Can voxel based morphometry, manual segmentation and automated segmentation
equally detect hippocampal volume differences in acute depression?
Loretxu Bergouignana,⁎, Marie Chupinb,g,i, Yvonne Czechowskac,d,e, Serge Kinkingnéhunf,g,h,i,
Cédric Lemognea,j, Guillaume Le Bastarda, Martin Lepagec,d,e, Line Garnerob,g,i,
Olivier Colliotb,g,i, Philippe Fossatia
aCNRS UMR 7593, Emotion Center, Department of Psychiatry, Groupe Hospitalier Pitié-Salpétrière, 47-83 Boulevard de l'Hôpital, Paris 75013 and IFR Neuroscience, France
bCNRS UPR 640, LENA, Cognitive Neuroscience and Brain Imaging Laboratory, 47 Boulevard de l'Hôpital, Paris 75013 and IFR Neuroscience, France
cBrain Imaging Group, Douglas Hospital Research Centre, Verdun, Canada
dDepartment of Psychiatry, McGill University, Montreal, PQ, Canada
eDepartment of Neurology, McGill University, Montreal, PQ, Canada
fINSERM U610, Paris, France
gIFR-49, Orsay, France
he(ye)BRAIN, Paris, France
iUPMC Univ Paris 06, France
jDepartment of C-L Psychiatry, European Georges Pompidou Hospital, 75015 Paris, France
a b s t r a c t a r t i c l ei n f o
Received 30 May 2008
Revised 5 November 2008
Accepted 10 November 2008
Available online 21 November 2008
Context: According to meta-analyses, depression is associated with a smaller hippocampus. Most magnetic
resonance imaging (MRI) studies among middle aged acute depressed patients are based on manual
segmentation of the hippocampus. Few studies used automated methods such as voxel-based morphometry
(VBM) or automated segmentation that can overcome certain drawbacks of manual segmentation
(essentially intra- and inter-rater variability and operator time consumption).
Objective: The aim of our study was to compare the sensitivity of manual segmentation, automated
segmentation and VBM to detect hippocampal structural changes in middle aged acute depressed
Method: Twenty-one middle aged depressed inpatients and 21 matched controls were compared regarding
their hippocampal structure using VBM with SPM5, manual segmentation and an automated segmentation
algorithm. The VBM-ROI analysis was performed using two different normalization methods: the standard
approach implemented in SPM5 and the most recent DARTEL algorithm.
Results: Using VBM-DARTEL, when corrected for multiple comparisons, significant volume differences were
detected between groups in different regions and more specifically in hippocampus with ROI analyses.
Whereas using standard VBM (without DARTEL), ROI analyses did not show bilateral volume between group
Significant hippocampal volume reductions between patients and controls were also detected using manual
segmentation (−11.6% volume reduction, pb0.05) and automated segmentation (−9.7% volume reduction,
pb0.05). VBM-DARTEL and automated segmentation show equal sensitivity in detecting hippocampal
differences in depressed patients, while standard VBM was unable to detect hippocampal changes. Both
VBM-DARTEL and automated segmentation could be used to perform large scale volumetric studies in
humans. The new automated segmentation technique could further explore and detect hippocampal subpart
differences that could be very useful for clarifying physiopathology of psychiatric disorders.
© 2008 Elsevier Inc. All rights reserved.
The hippocampus is a central component of the limbic system and
has a complex set of interconnections with limbic elements involved
in emotional processing (Nieuwenhuys et al., 1988). Depression is
characterized by emotional impairment. Hippocampal abnormalities
could then have a pathophysiologic role in depression regardless of
their ultimate etiology.
Several neuroimaging studies evaluating structural changes of the
hippocampus in acute depression have reported significant volume
reduction in patients compared to healthy subjects (Frodl et al., 2002,
2006; Bell-McGinty et al., 2002; Saylam et al., 2006; Maller et al.,
2007; Colla et al., 2007; Ballmaier et al., 2008; Vasic et al., 2008). Two
NeuroImage 45 (2009) 29–37
⁎ Corresponding author. Fax: +33 153790770.
E-mail address: email@example.com (L. Bergouignan).
1053-8119/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/ynimg
meta-analyses reviewing 12 studies and over 300 patients concluded
thathippocampalvolume is reduced in unipolardepression (Videbech
and Ravnkilde, 2004; Campbell et al., 2004). Although the patient
populations were highly heterogeneous regarding age, gender
distribution, age at onset of the disorder, average number of
depressive episodes, and response to treatment, depression was
associated with significant reduction in hippocampal volume in both
hemispheres. Videbech and Ravnkilde (2004) reported a weighted
average reduction of hippocampal volume of 8% on the left side and
10% on the right side.
Nevertheless, some studies did not observe any difference in
hippocampal volume between acute depressed patients and controls
(Mervaala et al., 2000; Vakili et al., 2000; Von Gunten et al., 2000;
Rusch et al., 2001; Posener et al., 2003; Caetano et al., 2004; Hastings
et al., 2004). Clinical characteristics of the studied acute depressed
populations, such as first vs. multiple depressive episodes, duration of
illness and presence of sexual abuse mayaccount for the discrepancies
between findings; however the divergent results could also be partly
explained by methodological differences.
In the vast majority of these studies, hippocampal volumetry was
performed using a manual segmentation protocol. Divergent measur-
ing protocols could explain the divergent findings (Geuze et al., 2005).
Furthermore, manual segmentation of the hippocampus is operator
timeconsuming,requires specific anatomicalexpertise and may result
in high intra- and inter-rater variability.
Two studies used voxel-based morphometry (VBM) with SPM
(Bell-McGinty et al., 2002; Vasic et al., 2008). VBM is a fully automatic
technique which allows an objective analysis of anatomical differ-
ences between groups across the whole-brain. It involves avoxel-wise
comparison between two groups of subjects of the local concentration
of gray matter or volume comparison using Jacobian modulation
(Ashburner and Friston, 2000). VBM has been applied in different
types of neuropsychiatric pathologies. Among psychiatric disorders,
several studies have used VBM in patients with schizophrenia
(Seidman et al., 1999; Kubicki et al., 2002; Job et al., 2002, 2003;
Moorhead et al., 2004; Borgwardt et al., 2007a; Rametti et al., 2007),
and all of these studies reported hippocampal volume differences
between patients and healthy subjects. To the best of our knowledge,
only Bell-McGinty et al. (2002) and Vasic et al. (2008) have applied
VBM in acute depressed patients. Bell-McGinty et al. (2002), using
SPM99, reported gray matter concentration differences in the
hippocampus between patients and controls but their p-value was
not corrected for multiple comparisons (p=0.001, uncorrected), and
their research population was composed exclusively of elderly
subjects. Recently, Vasic et al. (2008) interested in the relationship
between gray matter (GM) abnormalities, psychopathology and
cognitive impairment, studied GM concentration and volume differ-
ences between middle aged acute depressed patients and controls.
Using SPM5, they have shown significant volume differences for the
left hippocampus. However, as these volume differences were
observed with an uncorrected voxel level height threshold of
pb0.001, it is unclear whether, in middle aged depressed patients,
VBM can detect hippocampal volume differences when using a
properly corrected statistical threshold.
Also, a preprocessing step of the VBM in SPM has recently been
improved with the Diffeomorphic Anatomical Registration using
Exponentiated Lie algebra (DARTEL) registration method (Ashburner,
2007). This technique, being more deformable, notably improves the
realignment of small inner structures (Yassa et al. 2008). To the best
of our knowledge, no studies have yet used DARTEL in acute
depressed patients. More generally, this very recent technique has
only been used in two VBM studies (Martino et al. 2008; Stonnington
et al. 2008).
In addition, one study used a semi-automatic hippocampal
segmentation method (Posener et al., 2003). Automated and semi-
automatic methods have been developed for systematic segmentation
of thehippocampus.Thesemethodsareless operator time-consuming
and generate less inter and intra-rater variability than manual
segmentation (Csernansky et al., 2002; Fischl et al., 2002; Chupin et
al., 2007). Using a semi-automated segmentation of the hippocampus
(Csernansky et al., 2002, Posener et al. (2003)) were able to detect
hippocampal changesin acute depressedpatients.Chupinet al. (2007)
have recently developed an automated method, called SACHA
(“Segmentation Automatisée Compétitive de l'Hippocampe et de
l'Amygdale”), to segment the hippocampus and the amygdala. This
method has been validated by comparison with manual tracing in
healthy controls and in patients with Alzheimer's disease (AD)
(Chupin et al., 2007). It has also been successfully applied to detect
significant hippocampal volume reduction in patients with AD and in
patients with mild cognitive impairment (MCI) (Colliot et al., 2008).
However, hippocampal volume reductions in AD and in MCI are more
severe than in acute depressed patients, with the latter average
reduction ranging at 9% according to meta-analyses (Campbell et al.,
2004; Videbech et al., 2004).
The purpose of our study was to compare the sensitivity of VBM,
manual segmentation and automated segmentation to detect hippo-
campal volume differences in depression. To achieve this goal we
compared VBM with two segmentation methods (a manual segmen-
tation method and the automated method SACHA) in the same groups
of middle aged acute depressed patients and healthy controls.
Depressed inpatients fulfilling the DSM-IV criteria for a major
depressive episode (unipolar depressive disorder) were recruited
from the psychiatry department of the Pitié-Salpêtrière Hospital.
Healthy controls with no history of psychiatric disorders were
recruited from the community to match the patients for age and
level of education. All participants were right-handed. Participants
were screened for past or current DSM-IV axis I diagnoses by two
psychiatrists (CL and GL) with the Mini International Neuropsychiatric
Interview (Sheehan et al., 1998). Severity of depression was assessed
using the Montgomery–Asberg Depression Rating Scale (MADRS)
(Montgomery and Asberg, 1979) and the Beck Depression Inventory
(BDI) (Beck et al., 1961). Exclusion criteria were history of manic
episode, psychotic features, neurological illness, medical disorders or
medication likely to affect cognition, history of substance-related
disorders or electroconvulsive therapy in the previous 12 months.
Written informed consent was obtained for each participant. The
study was approved by the Ethics Committee for Biomedical Research
of the Pitié-Salpêtrière Hospital.
High-resolution three-dimensional (3D) T1-weighted images were
acquired on a 1.5-T whole-body scanner (SIGNA, GE, Milwaukee, WS,
USA). The MRI parameters of the 3D magnetization-prepared rapid
gradient-echo (3D-MPRAGE) (Mugler and Brookeman, 1991)
sequences were as follows: TR=10.2 ms; TE=2.04 ms; TI=400 ms;
FOV=256 mm, flip angle=10°. The images were acquired with an in-
plane spatial resolution of 0.9375 mm and with 106 contiguous
sagittal 1.5 mm thick slices for 39 participants and with 124
contiguous sagittal 1.3 mm thick slices for 3 participants. Thus, nearly
isotropic three-dimensional MR data sets were obtained, making
highly accurate volumetric MR measurements possible.
The three methodologies
The main goal of our study was to compare each method's
(manual, automated, VBM) sensitivity to detect subtle hippocampal
L. Bergouignan et al. / NeuroImage 45 (2009) 29–37
volume differences between the depressed and normal control
groups. We then matched the statistical designs of all three analyses
as close as possible. We have included gender and age as covariates in
the three statistical analyses.
Voxel-based morphometry (Ashburner and Friston, 2000) was
performed using SPM5 (Statistical Parametric Mapping, Wellcome
Department of Imaging Neuroscience, London, UK). The VBM pre-
processing included five steps:
(1) Check for scanner artefacts and gross anatomical abnormalities
for each subject. (2) Set image origin at the Anterior Commissure AC.
(3) Use of Hidden Markov Random Field (HMRF) option in the
segmentation part of the VBM5 toolbox to minimize the noise level of
the segmentation. (4) Use DARTEL toolbox to have a high-dimensional
normalization protocol. We followed John Ashburner's chapter in its
standard version including the MNI space transformation (Ashburner,
2007) (see Fig. 1), but we performed the MNI space transformation
with Donald McLaren's script that modifies John Ashburner's script of
code for transforming DARTEL templates and images to MNI space
(see Appendix A, Fig.1). (5) Check for homogeneity across sample and
use the standard version of the smoothing (i.e. 8). After this pre-
processing we obtained smoothed modulated normalized data that
we used for the statistical analysis.
Standard VBM analysis
For comparison on hippocampal difference sensitivity, we also
repeated the ROI analysis using the more widely used standard SPM5
segmentation code (Ashburner and Friston, 2005) instead of the
diffeomorphic registration algorithm. For this standard version we
used HMRF segmented modulated images (obtained with step 3) and
smoothed those images at 12.
Whole-brain volume comparison
The two groups (depressed patients and controls) were compared
using multiple factorial comparisons with the sample as studied
As the Jacobian modulation takes into account both local and
global normalization, we had to correct for head-size, we thus
added the Total Intracranial Volume (TIV) as covariate in the
statistical analysis. TIV was calculated as the sum of GM, White
Matter (WM) and Cerebrospinal Fluid (CSF) volumes, derived from
SPM5 toolbox's HMRF segmentations. Thus, an analysis of covar-
iance (ANCOVA) was computed to detect differences in gray matter
volume between groups. Gender, age and TIV were included as
nuisance covariates. To correct for multiple comparisons, we applied
the false discovery rate (FDR) approach (Genovese et al., 2002),
which controls the expected proportion of false positives among
suprathreshold voxels (Benjamini and Hochberg, 2000). We used a
threshold of pb0.05.
ROI analysis using hippocampal masks (MARINA)
Since the purpose of our study was to assess hippocampal changes,
we also performed a region of interest (ROI) analysis restricted to the
hippocampus. Indeed, when an a priori hypothesis is made on a
specific region, an ROI analysis is more sensitive than a whole-brain
comparison. We created a mask of right and left hippocampus with
the MARINA software (MAsks for Region of Interest Analysis, Version
Fig. 1A. HMRF and DARTEL preprocessing in VBM 5 in a subject of our control group.
L. Bergouignan et al. / NeuroImage 45 (2009) 29–37
0.6.1, B. Walter, Giessen, Germany, 2002). This mask was then inserted
as the explicit mask in our VBM factorial analyses.
When doing ROI analyses in standard VBM: analysis of covariance
(ANCOVA) computed on HMRF segmented modulated and smoothed
images between groups with gender, age and TIV included as nuisance
Within the hippocampal mask, weset significance at a threshold of
Manual segmentation of the hippocampus
We segmented the hippocampus according to the protocol
introduced by Pruessner et al. (2000). Prior to manual hippocampal
segmentation, the following preprocessing steps were performed:
non-uniformity correction, linear registration to standard stereotaxic
space (using the ICBM 152 template), and signal intensity normal-
ization. Volumetric analysis was performed with the interactive
software package DISPLAY developed at the Brain Imaging Center of
the Montreal Neurological Institute. This program allows visualization
of MR images in all orientations. The protocol includes the fimbria,
alveus, dentate gyrus, cornu ammonis and excludes the subiculum.
Measurements were performed by a trained rater (LB), who was blind
to group membership and clinical information. The whole procedure
takes about 90 min for each side. To assess intra-rater reliability,
volumetric measurement of the hippocampus was carried out twice
for 10 participants; the time between two successive sessions was
8–12 months. Intra-rater reliability was assessed using Pearson's
intra-class correlation (Fig. 2).
To test for differences between depressed patients and controls, a
repeated measures ANCOVA (analysis of covariance) was used, with
one between-subject grouping factor (patients and controls), one
within-subject factor (laterality: left, right) and gender and age as
Automated segmentation of the hippocampus using SACHA
Automated segmentation was performed using the SACHA soft-
ware (Chupin et al., 2007) included in the Brainvisa image analysis
platform (Institut Fédératif de Recherche de Neuroimagerie, IFR 49,
Saclay, France, Cointepas et al., 2001, http://www.brainvisa.info/).
SACHA extracts the hippocampus from a native scan according to the
protocol described in Chupin et al. (2007), Appendix C. As in the
manual protocol, the fimbria, alveus, dentate gyrus, cornu ammonis
are included and the subiculum excluded.
This software relies on region deformation, introducing a competi-
tion between the hippocampus and the amygdala. It gives the volume
of bothregions; hereweonly took hippocampalvolumes into account.
The method requires the following initialization from the operator.
First, a bounding box is manually defined around the amygdalo–
hippocampal complex (Fig. 3, 1st step) by selecting six slices that
correspond to the limits of the hippocampus and the amygdala. The
dimension of the bounding box is typically around 30×50×20 voxels.
Then, two seeds are placed, one in the hippocampus and one in the
amygdala (Fig. 3, 2nd step). These seeds constitute the starting points
of the deformation process. They are positioned close to the centre of
the amygdala and the center of the head of the hippocampus. Lastly,
starting from these two seeds, the algorithm automatically aggregates
voxels and converges tothe segmentation of the two structures (Fig. 3,
‘Final segmentation’). Additionally, two parameters of the algorithm
can be adjusted: a radiometric parameter and a geometric parameter.
1) The radiometric parameter controls the ratio between the intensity
characteristics (mean and Standard Deviation (SD)) of the hippocam-
pus and the amygdala and those of the gray matter. The ratio between
the mean and SD of the intensity of the amygdala and the
Fig. 2. Left hippocampus segmented with Pruessner's manual segmentation protocol in a subject of our patient group: l. HC=left hippocampus; r. HC=right hippocampus.
Fig. 1B. Used mask for VBM-ROI analyses.
L. Bergouignan et al. / NeuroImage 45 (2009) 29–37
hippocampus and those of gray matter may depend on image contrast.
For this reason, two preset values controlling these intensity ratios can
be chosen depending on the visual contrast of the image. 2) The
geometric parameter controls the degree of influence of an anisotropic
regularisation strategy taking into account the particular shape in the
a parameter called tail's anisotropy. These parameters can be adjusted
whenthehippocampus is atrophied. Thedefaultradiometric parameter
is 1 and the tail's anisotropy is at 1. The whole procedure takes about
10 min for each side. Measurements were performed by a trained rater
(LB) who was blind to group membership and clinical information.
The SACHA algorithm has been validated only on native data and
not on normalized data. We used this standard version and obtained
account brain inter-subject variability during the statistical analyses.
We took TIV into account in the statistical analyses. ANCOVA was
used to test for differences between depressed patients and controls,
with one two-level-between-subject factor (group: patients, con-
trols), one within-subject factor (laterality: left, right) and gender, age
and TIV as covariates.
We added an ANCOVA with one two-level-between-subject factor
(group: patients, controls), one within-subject factor (laterality: left,
ensure that the segmentation of the CSF did not influence the results.
Twenty-one unipolardepressed inpatients (17 women, 4 men) and
21 healthy controls (14 women, 7 men) were recruited. All controls
had no history of psychiatric disorder. Clinical and demographic
characteristics of the participants are displayed in Table 1.
All patients were taking antidepressants and were tested within
the first week of receiving their treatment. Sedative drugs were not
allowed on the experiment day. One male depressed inpatient was
excluded from the study population due to an abnormal enlargement
of the lateral ventricles revealed by the MRI scanning. This patient was
hospitalized due to suicidal thoughts and subsequently experienced
atypical memory symptoms.
Whole-brain analysis with DARTEL-VBM
After the segmentation and the normalization, no outlier was found
when checking for homogeneity. No subject was excluded from the
Clinical and demographic characteristics of participants
Mean s.d.Min Max Mean s.d.Min Max test t
Number of MDE
Age on the first
Duration of the
9.58 20 49 28.21 5.50
23.8 8.65 1346
4.56 3.221 12
28.71 6.82 18
19.36 4.67 13
MDE = Major Depression Episode. MADRS = Montgomeryand Asberg Depression Rating
Scale. BDI = Beck Depression Inventory.
Fig. 3. Right hippocampus segmented with SACHA software in a subject of our control group: HC=hippocampus; AMG=amygdala.
Regions of GM volume difference in VBM5 with FDR correction
Cingulate gyrus (BA 24)
Middle temporal gyrus
Superior parietal lobule
Inferior semi-lunar lobule
46 5.77 (7.68)
L. Bergouignan et al. / NeuroImage 45 (2009) 29–37
The whole-brain VBM analysis showed significant volume differences in
different regions including the parahippocampal gyrus, but not hippo-
campus (see Table 2), using a threshold of pb0.05 and the FDR method.
VBM-DARTEL. When doing ROI analyses with the SPM5 VBM-DARTEL
procedure: analysis of covariance (ANCOVA) with a threshold of
pb0.001 uncorrected showed significant bilateral hippocampal
differences (see Fig. 4A).
When doing ROI analyseswith the standard SPM5 VBM procedure:
analysis of covariance (ANCOVA) with a threshold of pb0.001
uncorrected showed slight significant differences restricted to the
left hippocampal region. However, the small region that appears is
outside the hippocampus, and it is therefore difficult to conclude that
this difference is related to hippocampal changes (see Fig. 4B).
Manual hippocampal volumetry
The intra-rater reliability (intra-class correlation) was 0.89 for the
left hippocampus and 0.82 for the right hippocampus.
Fig. 4A. Hippocampal volume difference observed in VBM5, including DARTEL preprocessing, in MARINA hippocampus restricted ROI analyses with pb0.001 uncorrected threshold.
Fig. 4B. Hippocampal volume difference observed in VBM5, without DARTEL preprocessing, in MARINA hippocampus restricted ROI analyses with pb0.001 uncorrected threshold.
L. Bergouignan et al. / NeuroImage 45 (2009) 29–37
Meanleft hippocampalvolumes were 3.95cm3±0.75(range: 2.25–
5.41) for depressed patients and 4.42 cm3±1.04 (range: 2.36–6.93) for
control participants. Right hippocampal volumes were 3.93 cm3±0.80
(range: 2.32–5.70) for depressed patients and 4.50 cm3±1.06 (range:
2.47–7.16) for control participants. The ANCOVA with hippocampal
volume as the dependent variable and gender and age as covariates
revealed no side effect (F=0.514, df=38, p=0.478) or group×side
interaction (F=0.031, df=38, p=0.861). However a significant group
effect was revealed (F=4.718, df=38, pb0.05). The mean hippocampal
volume reduction in patients compared to controls was 11.6% (10.6%
on the left and 12.6% on the right). Detailed results are presented in
Automated hippocampal volumetry
Seventy-four (74/84=88%) hippocampi were segmented with the
default parameters. Ten segmentations needed parameter adjust-
ments: for radiometric parameters, five were set at 0 (bad contrast);
for tail's geometric parameter two had the tail's anisotropy set at 0, six
at 2 and one at 3.
Left hippocampal volumes were 2.73 cm3±0.35 (range: 2.23–3.27)
for depressed patients and 2.95 cm3±0.31 (range: 2.45–3.47) for
control participants. Right hippocampal volumes were 2.71 cm3±0.47
(range: 1.38–3.40) for depressed patients and 3.10 cm3±0.33 (range:
2.63–3.73) for control participants. Detailed results are presented in
The ANCOVA with the hippocampal volume (cm3) as the
dependant variable and the TIV, gender and age as covariates revealed
no significant side effect (F=0.842, df=37, p=0.365) or group×side
interaction (Fb0.001, df=37, p=0.992). However a significant group
effect was revealed (F=6.504, df=37, pb0.05). The ANCOVA with the
hippocampal volume (cm3) as the dependant variable and the TCV,
gender and age as covariates revealed similar results [no significant
side effect (F=1.467, df=37, p=0.233), or group×side interaction
(F=0.022, df=37, p=0.884), but a significant group effect (F=4.234,
The mean hippocampal volume reduction in patients compared to
controls was 9.7% (7.8% on the left and 11.6% on the right). T testonTIV
values showed no significant difference between group (t=0.836,
When analysing our datawith the standard normalization of SPM5
(i.e. without DARTEL normalization) VBM analyses were unable to
reveal bilateral hippocampal changes. This could be due to the fact
the posterior subparts of the hippocampus, the hippocampal anterior
part being preserved (Maller et al., 2007) and that this posterior part
needs a good realignment. This may account for the discrepancies
between VBM without DARTEL and with DARTEL. Since the posterior
part of the hippocampus is a thin elongated structure, it might not be
accurate for VBM analyses without DARTEL realignment. To test this
hypothesis, we used the SACHA software and divided each automated
hippocampal labels into head, body and tail. We used ANCOVA with
one two-level-between-subject factor (group: patients, controls), two
within-subject factors [1) subpart: head, body, tail; 2) laterality: left,
right], and sex and TIV as covariates. This revealed a significant group
by subpart effect (p=0.018). We then performed distinct ANCOVAs for
each subpart of the hippocampus to determine which one triggered
this effect. There was no significant effect on the head (p=0.299).
There was a significant group effect on the body (p=0.003) and a
tendency of group effect on the tail (p=0.093). Detailed results are
presented in Table 5.
normalization) as well as manual and automated segmentation of the
hippocampus were able to detect significant bilateral hippocampal
volume reduction in acute depressed patients compared to controls.
There were comparable degrees of hippocampal reduction with
manual segmentation and automated segmentation.
This significant hippocampal volume reduction in acute depressed
patients is consistent with the two meta-analyses of Videbech and
Ravnkilde (2004) and Campbell et al. (2004).
Thus, according to our results the three methods (ROI-based VBM-
DARTEL, manual and automated segmentation) are sensitive enough
to detect hippocampal volume differences in acute depression. As
specified in the introduction, only two studies using VBM (the
standard version) have been published in acute depressed patients.
The study of Bell-McGinty et al. did not find any significant gray
matter differences in the hippocampal region with corrected p values
but found some with an uncorrected p value at p=0.001 (Bell-
McGinty et al., 2002) and Vasic et al. (2008) have recently found
positive results in middle aged acute depressed patients (Vasic et al.,
2008). A study in chronic patients found positive results with VBM
(Shah et al., 1998). Some other works might have been done with
standard-VBM in acute depression, but these may have not been
published due to negative results.
In our study, the results of whole brain analysis with VBM-DARTEL
wereclose tothe results of Vasicetal. (2008). In ourstudy, whenusing
a FDR corrected threshold of pb0.05, there were volume differences
on middle temporal gyrus, parahippocampal gyrus, superior parietal
lobule, and cingulate gyrus.
of anatomical differences between groups across the whole-brain. Its
main advantage is to simultaneously analyse many different brain
regions. With its diffeomorphic image registration algorithm (DARTEL),
it has certainly improved the realignment. Even if it is not yet totally
user-friendly and needs to run during few days to get the template,
DARTEL is a real improvement over the standard approach specially for
the medial temporal lobe (Yassa et al. 2008), including, as we can see
with our results, an improvement for the hippocampus analyses.
Using a subdivision of the automated segmentation into hippo-
campal head, body and tail, we observed that our depressed patients
had no significant volume reduction in the head of the hippocampus
but showed a significant difference in the body and a trend in the tail.
This is consistent with the results of Maller et al. (2007), who found a
more pronounced difference in the posterior part of the hippocampus.
On the contrary, in schizophrenia, atrophy seems to mainly affect the
hippocampal head (Csernansky et al., 2002). Several studies using the
standard version of VBM with SPM have shown hippocampal gray
matter differences when comparing schizophrenic patients with
healthycontrols (Kubicki et al., 2002; Job et al., 2002, 2003; Moorhead
Hippocampal volumes (cm3) using manual segmentation
Left Hc manual segmentationRight Hc manual segmentation
Means.d. MinMax Means.d. MinMax
Hippocampal volumes (cm3) using automated segmentation
Left Hc automated
Right Hc automated
Mean s.d.Min MaxMean s.d. MinMax
L. Bergouignan et al. / NeuroImage 45 (2009) 29–37
et al., 2004; Borgwardt et al., 2007b; Rametti et al., 2007). These
different spatial patterns of hippocampal atrophy between depression
and schizophrenia might have influence VBM standard version's
sensitivity. With the subdivision in head-body-tail with the auto-
mated segmentation, and in the VBM-DARTEL ROI analyses, we can
observe that the difference is larger in the posterior part of the
hippocampus in our population (see Fig. 4).
Manual and automated segmentation resulted in comparable
volume reductions (9.7% vs. 11.6%). However, in both populations,
manual volumes were larger than non TIV corrected automated
volumes (mean=4.20 cm3for manual segmentation, mean=2.87 cm3
for automated segmentation). Considering that the manual and the
automated hippocampus segmentation protocols are highly similar,
this difference is likely to be due to the normalization in the MNI
stereotactic space that was performed prior to manual segmentation
with the linear ICBM 152 template (Lancaster et al., 2007).
Our study has some limitations: we did not perform any direct
comparison between the methods. The automated segmentation
method was done on native scans whereas manual segmentations and
VBM after normalization to a standard space (as specified on the
methodology, with the modulation we also needed to take into
account global differences during statistical analyses with the VBM).
One would expect that the sharper edges of the native scans would be
as beneficial to manual segmentation as they are to automated
segmentation. However, the use of normalization may be beneficial to
the manual segmentation because it improves reproducibility and
Pruessner's manual segmentation has not been adapted for a
segmentation on native MRI images. Whatsoever repeating the
manual segmentation on the native scans to have a direct comparison
of automated and manual segmentations would in turn not allow for
direct comparison to the VBM-DARTEL method that includes spatial
normalization as part of the preprocessing. Beyond this limitation, our
results confirm the hippocampal difference between patients and
controls and highlight the posterior aspect of this difference.
Episodic memory retrieval is critically dependent upon hippo-
campal integrity (Sapolsky et al.,1990). According to the HIPER model
(HIPpocampal Encoding/Retrieval model), activations in the hippo-
campal region associated with episodic memory encoding are located
in the rostral part of the region (i.e. head), whereas activations
associated with episodic memory retrieval are located in the caudal
part (i.e. bodyand tail) (Lepage et al.,1998). Episodic memory retrieval
is impaired in depressed patients not only in acute state (Lemogne
et al., 2006), but also in euthymic state (Bergouignan et al., 2008). The
specific anatomical abnormality found in the posterior hippocampus
in our study in acute depression may be associated with specific
retrieval deficits in depression.
Structural changes in the hippocampus could be due to remodeling
of key cellular elements, involving retraction of dendrites, decreased
neurogenesis in the dentate gyrus, and loss of glial cells (Cameron
et al., 1998; Magarinos et al., 1999; Malberg et al., 2000; McEwen,
1999; Rajkowska, 2000; Rogatsky et al., 1997). According to post
mortem clinical studies and animal model studies, most antidepres-
sants stimulateadult hippocampal neurogenesis(Malberget al.,2000;
Perera et al., 2000; Madsen et al., 2000; Van Praag et al.,1999; Duman,
2004). Hippocampal changes could be associated with acute
depressed patient's responsiveness to antidepressants. Large scale
longitudinal studies evaluating the link between hippocampal
atrophy and medication responsiveness are necessary to test this
To summarize, our results demonstrate that automated segmenta-
tion and VBM with DARTEL can constitute a viable alternative to
manual segmentation to detect hippocampal atrophy in acute
depressed patients. The two automated techniques can be used to
performlarge scalevolumetric studies in humans.The newautomated
segmentation technique could further explore and detect hippocam-
pal subpart differences that could be very useful for clarifying
physiopathology and providing further light on the clinical implica-
tions of these structural brain abnormalities.
We are particularly grateful to the reviewers for their requirement
in the VBM part of the paper that has made us notably improve the
manuscript. We are also grateful to Lisa Buchy and thank her for the
helpful comments and corrections on English language regarding the
L.B. is supported by funding from the Institut National du Cancer
(The Cancer National Institute).
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.neuroimage.2008.11.006.
Ashburner, J., 2007t. A fast diffeomorphic image registration algorithm. NeuroImage 15
38 (1), 95–113 Oct.
Ashburner, J., Friston, K.J., 2000. Voxel-based morphometry—the methods. NeuroImage
Ashburner, J., Friston, K.J., 2005. Unified segmentation. NeuroImage 26, 839–851 2005.
Ballmaier, M., Narr, K.L., Toga, A.W., Elderkin-Thompson, V., Thompson, P.M., Hamilton,
L., Haroon, E., Pham, D., Heinz, A., Kumar, A., 2008. Hippocampal morphology and
distinguishing late-onset from early-onset elderly depression. Am. J. Psychiatry 165
Beck, A.T., Ward, C.H., Mendelson, M., Mock, J., Erbauch, J., 1961. An inventory for
measuring depression. Arch. Gen. Psychiatry 4, 561–571.
Bell-McGinty, S., Butters, M.A., Meltzer, C.C., Greer, P.J., Reynolds III, C.F., Becker, J.T.,
2002. Brain morphometric abnormalities in geriatric depression: long-term
neurobiological effects of illness duration. Am. J. Psychiatry 159 (8), 1424–1427.
Benjamini, Y., Hochberg, Y., 2000. On the adaptive control of the false discovery rate in
multiple testing with independent statistics. J. Educ. Behav. Stat. 25 (1), 60–83.
Bergouignan, L., Lemogne, C., Foucher, A., Longin, E., Vistoli, D., Allilaire, J.F., Fossati,
P., 2008. Field perspective deficit for positive memories characterizes auto-
biographical memory in euthymic depressed patients. Behav. Res. Ther. 46 (3),
Borgwardt, S.J., Riecher-Rössler, A., Dazzan, P., Chitnis, X., Aston, J., Drewe, M.,
Gschwandtner, U., Haller, S., Pflüger, M., Rechsteiner, E., D'Souza, M., Stieglitz,
R.D., Radü, E.W., McGuire, P.K., 2007a. Regional gray matter volume abnormalities
in the at risk mental state. Biol. Psychiatry 61 (10), 1148–1156 15.
Borgwardt, S.J., McGuire, P.K., Aston, J., Berger, G., Dazzan, P., Gschwandtner, U., Pflüger,
M., D'Souza, M., Radue, E.W., Riecher-Rössler, A., 2007b. Structural brain
abnormalities in individuals with an at-risk mental state who later develop
psychosis. Br. J. Psychiatr. (Suppl.) 51, 69–75.
Caetano, S.C., Hatch, J.P., Brambilla, P., Sassi, R.B., Nicoletti, M., Mallinger, A.G., Frank, E.,
Kupfer, D.J., Keshavan, M.S., Soares, J.C., 2004. Anatomical MRI study of
hippocampus and amygdala in patients with current and remitted major
depression. Psychiatry Res. 132 (2), 141–147 15.
Cameron, H.A., Tanapat, P., Gould, E., 1998. Adrenal steroids and N-methyl-D-aspartate
receptor activation regulate neurogenesis in the dentate gyrus of adult rats through
a common pathway. Neuroscience 82 (2), 349–354.
Hippocampal head, body and tail volumes (cm3) using automated segmentation
Left hippocampus Right hippocampus
Head BodyTail HeadBodyTail
Means.d. Means.d.Means.d. Means.d.Means.d. Mean s.d.
L. Bergouignan et al. / NeuroImage 45 (2009) 29–37
Campbell, S., Marriott, M., Nahmias, C., MacQueen, G.M., 2004. Lower hippocampal Download full-text
volumeinpatients suffering fromdepression:a meta-analysis. Am.J. Psychiatry 161
Chupin, M., Mukuna-Bantumbakulu, A.R., Hasboun, D., Bardinet, E., Baillet, S.,
Kinkingnéhun, S., Lemieux, L., Dubois, B., Garnero, L., 2007. Anatomically
constrained region deformation for the automated segmentation of the hippo-
campus and the amygdala: method and validation on controls and patients with
Alzheimer's disease. NeuroImage 34 (3), 996–1019 1.
Colla, M., Kronenberg, G., Deuschle, M., Meichel, K., Hagen, T., Bohrer, M., Heuser, I.,
2007.Hippocampal volumereduction and HPA-system activityin majordepression.
J. Psychiatr. Res. 41 (7), 553–560.
Colliot, O., Chételat, G., Chupin, M., Desgranges, B., Magnin, B., Benali, H., Dubois, B.,
Garnero, L., Eustache, F., Lehéricy, S., 2008. Discrimination between Alzheimer
disease, mild cognitive impairment and normal aging by using automated
segmentation of the hippocampus. Radiology 248 (1), 194–201.
Csernansky, J.G., Wang, L., Jones, D., Rastogi-Cruz, D., Posener, J.A., Heydebrand, G.,
Miller, J.P., Miller, M.I., 2002. Hippocampal deformities in schizophrenia character-
ized by high dimensional brain mapping. Am. J. Psychiatry 159 (12), 2000–2006.
Duman, R.S., 2004. Depression: a case of neuronal life and death? Biol. Psychiatry 1 56
(3), 140–145 Review.
Fischl, B., Salat, D.H., Busa, E., Albert, M., Dieterich, M., Haselgrove, C., van der Kouwe, A.,
Killiany, R., Kennedy, D., Klaveness, S., Montillo, A., Makris, N., Rosen, B., Dale, A.M.,
2002. Whole brain segmentation: automated labeling of neuroanatomical
structures in the human brain. Neuron 33 (3), 341–355 31.
Frodl, T., Meisenzahl, E.M., Zetzsche, T., Born, C., Groll, C., Jäger, M., Leinsinger, G.,
Bottlender, R., Hahn, K., Möller, H.J., 2002. Hippocampal changes in patients with a
first episode of major depression. Am. J. Psychiatry 159 (7), 1112–1118.
Frodl, T., Schaub, A., Banac, S., Charypar, M., Jäger, M., Kümmler, P., Bottlender, R.,
Zetzsche, T., Born, C., Leinsinger, G., Reiser, M., Möller, H.J., Meisenzahl, E.M., 2006.
Reduced hippocampal volume correlates with executive dysfunctioning in major
depression. J. Psychiatry Neurosci. 31 (5), 316–323.
Genovese, C.R., Lazar, N.A., Nichols, T., 2002. Thresholding of statistical maps in
functional neuroimaging using the false discovery rate. 15(4):870–8.
Geuze, E.,Vermetten,E.,Bremner, J.D., 2005.MR-basedinvivohippocampalvolumetrics:
review of methodologies currently employed. Mol. Psychiatry 10 (2),147–159.
Hastings, R.S., Parsey, R.V., Oquendo, M.A., Arango, V., Mann, J.J., 2004. Volumetric
analysis of the prefrontal cortex, amygdala, and hippocampus in major depression.
Neuropsychopharmacology 29 (5), 952–959.
Job, D.E., Whalley, H.C., McConnell, S., Glabus, M., Johnstone, E.C., Lawrie, S.M., 2002.
Structural gray matter differences between first-episode schizophrenics and
normal controls using voxel-based morphometry. NeuroImage 17 (2), 880–889.
Job, D.E., Whalley, H.C., McConnell, S., Glabus, M., Johnstone, E.C., Lawrie, S.M., 2003.
Voxel-based morphometry of gray matter densities in subjects at high risk of
schizophrenia. Schizophr. Res. 64 (1), 1–13 1.
Kubicki, M., Shenton, M.E., Salisbury, D.F., Hirayasu, Y., Kasai, K., Kikinis, R., Jolesz, F.A.,
McCarley, R.W., 2002. Voxel-based morphometric analysis of gray matter in first
episode schizophrenia. NeuroImage 17 (4), 1711–1719.
Lancaster, J.L., Tordesillas-Gutiérrez, D., Martinez, M., Salinas, F., Evans, A., Zilles, K.,
Mazziotta, J.C., Fox, P.T., 2007. Bias between MNI and Talairach coordinates analyzed
using the ICBM-152 brain template. Hum. Brain Mapp. 28 (11), 1194–1205.
Lemogne, C., Piolino, P., Friszer, S., Claret, A., Girault, N., Jouvent, R., Allilaire, J.F., Fossati,
P., 2006. Episodic autobiographical memory in depression: specificity, autonoetic
consciousness, and self-perspective. Conscious. Cogn. 15 (2), 258–268.
Lepage, M., Habib, R., Tulving, E., 1998. Hippocampal PET activations of memory
encoding and retrieval: the HIPER model. Hippocampus 8 (4), 313–322.
Madsen, T.M., Greisen, M.H., Nielsen, S.M., Bolwig, T.G., Mikkelsen, J.D., 2000.
Electroconvulsive stimuli enhance both neuropeptide Y receptor Y1 and Y2
messenger RNA expression and levels of binding in the rat hippocampus.
Neuroscience 98 (1), 33–39.
Magariños, A.M., Deslandes, A., McEwen, B.S., 1999. Effects of antidepressants and
benzodiazepine treatments on the dendritic structure of CA3 pyramidal neurons
after chronic stress. Eur. J. Pharmacol. 371 (2–3), 113–122 29.
Malberg, J.E., Eisch, A.J., Nestler, E.J., Duman, R.S., 2000. Chronic antidepressant
treatment increases neurogenesis in adult rat hippocampus. J. Neurosci. 20 (24),
Maller, J.J., Daskalakis, Z.J., Fitzgerald, P.B., 2007. Hippocampal volumetrics in
depression: the importance of the posterior tail. Hippocampus 17 (11), 1023–1027.
Martino, D., Draganski, B., Cavanna, A., Church, A., Defazio, G., Robertson, M.M.,
Frackowiak, R.S., Giovannoni, G., Critchley, H.D., 2008. Anti-basal ganglia antibodies
and Tourette's syndrome: a voxel-based morphometry and diffusion tensor
imaging study in an adult population. J. Neurol. Neurosurg. Psychiatry 79 (7),
McEwen, B.S., Tanapat, P., Weiland, N.G.,1999. Inhibition of dendritic spine induction on
hippocampal CA1 pyramidal neurons by a nonsteroidal estrogen antagonist in
female rats. Endocrinology 140 (3), 1044–1047.
Mervaala, E., Föhr, J., Könönen, M., Valkonen-Korhonen, M., Vainio, P., Partanen, K.,
Partanen, J., Tiihonen, J., Viinamäki, H., Karjalainen, A.K., Lehtonen, J., 2000.
Quantitative MRI of the hippocampus and amygdala in severe depression. Psychol.
Med. 30 (1), 117–125.
Montgomery, S.A., Asberg, M.,1979. A new depression scale designed to be sensitive to
change. Br. J. Psychiatry 134, 382–389.
Moorhead, T.W., Job, D.E., Whalley, H.C., Sanderson, T.L., Johnstone, E.C., Lawrie, S.M.,
2004. Voxel-based morphometry of comorbid schizophrenia and learning
disability: analyses in normalized and native spaces using parametric and
nonparametric statistical methods. NeuroImage 22 (1), 188–202.
Mugler III, J.P., Brookeman, J.R.,1991. Rapid three-dimensional T1-weighted MR imaging
with the MP-RAGE sequence. J. Magn. Reson. Imaging 1 (5), 561–567.
Nieuwenhuys, R., Voogd, J., van Huijzen, C.,1988. The Human Central Nervous System: a
Synopsis and Atlas3rd revised ed. Springer-Verlag, Berlin.
Perera, T.D., Lisanby, S.H., 2000. Neurogenesis and depression. J. Psychiatr. Pract. 6 (6),
Posener, J.A., Wang, L., Price, J.L., Gado, M.H., Province, M.A., Miller, M.I., Babb, C.M.,
Csernansky, J.G., 2003. High-dimensional mapping of the hippocampus in
depression. Am. J. Psychiatry 160 (1), 83–89.
Pruessner, J.C., Li, L.M., Serles, W., Pruessner, M., Collins, D.L., Kabani, N., Lupien, S.,
Evans, A.C., 2000. Volumetry of hippocampus and amygdala with high-resolution
MRI and three-dimensional analysis software: minimizing the discrepancies
between laboratories. Cereb. Cortex 10 (4), 433–442.
Rajkowska, G., 2000. Postmortem studies in mood disorders indicate altered numbers
of neurons and glial cells. Biol. Psychiatry 48 (8), 766–777 15.
Rametti, G., Segarra, N., Junqué, C., Bargalló, N., Caldú, X., Ibarretxe, N., Bernardo, M.,
2007. Left posterior hippocampal density reduction using VBM and stereological
MRI procedures in schizophrenia. Schizophr. Res. 96 (1–3), 62–71.
Rogatsky, I., Trowbridge, J.M., Garabedian, M.J., 1997. Glucocorticoid receptor-mediated
cell cycle arrest is achieved through distinct cell-specific transcriptional regulatory
mechanisms. Mol. Cell. Biol. 17 (6), 3181–3193.
Rusch, B.D., Abercrombie, H.C., Oakes, T.R., Schaefer, S.M., Davidson, R.J., 2001.
Hippocampal morphometry in depressed patients and control subjects: relations
to anxiety symptoms. Biol. Psychiatry 50 (12), 960–964 15.
Sapolsky, R.M., Uno, H., Rebert, C.S., Finch, C.E., 1990. Hippocampal damage associated
with prolonged glucocorticoid exposure in primates. J. Neurosci.10 (9), 2897–2902.
Saylam, C., Uçerler, H., Kitiş, O., Ozand, E., Gönül, A.S., 2006. Reduced hippocampal
volume in drug-free depressed patients. Surg. Radiol. Anat. 28 (1), 82–87.
Seidman, L.J., Faraone, S.V., Goldstein, J.M., Goodman, J.M., Kremen, W.S., Toomey, R.,
Tourville, J., Kennedy, D., Makris, N., Caviness, V.S., Tsuang, M.T.,1999. Thalamic and
amygdala-hippocampal volume reductions in first-degree relatives of patients with
schizophrenia: an MRI-based morphometric analysis. Biol. Psychiatry 46 (7),
Shah, P.J., Ebmeier, K.P., Glabus, M.F., Goodwin, G.M., 1998. Cortical gray matter
reductions associated with treatment-resistant chronic unipolar depression.
Controlled magnetic resonance imaging study. Br. J. Psychiatry 172, 527–532.
Sheehan, D.V., Lecrubier, Y., Sheehan, K.H., Amorim, P., Janavs, J., Weiller, E., Hergueta, T.,
Baker, R., Dunbar, G.C.,1998. The Mini-International Neuropsychiatric Interview (M.I.
N.I.): the development and validation of a structured diagnostic psychiatric interview
for DSM-IV and ICD-10. J. Clin. Psychiatry 59 (Suppl. 20), 22–33 quiz 34–57.
Stonnington, C.M., Tan, G., Klöppel, S., Chu, C., Draganski, B., Jack Jr, C.R., Chen, K.,
Ashburner, J., Frackowiak, R.S., 2008. Interpreting scan data acquired from multiple
scanners: a study with Alzheimer's disease. NeuroImage 39 (3), 1180–1185 Feb 1.
Vakili, K., Pillay, S.S., Lafer, B., Fava, M., Renshaw, P.F., Bonello-Cintron, C.M., Yurgelun-
Todd, D.A., 2000. Hippocampal volume in primary unipolar major depression: a
magnetic resonance imaging study. Biol. Psychiatry 47 (12), 1087–1090 15.
Van Praag, H., Kempermann, G., Gage, F.H., 1999. Running increases cell proliferation
and neurogenesis in the adult mouse dentate gyrus. Nat. Neurosci. 2 (3), 266–270.
Vasic, N., Walter, H., Höse, A., Wolf, R.C., 2008. Gray matter reduction associated with
psychopathology and cognitive dysfunction in unipolar depression: a voxel-based
morphometry study. J. Affect. Disord. 109 (1–2), 107–116.
Videbech, P., Ravnkilde, B., 2004. Hippocampal volume and depression: a meta-analysis
of MRI studies. Am. J. Psychiatry 161 (11), 1957–1966.
Von Gunten, A., Fox, N.C., Cipolotti, L., Ron, M.A., 2000. A volumetric study of
hippocampus and amygdala in depressed patients with subjective memory
problems. J. Neuropsychiatry Clin. Neurosci. 12 (4), 493–498.
Yassa, M.A., Stark, C.E., 2009. A quantitative evaluation of cross-participant registration
techniques for MRI studies of the medial temporal lobe. NeuroImage 44 (2),
L. Bergouignan et al. / NeuroImage 45 (2009) 29–37