
3
Dep<
NC
-4.0
3
0.037
dFC, dynamic functional connectivity; CM, centromedial; LB, laterobasal; SF, superficial; BA,
Brodmann area; SMA, supplementary motor area; L, left; R, right; MNI, Montreal Neurological Institute;
Dep, depressive patients; NC, normal controls; FWE, familywise error. The coordinates represent the
position of the voxel with the highest intensity in Montreal Neurological Institute standard space. a Uncorrected cluster P-value.
15
Figure S7: The group differences in the variability of the dFC in the amygdalar subregions between the
right CM and left supplementary motor area (top row) and the left LB and bilateral supplementary motor
area (bottom row) by using a sliding-window length of 80 time points. MNI coordinates: z=48, y=3,
x=-9 (top row); z=63, y=-12, x=3 (bottom row).
16
Figure S8: The group differences in the decreased negative dFC between the left CM and right superior
medial frontal gyrus (top row); the group differences in the decreased positive dFC between the left CM
and brainstem (second row), left SF and brainstem (third row), and left SF and left thalamus (bottom
row) by using a sliding-window length of 80 time points. MNI coordinates: z=12, y=66, x=15 (top row);
z=-45, y=-30, x=12 (second row); z=-27, y=-15, x=-9 (third row); z=3, y=-27, x=-12 (bottom row).
17
Table S5: Regions showing significant between-group dFC differences in amygdalar subregions by
using a sliding-window length of 120 time points
Hemisph
ere
Subregi
on
dFC
metr
ic
dFC
type
Region B
A
Clust
er
size
MNI
coordina
tes
Directi
on
T
valu
e
Cluste
r
pFWE-c
orr
Left CM mea
n
negati
ve
R.
Frontal_Sup_M
edial
10 46 15 66 12 Dep<N
C
-4.4
5
0.048
Left CM mea
n
positi
ve
Brainstem N
A
99 12 -30
-45
Dep<N
C
-4.6
5
0.001
Right CM mea
n
positi
ve
Brainstem N
A
41 -12 -51
-33
Dep<N
C
-4.0
1
0.062
(0.01
8a)
Left SF mea
n
positi
ve
Brainstem 35 50 -9 -15
-27
Dep<N
C
-4.6
8
0.015
Left SF mea
n
positi
ve
L. Thalamus N
A
39 -12 -27 3 Dep<N
C
-4.0
3
0.034
dFC, dynamic functional connectivity; CM, centromedial; SF, superficial; BA, Brodmann area; L, left; R,
right; MNI, Montreal Neurological Institute; Dep, depressive patients; NC, normal controls; FWE,
familywise error. a Uncorrected cluster P-value.
18
Figure S9: The group differences in the decreased negative dFC between the left CM and right superior
medial frontal gyrus (top row); the group differences in the decreased positive dFC between the left CM
and brainstem (second row), left SF and brainstem (third row), and left SF and left thalamus (bottom
row) by using a sliding-window length of 120 time points. MNI coordinates: z=12, y=66, x=15 (top
row); z=-45, y=-30, x=12 (second row); z=-27, y=-15, x=-9 (third row); z=3, y=-27, x=-12 (bottom
row).
19
Figure S10: Compared with the NC group, the MDD group showed a trend toward a decreased positive
mean dFC between the right CM and right brainstem (P = 0.024, uncorrected, top row) as well as
increased variability of the positive dFC between the left LB and right SMA (P = 0.010, uncorrected,
bottom row). MNI coordinates: z=-33, y=-51, x=-12 (top row); z=63, y=-12, x=3 (bottom row).
20
Validation analyses by using FD as an additional covariate
Because the mean framewise displacement (FD) differed between the two groups (P=0.03), we
reperformed the between-group GLM tests on the dFC metrics by adding FD as a covariate to reduce the
motion effect.
A total 30 first-episode, drug-naive patients with MDD (18-60 years old) and 62 age- and
sex-matched NCs (16-81 years old) were included for further analysis. The image preprocessing,
definition of amygdalar subregions, and statistical analysis were the same. Statistical group difference
maps were constructed using a GLM with age, sex and FD as covariates. The statistical significance
threshold was set for P < 0.001 at the voxel level with a FWE-corrected P-value < 0.05 at the cluster
level using SPM.
Compared with the controls, MDD patients showed decreased positive dFC between the left CM
and brainstem and decreased negative dFC between the right LB and left orbital frontal cortex by adding
FD as a covariate (Table S6 and Figure S11).
Decreased positive dFC between the left CM and brainstem in MDD remained in the additional
analysis by using FD as an additional covariate. Furthermore, decreased negative dFC between the right
LB and left orbital frontal cortex was also found in MDD patients by adding FD as a covariate, which
was not found in the main result. However, we noticed that most of the other connections with
between-group differences occurred with reduced significance at an uncorrected cluster P < 0.05, which
might imply potential effects of the mean FD on dFC. A study with a larger sample size might overcome
this issue.
21
Table S6: Regions showing significant between-group dFC differences by using FD as an
additional covariate
Hemisphe
re
Subregio
n
dFC
metri
c
dFC
type
Region B
A
Cluste
r size
MNI
coordinat
es
Directio
n
T
valu
e
Cluste
r
pFWE-co
rr
Left CM mean positiv
e
Brainste
m
N
A
73 -6 -36 -45 Dep<N
C
-4.2
5
0.006
Right LB mean negativ
e
L. OFC 11 50 -9 63 -15 Dep<N
C
4.22 0.046
FC, functional connectivity; CM, centromedial; LB, laterobasal; OFC, orbital frontal cortex; BA,
Brodmann area; L, left; MNI, Montreal Neurological Institute; Dep, depressive patients; NC, normal
controls; FWE, familywise error.
22
Figure S11: The group dFC differences in decreased positive dFC between the left CM and brainstem
(top row) and decreased negative dFC between the right LB and left orbital frontal cortex (bottom row)
by using FD as an additional covariate. MNI coordinates: z=-45, y=-36, x=-6 (top row); z=-15, y=63,
x=-9 (bottom row).
23
The correlation between the time series of the amygdalar subregions
We did not perform orthogonalization on the time series of the amygdalar subregions for two main
reasons. First, the orthogonalization might alter the dFC between the seeds and the other brain voxels
since the time series of the other voxels was not processed. Thus, the results could not reflect the
original characteristics of whole-brain dFC of the amygdalar subregions. Second, according to previous
studies, the patterns of resting-state functional connectivity exhibited both convergence and divergence
across different amygdalar subregions.9,10
Our results revealed that the dFC patterns were quite similar
between the CM and SF, which is in line with previous findings. The procedure of orthogonalization
might reduce the convergence between the subregions and thus bias the results. Here, we estimated the
similarity between the time series of these seeds for each subject by calculating Pearson’s correlation. As
expected, we found that the bilaterally symmetric subregions had the most similar brain activities
(correlation coefficients in NCs: CM, 0.61 ± 0.16; LB, 0.71 ± 0.13; SF, 0.69 ± 0.16. in MDDs: CM, 0.58
± 0.20; LB, 0.63 ± 0.12; SF, 0.65 ± 0.15). The correlations between the CM and LB were the smallest (in
NCs: 0.33-0.44; in MDDs: 0.29-0.43), while the SF exhibited relatively similar spontaneous activity
patterns to those of the other areas (in NCs: 0.43-0.63; in MDDs: 0.38-0.63) (Table S7).
24
Table S7: Correlation between the time series of the seeds
In Normal Controls
In MDD
Patients
Left CM Right CM Left LB Right LB Left SF Right SF
Left CM NA 0.61 ±
0.16
0.44 ±
0.19
0.33 ±
0.19
0.58 ±
0.16
0.51 ±
0.19
Right
CM
0.58 ±
0.20
NA 0.34 ±
0.20
0.42 ±
0.18
0.43 ±
0.20
0.59 ±
0.17
Left LB 0.41 ±
0.21
0.33 ±
0.21
NA 0.71 ±
0.13
0.63 ±
0.13
0.56 ±
0.17
Right LB 0.29 ±
0.19
0.43 ±
0.16
0.63 ±
0.16
NA 0.43 ±
0.22
0.61 ±
0.14
Left SF 0.52 ±
0.16
0.38 ±
0.19
0.63 ±
0.12
0.38 ±
0.16
NA 0.69 ±
0.16
Right SF 0.47 ±
0.19
0.58 ±
0.15
0.54 ±
0.19
0.57 ±
0.15
0.65 ±
0.15
NA
CM, centromedial; LB, laterobasal; SF, superficial; MDD, major depressive disorder.
25
References
1. Fransson P. Spontaneous low-frequency BOLD signal fluctuations: an fMRI investigation of the
resting-state default mode of brain function hypothesis. Hum Brain Mapp 2005;26(1):15-29.
2. Fox MD, Zhang D, Snyder AZ, et al. The global signal and observed anticorrelated resting state
brain networks. J Neurophysiol 2009;101(6):3270-83.
3. Birn RM, Diamond JB, Smith MA, et al. Separating respiratory-variation-related fluctuations
from neuronal-activity-related fluctuations in fMRI. Neuroimage 2006;31(4):1536-48.
4. Murphy K, Birn RM, Handwerker DA, et al. The impact of global signal regression on resting
state correlations: are anti-correlated networks introduced? Neuroimage 2009;44(3):893-905.
5. Weissenbacher A, Kasess C, Gerstl F, et al. Correlations and anticorrelations in resting-state
functional connectivity MRI: a quantitative comparison of preprocessing strategies. Neuroimage
2009;47(4):1408-16.
6. Yang GJ, Murray JD, Repovs G, et al. Altered global brain signal in schizophrenia. Proc Natl
Acad Sci U S A 2014;111(20):7438-43.
7. Wilson RS, Mayhew SD, Rollings DT, et al. Influence of epoch length on measurement of
dynamic functional connectivity in wakefulness and behavioural validation in sleep. Neuroimage
2015;112:169-79.
8. Hutchison RM, Womelsdorf T, Gati JS, et al. Resting-state networks show dynamic functional
connectivity in awake humans and anesthetized macaques. Hum Brain Mapp
2013;34(9):2154-77.
9. Roy AK, Shehzad Z, Margulies DS, et al. Functional connectivity of the human amygdala using
resting state fMRI. Neuroimage 2009;45(2):614-26.
10. Etkin A, Prater KE, Schatzberg AF, et al. Disrupted amygdalar subregion functional connectivity
and evidence of a compensatory network in generalized anxiety disorder. Arch Gen Psychiatry
2009;66(12):1361-72.