Annu. Rev. Med. 1998. 49:341–61Copyright © 1998 by Annual Reviews Inc. All rights reserved
FUNCTIONAL NEUROIMAGING
STUDIES OF DEPRESSION: The
Anatomy of Melancholia
Wayne C. Drevets, MDDepartments of Psychiatry and Radiology, University of Pittsburgh Medical Center,Pittsburgh, Pennsylvania 15213
KEY WORDS: PET (positron emission tomography), cerebral glucose metabolism, cerebral
blood flow, MRI (magnetic resonance imaging), major depression
ABSTRACT
Functional brain imaging techniques, which permit noninvasive measures ofneurophysiology and neuroreceptor binding, are powerful and sensitive toolsfor research aimed at elucidating the pathophysiology of major depression.The application of these technologies in depression research has producedseveral studies of resting cerebral blood flow (BF) and glucose metabolismin subjects imaged during various phases of illness and treatment. This re-view examines these data and the principles relevant to their interpretationand discusses the insights they provide into the anatomical correlates of de-pression. Within the anatomical networks implicated in emotional process-ing by other types of evidence, these BF and metabolic data demonstrate thatmajor depression is associated with reversible, mood state–dependent, neu-rophysiological abnormalities in some structures and irreversible, trait-likeabnormalities in other structures. In some of the regions in which trait-likeabnormalities appear, abnormal metabolic activity appears at least partly re-lated to the anatomical abnormalities identified in magnetic resonance imag-ing (MRI) studies of depression.
APPLICATION OF NEUROIMAGING IN DEPRESSIONRESEARCH
The major depressive disorders appear particularly tractable to functional im-
aging approaches aimed at elucidating their pathophysiology because they
0066-4219/98/0201-0341$08.00
341
have been associated with disruptions of brain function in the absence of grossneuropathology (1). Their episodic nature and responsiveness to treatmentpermit imaging in both symptomatic and asymptomatic states. The similaritybetween some depressive symptoms and emotional states that can be ex-pressed by nondepressed subjects permits comparisons of the hemodynamicchanges associated with the depressed state with those occurring in healthysubjects during experimentally induced sadness or anxiety (2). The insightsprovided by functional neuroimaging studies of depression may ultimately lo-calize specific brain regions for histopathological assessment, elucidate anti-depressant treatment mechanisms, and guide pathophysiology-based classifi-cation of depression.
The clinical capabilities of these tools for determining diagnoses or guidingtreatment decisions have not been established (3). The abnormalities identifiedin BF and metabolic imaging studies of depression have consisted of differ-ences in mean tracer-uptake values between groups of depressives and con-trols. The sensitivity for detecting abnormalities in individual cases has beenlimited by the small magnitude of such differences relative to the variability ofimaging measures. None of these abnormalities has been shown to sensitivelyand specifically distinguish individual depressives from healthy subjects andfrom subjects who have other illnesses.
SIGNIFICANCE OF REGIONAL BLOOD FLOW ANDGLUCOSE METABOLISM
Changes in local BF and glucose metabolism during mental activity predomi-nantly reflect net differences in the summed metabolic activity associated withterminal field synaptic transmission (4, 5). Dynamic brain imaging can thus beused to provide maps of regional neural function (4). However, these measuresare also affected by cerebrovascular disease, thyroid dysfunction, gray-matterreductions associated with atrophy or hypoplasia, and other abnormalities thatexist in some depressive subgroups (6, 7). Regional BF or metabolic differ-ences between depressives and controls may thus reflect either the physiologi-cal correlates of depressive emotions and thoughts or pathophysiologicalchanges that predispose subjects to or result from affective disease.
The physiological correlates of depressive symptoms presumably appear as
reversible changes in local BF and metabolism that normalize after effective
treatment. These changes can to some extent be reproduced in healthy subjects
imaged while performing tasks that mimic the corresponding depressive mani-
festation. For example, in primary major depressive disorder (MDD), the me-
dial orbital cortex shows abnormally elevated BF and metabolism in the de-
pressed phase relative to the remitted phase (Figure 1a, see color section, p. C-
342 DREVETS
8). Medial orbital BF also increases during experimentally induced states of
anxiety or sadness in healthy subjects (2, 8, 9, 10).In contrast, BF and metabolic abnormalities in depression may also reflect
pathophysiological changes in synaptic transmission associated with altered
neurotransmitter synthesis or receptor sensitivity, or with alterations in the
number of synaptic connections resulting from regional atrophy or hypoplasia.
Such abnormalities may be evident regardless of whether subjects are exhibit-
ing symptoms of depression (14, 15).
TECHNICAL ISSUES RELEVANT TO THEINTERPRETATION OF IMAGE DATA
Consensus regarding the functional anatomical correlates of depression hasbeen difficult to achieve because the conclusions disagree across studies.However, many of the conflicts within the literature can be resolved by consid-ering technical issues related to subject selection, image acquisition, and im-age analysis (3).
Subject Selection
Most functional imaging studies of depression involve relatively small samplesizes. This design limitation coupled with the subtle magnitude of the BF andmetabolic differences between depressives and controls relative to the vari-ability of such measures reduces the sensitivity of studies for detecting inter-group differences and for replicating findings across studies. Consequently,subject selection criteria that reduce the variability of imaging measures areoften required to improve statistical sensitivity.
Medication effects are an important source of variability in functional im-
aging studies. Regional BF and metabolism in some prefrontal cortical and
limbic areas of interest in depression are reduced by antidepressant, antipsy-
chotic, and antianxiety drugs (3, 16, 17). Blood flow or metabolic image data
acquired from depressives medicated with these drugs are thus difficult to in-
terpret without the availability of unmedicated baseline scans for comparison.
Nevertheless, most published studies of depression report image data con-
founded by such medication effects, potentially introducing artifactual differ-
ences or obscuring true biological differences between depressives and con-
trols.Another source of variability is introduced by the clinical heterogeneity in-
herent within the depressive syndrome, as diverse signs and symptoms may
have distinct neurophysiological correlates (1). For example, a depressed pa-
tient exhibiting prominent anxiety, obsessive ruminations, insomnia, and psy-
NEUROIMAGING STUDIES OF DEPRESSION 343
chomotor agitation may show dissimilar imaging findings from one who pre-
dominantly manifests apathy, inactivity, excessive sleep, and psychomotor
slowing. Nevertheless, few studies have had sufficiently large subsamples of
depressives manifesting clinical factors to sort out the relative contributions of
each factor to the variability of imaging measures.A related concern for imaging studies of depression is that the major de-
pressive syndrome is heterogenous with regard to etiology and pathology (1).
Biological heterogeneity is evidenced by the variety of antecedents to depres-
sive syndromes; the diversity of responses to somatic or psychological thera-
pies; and the variable presence of neuroendocrine, neurochemical, and cir-
cadian rhythm disturbances in depressive samples (1). If depression is associ-
ated with multiple pathophysiological states, it can presumably be character-
ized by an assortment of distinct functional imaging abnormalities. The litera-
ture confirms this hypothesis, since some subtyping of depressed subjects ap-
pears critical for reducing the variability of image data. Examples discussed
below include the distinction between primary and secondary depressive dis-
orders and between elderly depressives with a late versus an early age-of-
illness-onset.
Implications of Structural Abnormalities for FunctionalNeuroimaging
Presumably related to this biological diversity, some subsets of depressed sub-jects have neuromorphological abnormalities based on structural MRI or CTimaging studies (3). Depressives who are bipolar, psychotic, or elderly (with alate age-of-depression-onset) have an elevated incidence of ventricular andsulcal enlargement and have reductions in some lobar or gyral volumes (3).The tissue reductions reflected by such structural imaging abnormalities de-crease the magnitude of imaging measures from positron emission tomogra-phy (PET) and single photon emission tomography (SPET) images because ofthe low spatial resolution of these images (6). Where atrophy or hypoplasia ex-ists, the tracer uptake measured within the corresponding PET or SPET imagevoxel (volume element) reflects lower proportions of gray and white matterrelative to the proportion of cerebrospinal fluid (CSF). The measured BF ormetabolism is thus reduced via partial volume averaging effects of emphasiz-ing CSF, which is metabolically inactive (6).
As a result, findings of decreased BF and metabolism in depressives who
exhibit such abnormalities are difficult to interpret. Since the incidence of ven-
tricular and sulcal enlargement does not appear increased in nondelusional,
unipolar depressives who developed the disorder at a young age, the image
data from this group must constitute the benchmark against which physiologi-
344 DREVETS
cal imaging measures in other depressive subtypes are compared (1, 3). Nev-
ertheless, some focal areas of reduced gray matter volume appear to exist even
in young, nondelusional, familial unipolar depressives. Such findings empha-
size the importance of evaluating areas in which BF and metabolism are irre-
versibly decreased in depressives relative to controls using MRI-based mor-
phometric and postmortem histopathological studies (14; see also Figure 2, p.
C-10).Another type of structural abnormality that typically shows up in MRI im-
ages from elderly depressives who have a late age-of-depression-onset in-
volves white matter hyperintensities (WMH) and lacunar infarctions (18).
Postmortem evidence indicates that these hyperintensities reflect arterioscle-
rotic or ischemic disease (18–20). As expected, imaging studies demonstrate
that elderly depressives who have WMH have decreased BF and metabolism
in the involved areas relative to age-matched depressives who do not have
WMH (21). In the presence of cerebrovascular disease, the findings of reduced
BF and metabolism cannot be interpreted as providing information related to
local synaptic transmission. Nevertheless, most functional imaging studies of
depression in the elderly have not excluded (or even identified) subjects who
show MRI evidence of cerebrovascular disease.
Image Acquisition and Analysis Issues
Functional imaging studies of depression have been performed using PET,SPET, and nontomographic multidetector systems, which each present theirown set of unique technical considerations (3, 4). The precision and resolutionof localizing functional imaging measures has been enhanced by progressiveimprovements in spatial resolution [to 4 mm for the newest PET cameras and0.5 mm for functional MRI (fMRI)], techniques for co-registering PET andMRI images, and statistical mapping methods that delimit inherent physiologi-cal differences between depressives and controls (3; see also Figures 1a–c and2, pp. C-8, C-9, C-10). Localization is now limited as much by the anatomicalvariability across individuals as by the spatial resolution of imaging technolo-gies. Nevertheless, most published studies were performed using techniquesthat more severely constrain the size and the location of brain structures fromwhich measures can be obtained (3, 4). For example, images acquired usingSPET or nontomographic multidetector systems and inhalation of 133Xe pro-vide BF measures limited to the cortical gray matter lying near the scalp (4).Moreover, although PET affords relatively high spatial resolution and sensi-tivity for deep structures, in most PET studies, the images have been blurred(by filtering) to spatial resolutions of 2 cm or more prior to analysis. This prac-tice reduces the effects of anatomical variability across subjects, but it also sac-rifices the method's ability for resolving small structures. Such considerations
NEUROIMAGING STUDIES OF DEPRESSION 345
thus influence the size of regions that can be resolved in functional imagingstudies, the boundaries for defining regions of interest, and the sensitivity forreplicating results across studies.
Another technical issue applied variably across imaging studies involvesthe normalization of regional measures by dividing by global or hemisphericmeasures (4). Because of the small subject sample sizes involved in imagingstudies, normalization is usually required to reduce the variability of regionalvalues so that intergroup differences can be detected. Whole-brain BF and glu-cose metabolism have not significantly differed between depressed and con-trol samples in most studies of unmedicated, mid-life depressives (e.g. 14, 16,22). In contrast, global BF and metabolism correlated inversely with thyroidstimulating hormone (TSH) levels and were decreased in a depressed samplewhose mean TSH value exceeded that of the control sample (7). Since modestelevations of TSH are common in major depression, these data suggest that ab-solute BF and metabolic measures in depressive samples may be difficult to in-terpret unless covaried for TSH levels. This problem further emphasizes thebenefits of normalizing regional values to exclude global effects, which wouldminimize the global impact of differences in mean TSH levels between depres-sives and controls on regional BF and metabolic values (7).
ABNORMALITIES OF CEREBRAL BLOOD FLOW ANDMETABOLISM IN DEPRESSION
Studies comparing depressed patients to healthy controls have identified BFand metabolic differences between groups in multiple regions, consistent withthe expectation that the emotional, cognitive, psychomotor, neurovegetative,neuroendocrine, and neurochemical disturbances associated with depressionimplicate extended anatomical networks involving numerous brain structures.This review highlights the major findings delineated by well-designed studiesand replicated across neuroimaging laboratories. The results of studies that areconfounded by medication effects (which occupy a large proportion of the lit-erature) are not included.
Prefrontal Cortical Abnormalities in Mood Disorders
The prefrontal cortex (PFC) represents 40–50% of the volume of the human
brain. Numerous anatomical and functional subdivisions of the PFC have been
based on anatomical connectivity, cytoarchitectonic distinctions, electro-
physiological studies, lesion analyses, and PET/fMRI brain mapping studies
(12, 24). Within the PFC, performance of either cognitive or emotional proc-
essing tasks generally results in BF increases in multiple, distinct areas (in pat-
terns specific to the mental operations demanded by each type of task) and in
346 DREVETS
focal BF decreases in additional prefrontal areas (4, 25, 26). The latter puta-
tively reflect regions that are deactivated to facilitate task performance via
suppression of unattended, potentially competing background processes (2,
27). The complex activation/deactivation patterns formed by these increases
and decreases in physiological activity are interpreted within the context of
data from other brain mapping and lesion-analysis studies.Many studies of the dorsolateral and the dorsomedial PFC found that BF
and metabolism are decreased in depressives relative to controls (Table 1).These abnormalities have been reversible with effective antidepressant ther-apy in some studies (28, 29; but see 30, 31).
Another prefrontal area in which BF and metabolism are decreased in uni-polar and bipolar depressives is in the anterior cingulate gyrus ventral to thegenu of the corpus callosum (14; see also Figure 2). This decrement appears tobe at least partly explained by a corresponding reduction in cortex, as MRI-based neuromorphometric measures demonstrate an abnormal reduction in themean gray matter volume of the left subgenual PFC in familial bipolar and uni-polar depressed samples (6, 14). Consistent with the hypothesis that an ana-tomical abnormality partly accounts for the metabolic reduction in the subge-nual PFC, metabolism in this region does not normalize following treatment(14).
Neurophysiological activity generally increases abnormally in the ventro-lateral, lateral orbital, and posteromedial orbital portions of the PFC (Figure1a,b, p. C-8) and in the portion of the anterior cingulate gyrus anterior to thegenu of the corpus callosum (i.e. the pregenual anterior cingulate) in unmedi-cated, primary, unipolar depressives (Table 1). When depressives are imagedboth before and during effective antidepressant treatment, ventrolateral pre-frontal and orbital cortex activity decrease in the remitted phase relative to thedepressed phase (Table 2). In contrast, the effects of treatment upon activityin the pregenual anterior cingulate have been more variable: Two SPET stud-ies have reported that anterior cingulate BF increases rather than decreasesfollowing treatment (29, 46; see Table 2).
Abnormalities in Related Subcortical and Limbic Structures
The ventrolateral, ventromedial (i.e. anterior cingulate), and orbital areas of
the PFC where BF and metabolism are abnormal in the depressed phase of
MDD all share extensive interconnections with the amygdala, the mediodorsal
nucleus of the thalamus, and the ventral striatum (the ventromedial caudate
and the nucleus accumbens), structures consistently implicated by other types
of evidence in emotional behavior (12, 50). In the left (and possibly the right)
amygdala and the medial thalamus, BF and metabolism are abnormally in-
creased in unipolar and bipolar depression (37, 38, 41, 42, 51, 52; see also Fig-
NEUROIMAGING STUDIES OF DEPRESSION 347
348 DREVETS
Ta
ble
1F
un
ctio
nal
imag
ing
resu
lts
inp
refr
on
tal
cort
ex(P
FC
)fo
ru
nm
edic
ated
,p
rim
ary
,u
nip
ola
rd
epre
ssiv
esa
Sam
ple
Siz
eB
loo
dfl
ow
or
glu
cose
met
abo
lism
inD
EP
rela
tiv
eto
CO
N
Ref
eren
ces
DE
PC
ON
Imag
ing
Tec
hn
iqu
eV
entr
alP
FC
Do
rsal
PF
C
Bax
ter
etal
(32
)1
41
4P
ET
,1
8F
DG
↑m
edia
lo
rbit
al(B
L)b
N/A
Bax
ter
etal
(28
)1
01
2P
ET
,1
8F
DG
N/A
↓d
ors
alan
tero
late
ral
(BL
)
Ben
chet
al(3
3)
33
c2
3P
ET
,H
21
5O
N/A
d↓
do
rso
late
ral(
L);
↓d
ors
alan
t.ci
ng
ula
te
Biv
eret
al(3
4)
12
12
PE
T,
18F
DG
↑m
edia
lan
dla
tera
lo
rbit
al(B
L)
↓d
ors
ola
tera
l(L
)
Bu
chsb
aum
etal
(35
)4
24
PE
T,
18F
DG
↑fr
on
tal/
occ
ipit
alra
tio
(↑v
entr
alfr
on
tal
met
abo
lism
sho
wn
inA
NO
VA
imag
e)N
/A
Co
hen
etal
(36
)7
e3
8P
ET
,1
8F
DG
↑m
edia
lo
rbit
al↓
do
rsal
med
ial
Dre
vet
set
al(3
7)
13
33
PE
T,
H2
15O
↑v
entr
ola
tera
l,la
tera
lo
rbit
al(L
),↑
pre
gen
ual
ant.
cin
gu
late
NS
Dre
vet
set
al(3
8)
31
17
PE
T,
18F
DG
↑la
tera
l&
med
ial
orb
ital
(BL
)f ;↓
sub
gen
ual
ant.
cin
gu
late
f↓
do
rsal
ante
rola
tera
l(L
)f
Eb
ert
etal
(39
)1
08
SP
ET
,9
9m
TcH
MP
AO
↑o
rbit
al(B
L)g
↓d
ors
alan
tero
late
ral
(L)
Mae
set
al(1
7)
30
h1
2S
PE
T,
99
mT
cHM
PA
ON
/AN
S
May
ber
get
al(4
0)
18
i1
5P
ET
,1
8F
DG
↑p
reg
enu
alan
t.ci
ng
ula
tein
trea
tmen
tre
spo
nd
ers,
bu
t↓
inn
on
resp
on
der
s;N
So
rbit
ali
↓d
ors
ola
tera
l(B
L)
Men
tis
etal
(41
)1
11
2P
ET
,1
8F
DG
↑o
rbit
alan
dla
tera
lp
refr
on
tal
NS
No
fzin
ger
etal
(42
)6
10
PE
T,
18F
DG
↑m
edia
lo
rbit
al(L
)↑
late
ral
orb
ital
(R)
↓d
ors
ola
tera
l(L
)
Sil
fver
skiö
ld&
Ris
ber
g(1
6)
31
31
Mu
ltid
etec
tor
pro
bes
,1
33X
ejN
S(t
ren
dto
war
d↑
left
ven
tro
late
ral)
NS
Tri
ved
iet
al(4
3)
24
24
SP
ET
,9
9m
TcH
MP
AO
↑in
feri
or
fro
nta
lg
.(R
)↓
sup
erio
ran
dm
idd
lefr
on
tal
g.
(BL
)
Uy
tden
ho
efet
al(4
4)
16
20
Mu
ltid
etec
tor
pro
bes
,1
33X
ej↑
left
“fro
nta
lra
tio
”(w
hic
hin
clu
ded
ven
tro
late
ral)
↑su
per
ior
fro
nta
lg
.(L
)
Wu
etal
(45
)1
51
5P
ET
,1
8F
DG
↑an
t.ci
ng
ula
teg,
orb
ital
N/A
N/A
NEUROIMAGING STUDIES OF DEPRESSION 349
a Ab
bre
via
tio
ns:
DE
P,
dep
ress
edsu
bje
cts;
CO
N,
con
tro
lsu
bje
cts;
18F
DG
,1
8F
-flu
oro
deo
xy
glu
cose
isu
sed
tom
easu
reg
luco
sem
etab
oli
sm;
H2
15O
(ox
yg
en-1
5w
ater
),9
9m
Tc-
HM
PA
O(t
ech
net
ium
-99
HM
PA
O),
and
13
3X
e(x
eno
n-1
33
)ar
eu
sed
tom
easu
reb
loo
dfl
ow
;an
t,an
teri
or;
g,g
yru
s;B
L,b
i-la
tera
l;L
,le
ft;R
,ri
gh
t;N
S,d
iffe
ren
ceas
sess
edan
dn
otsi
gn
ific
ant;
N/A
,re
gio
nn
otas
sess
ed.↑
and
↓in
dic
ate
incr
ease
san
dd
ecre
ases
,re
spec
tiv
ely
,in
the
dep
ress
ives
rela
tiv
eto
the
con
tro
ls.Im
age
dat
afr
om
stu
die
sw
hic
har
eu
nin
terp
reta
ble
du
eto
con
fou
nd
ing
med
icat
ion
effe
cts
are
no
tre
vie
wed
un
less
dat
aw
ere
sep
arat
ely
asse
ssed
for
the
un
med
icat
edsu
bsa
mp
le.
Dat
afr
om
stu
die
so
f“s
eco
nd
ary
”d
epre
ssio
nan
db
ipo
lar
dep
ress
ion
are
ad-
dre
ssed
sep
arat
ely
(see
tex
tan
dT
able
3).
So
me
of
thes
est
ud
ies
incl
ud
edb
ipo
lar
sub
ject
sin
thei
rd
epre
ssed
sam
ple
s,b
uto
nly
the
resu
lts
fro
mth
eu
ni-
po
lar
sub
sam
ple
are
rep
ort
edh
ere
(un
less
dat
aw
ere
no
tp
rese
nte
dse
par
atel
yfo
rb
ipo
lar
and
un
ipo
lar
sub
ject
s).T
he
des
crip
tiv
ete
rms
use
dto
loca
tere
gio
ns
var
yac
ross
stu
dy
,an
din
som
eca
ses
dis
tin
ctte
rms
may
be
use
dto
des
crib
eth
esa
me
area
.Co
nv
erse
ly,s
tud
ies
usi
ng
the
sam
ete
rmm
ayb
ed
e-sc
rib
ing
dif
fere
nt
cort
ical
area
s(e
.g.
the
“do
rso
late
ral
pre
fro
nta
lco
rtex
”sp
ans
sev
eral
cm2
and
enco
mp
asse
sn
um
ero
us
reso
luti
on
elem
ents
)si
nce
mo
stst
ud
ies
hav
en
ot
use
dlo
cali
zati
on
tech
niq
ues
that
emp
loy
ast
and
ard
ized
coo
rdin
ate
syst
emo
rM
RI-
bas
edla
nd
mar
ks.
bT
his
pap
erco
mp
ared
reg
ion
alm
etab
oli
smb
etw
een
sub
ject
sw
ith
ob
sess
ive-
com
pu
lsiv
ed
iso
rder
and
sub
ject
sw
ho
wer
eei
ther
pri
mar
yu
nip
ola
rd
e-p
ress
ives
or
hea
lth
yco
ntr
ols
.W
hil
eth
ed
iffe
ren
ceb
etw
een
dep
ress
ives
and
con
tro
lsw
asn
ot
stat
isti
call
yas
sess
edin
this
pap
er,th
ep
ub
lish
edv
alu
esfo
rth
eo
rbit
al-t
o-h
emis
ph
ere
rati
oin
the
dep
ress
ives
(N=
14
;ri
gh
t:1
.17
±0
.06
,le
ft:
1.1
4±
0.0
5)
and
the
con
tro
ls(N
=1
3;
rig
ht
1.1
1±
0.0
8,
left
:1
.09
±0
.06
)sh
ow
edd
iffe
ren
ces
wh
ich
wer
esi
mil
arin
mag
nit
ud
ean
dv
aria
bil
ity
toth
ed
iffe
ren
ces
fou
nd
ino
ther
stu
die
su
sin
gim
ages
of
sim
ilar
reso
-lu
tio
n,
and
wo
uld
be
sig
nif
ican
tb
yt-
test
(rig
ht:
P<0
.05
,le
ft:
P<0
.02
)[f
rom
Tab
le3
of
Bax
ter
etal
(27
)].
c Nin
etee
no
fth
ed
epre
ssiv
esw
ere
imag
edw
hil
ere
ceiv
ing
var
iou
sm
edic
atio
ns.
Th
ere
po
rted
abn
orm
alit
ies
wer
ein
itia
lly
iden
tifi
edu
sin
gim
age
dat
afr
om
the
enti
red
epre
ssed
sam
ple
(N=
33
),b
ut
the
dif
fere
nce
bet
wee
nD
EP
and
CO
Nw
asal
soco
nfi
rmed
inth
eu
nm
edic
ated
sub
sam
ple
(N=
14
)p
ost
ho
c.dT
his
stu
dy
iden
tifi
edab
no
rmal
itie
su
sin
ga
stat
isti
cal
imag
eth
atex
clu
ded
pix
els
inth
eo
rbit
alco
rtex
,b
ecau
seth
eim
ages
acq
uir
edd
idn
ot
exte
nd
into
this
ven
tral
stru
ctu
refo
ral
lsu
bje
cts
(Do
lan
,p
erso
nal
com
mu
nic
atio
n).
e Su
bje
cts
also
met
crit
eria
for
win
ter
seas
on
alaf
fect
ive
dis
ord
er.
f Ab
no
rmal
itie
ssi
gn
ific
ant
insu
bg
rou
pm
eeti
ng
crit
eria
for
fam
ilia
lp
ure
dep
ress
ive
dis
ease
,b
ut
no
tfo
rsu
bg
rou
pw
ith
fam
ilia
llo
adin
gfo
ral
coh
ol-
ism
.gA
bn
orm
alit
yfo
un
do
nly
insu
bg
rou
pw
ho
pro
ved
resp
on
siv
eto
slee
pd
epri
vat
ion
.hR
efle
cts
the
sam
ple
size
for
the
un
med
icat
edsu
bje
cts
on
ly,
wh
ow
ere
ind
epen
den
tly
com
par
edto
con
tro
lsan
dto
ano
ther
13
ben
zod
iaze
pin
e-tr
eate
dsu
bje
cts.
i Fiv
eo
fth
ese
sub
ject
sh
adb
een
sub
chro
nic
ally
trea
ted
wit
han
tid
epre
ssan
td
rug
sp
rio
rto
scan
nin
g.It
isu
ncl
ear
wh
eth
ersu
bch
ron
ican
tid
epre
ssan
td
rug
ther
apy
has
the
sam
eef
fect
of
dec
reas
ing
orb
ital
met
abo
lism
asd
oes
chro
nic
trea
tmen
t(T
able
2).
j Blo
od
flo
wis
on
lym
easu
red
nea
rth
esc
alp
usi
ng
this
tech
niq
ue,
son
ore
sult
sar
eav
aila
ble
for
the
orb
ital
or
the
cin
gu
late
cort
ices
(see
tex
t).
ure 1a,c, pp. C-8, C-9). In contrast, in unipolar—but not bipolar—depressives,
BF and metabolism are reduced abnormally in the caudate (37, 22, 53).
Abnormalities in Other Brain Areas
Regional BF and metabolic abnormalities in other brain areas have been repli-
cated less consistently. In areas of the lateral temporal and the parietal cortex,
some studies have found reduced regional BF and metabolism (37, 34, 36, 54).
Some of these areas appear to be involved in processing sensory information.
The significance of BF reductions in such areas in depression is unclear.
These abnormalities may reflect areas that become deactivated as the brain en-
gages in emotional processing, and they may relate to some of the neuropsy-
chological impairments seen in depression (2).
350 DREVETS
Table 2 Antidepressant treatment effects upon ventral prefrontal cortical blood flow and me-tabolism in major depressiona
References Treatment modality Change in BF or glucose metabolismpost- vs pretreatment scans
Bonne et al (29) ECT ↑ anterior cingulate in responders,NS in ventral anterolateral PFC
Cohen et al (36) phototherapy ↓ medial orbital Cb
Drevets & Raichle (47) desipramine ↓ left ventrolateral PFC
Drevets et al (48) sertraline ↓ ventrolateral PFC and orbital C
Ebert et al (39) sleep deprivation ↓ orbital C in respondersb
George et al (49) RTMS ↓ orbital C in responders
Goodwin et al (46) various drug treatments ↑ anterior cingulate, NS in ventralanterolateral
Nobler et al (30) ECT ↓ left ventrolateral PFC, respondersc
Rubin et al (31) nortriptyline or sertraline ↓ left ventrolateral PFC, respondersc
Trivedi et al (43) fluoxetine ↓ left orbital C
Wu et al (45) sleep deprivation ↓ anterior cingulate, responders
aAbbreviations: C, cortex; ECT, electroconvulsive therapy; PFC, prefrontal cortex; RTMS, re-peated transcranial magnetic stimulation.↑ , ↓ , and NS indicate increases, decreases, or no signifi-cant changes, respectively, in the treated relative to the untreated state for regions assessed. Notall studies examined the same regions, and the absence of a listed result for a specific region indi-cates that no image data were provided for that region. The changes in these ventral PFC regionsshow similar results to studies of antidepressant drug treatment in obsessive compulsive disor-dered samples (3). In contrast to these ventral prefrontal changes in depression, BF and metabo-lism in the dorsal anterior cingulate and the dorsolateral PFC have been shown to increase insome studies of depression (e.g. 28, 40), but to decrease in others (41, 42) following effective an-tidepressant treatment.
bThe treatment-associated change reported in this study was not shown by paired statisticaltests but rather by the observation that in the treatment responders, the abnormal increase that wasevident pretreatment was not present post-treatment.
cThese studies were performed using the radiotracer, xenon-133, which only provides BFmeasures near the scalp. Thus results were not available for the orbital or the cingulate cortices(see text).
Finally, abnormally increased BF has been reported in the cerebellar vermis
in major depression (33). As in the case of the orbital cortex, BF in the vermis
increases in experimentally induced states of anxiety or sadness in healthy
subjects and in anxiety states elicited in subjects who have anxiety disorders
(8, 55, 68). This finding affects the interpretation of the results of early SPET
studies of depression, which normalized flow in all other brain regions by cere-
bellar flow. For example, Philpot et al (54) showed that while comparisons of
regional-to-cerebellar ratios resulted in apparent flow reductions in most re-
gions assessed in depressives relative to controls, when the same data were re-
calculated as ratios relative to the ipsilateral occipital cortex (where flow gen-
erally does not differ between depressives and controls), only the right parietal
cortex BF ratio remained significantly decreased in depression.
Primary Major Depressive Subtypes Requiring SpecialConsiderations
The abnormal BF and metabolic increases in the ventral PFC and the amygdalahave not been demonstrable in all primary depressive samples. Elderly depres-sives who have WMH have not demonstrated abnormal elevations in these ar-eas (e.g. 21). In addition, depressive subgroups who are nonresponsive to anti-depressant treatments have shown either reductions (40) or no differences (39,45) in anterior cingulate and orbital cortex activity with respect to controlgroups.
Elderly depressives with MRI evidence of cerebrovascular disease (seen as
WMH or lacunae in T2 weighted images) show reduced BF and metabolic
measures, putatively on a vascular/ischemic basis (18, 21). To examine the
physiological correlates of depression in elderly samples, it is thus necessary
either to exclude such subjects (who are clinically characterized by having an
age-of-depression-onset after age 60 and relatively greater degrees of cogni-
tive impairment) or to study such subjects as their own controls in pre- versus
posttreatment comparisons. A study of elderly depressives that excluded sub-
jects who had moderate or severe WMH reported the same findings as found in
young depressives, namely, increased BF in the orbital cortex, the ventro-
lateral PFC, and the amygdala (41). In a study of elderly depressed samples
that included subjects who had WMH, successful treatment with either elec-
troconvulsive therapy (ECT) or antidepressant drug therapy resulted in a fur-
ther decrease in BF in the left ventrolateral PFC (30, 31). These data converge
with those presented in Table 2 to indicate that the depressed state is associated
with elevated ventral prefrontal activity relative to the remitted state.Imaging studies in bipolar depressives also appear to be confounded by
neuromorphological abnormalities, as volumetric reductions in some prefron-
NEUROIMAGING STUDIES OF DEPRESSION 351
tal cortical and limbic structures have been reported in bipolar samples (14,
56). Such morphometric abnormalities presumably reduce the magnitude of
PET and SPET measures in the vicinity of these structures (6). Consistent with
this concern, identification of hypermetabolism in the orbital cortex, the
amygdala, and the medial thalamus in bipolar depressives has in some studies
depended on the use of high (e.g. 5-mm full-width at half-width maximum
[FWMH; see Reference 4 for description]) rather than low spatial resolution
PET images (e.g. 17-mm FWHM), since the effects of partial volume averag-
ing diminish as resolution increases (51).Finally, some means of enriching subject samples for depressives likely to
have biological markers for depression may be needed to demonstrate evi-
dence of functional neuroimaging abnormalities in small samples. For exam-
ple, unipolar depressives with familial loading for alcoholism are also less
likely to have neuroendocrine and sleep EEG abnormalities (i.e. other biologi-
cal markers for depression) as compared to familial pure depressives (1). Simi-
larly, samples of unipolar depressives with familial loading for alcoholism
have mean metabolic values in the above-mentioned regions, between those of
healthy controls and of unipolar depressives with familial-pure depressive dis-
ease (38). In other studies, elevated anterior cingulate, orbital, and amygdala
activity have been shown in subsets whose depressive symptoms improve with
total sleep deprivation but not in subgroups who are unresponsive to sleep dep-
rivation (39, 45).
Anatomical Circuits Related to Depression
Since alterations in BF and metabolism primarily reflect changes in local syn-aptic activity, interpreting the regional abnormalities in depression requiresconsideration of anatomical connectivity. Elevated regional BF and metabo-lism may signify increased neurotransmission (excitatory or inhibitory) fromafferent projections arising from within the same structure or from a distalstructure (4, 5). Conversely, reductions in regional BF or metabolism may re-flect a decrease in afferent transmission (e.g. owing to active inhibition at up-stream synapses).
The imaging data in primary major depression converge with evidence
from lesion analyses to implicate circuits involving parts of the frontal and
temporal lobes along with related parts of the striatum, pallidum, and thalamus
in the pathophysiology of depression (1, 37, 12). For example, the findings in
primary unipolar depressives of abnormally increased BF and metabolism in
the ventrolateral and orbital portions of the PFC, the amygdala, and the medial
thalamus (Figure 1a–c, pp. C-8, C-9), coupled with indications of reduced
flow in the medial caudate, implicate two interconnected circuits in the patho-
352 DREVETS
physiology of depression. A limbic-thalamo-cortical circuit involves the
amygdala, the mediodorsal nucleus of the thalamus (in the medial thalamus),
and the ventral PFC; and a limbic-striatal-pallidal-thalamic circuit involves re-
lated parts of the striatum and the ventral pallidum as well as the components
of the other circuit (37).The amygdala and the prefrontal cortical regions are connected by excita-
tory projections with each other and with the mediodorsal nucleus of the thala-
mus. As a result, increased metabolic activity in these structures would pre-
sumably reflect increased synaptic transmission through the limbic-thalamo-
cortical circuit. Consistent with this hypothesis, neurosurgical procedures that
ameliorate treatment-resistant depression interrupt projections within these
circuits, and effective antidepressant drug treatments reduce BF and metabo-
lism in the amygdala, the ventral PFC, and the medial thalamus (37, 47, 48).
However, transmission through these circuits likely differs across depressive
subtypes, since the lesions involving the PFC (i.e. tumors or infarctions) and
the diseases of the basal ganglia (e.g. Parkinson’s or Huntington’s diseases),
which are associated with higher rates of depression than other similarly de-
bilitating conditions, result in dysfunction at distinct points within these cir-
cuits that affect synaptic transmission in different ways (1). This concept finds
an analogy in the more dorsal motor circuitry, in which lesions at various
points can result in paraplegia but are associated with dissimilar patterns of ab-
normal synaptic transmission.
Functional Imaging Studies of Secondary Depression
Consistent with a circuitry-based approach to the functional anatomy of de-pression, results of imaging studies of depressive syndromes arising secondaryto neurological disorders generally differ from those reported for primary de-pressive syndromes. In contrast to the findings of increased BF or metabolismin parts of the orbital cortex in primary depressives (Table 1), orbital BF or me-tabolism is reportedly decreased or not significantly different in subjects whohave depressive syndromes arising secondary to Parkinson's disease, Hunting-ton's disease, or basal ganglia infarction relative to nondepressed subjects whohave the same illnesses (Table 3). Similarly, in the dorsolateral PFC, BF andmetabolism are generally not significantly different between subjects whohave neurological diseases who meet criteria for major depression and thosewho do not, which contrasts with the findings in primary depression (Table 1).
Studies of the physiological correlates of depressive symptoms in subjects
who have primary psychiatric disorders other than mood disorders have re-
ported abnormalities that more closely resemble those found in primary de-
pressive samples. In a study of subjects who had bulimia nervosa (63), metabo-
NEUROIMAGING STUDIES OF DEPRESSION 353
lism trended toward being abnormally reduced in the dorsolateral PFC, where
it correlated negatively with depression ratings [as reported in primary depres-
sion (28)]. Metabolism tended to be abnormally elevated in the orbital cortex,
where it correlated negatively with obsessive-compulsive ratings. These re-
sults are similar to the relationships found between posterior orbital BF and
obsession ratings in primary obsessive-compulsive disorder (9) and between
orbital metabolism and depressive ideation ratings in primary depression (38).In addition, cocaine-dependent subjects scanned within the first week of co-
caine withdrawal as they experienced both depressive symptoms and cocaine
craving had elevated glucose metabolism in the medial orbital cortex and the
basal ganglia relative to healthy controls (64). The direction of the metabolic
abnormality in the orbital cortex resembled that seen in primary depression
(Table 1), but the direction of the abnormality in the basal ganglia was opposite
that found in the basal ganglia of primary unipolar depressives (37, 22, 35, 53).Primary and secondary depression thus appear to involve the same general
structures, though the patterns and the direction of the abnormalities within
these structures in some cases differs between primary and secondary depres-
354 DREVETS
Table 3 Regional blood flow and metabolic abnormalities in the ventral prefrontal cortex inprimary versus secondary neuropsychiatric syndromesa
Primary Secondary References (and primary diagnoses)
Unipolar majordepression ⇑ b
⇓
NS
Mayberg et al (57) (Parkinson’sdisease)c
Mayberg et al (58) (Huntington’sdisease)c
Mayberg et al (59) (basal gangliainfarction)c
Ring et al (60) (Parkinson’s disease)c
Obsessive-compulsivedisorder ⇑ d
⇓
NS
LaPlane et al (61) (basal ganglialesions)e
George et al (62) (Tourette’s syndrome)f
aAbbreviations: D, disease; NS, difference not significant. Additional studies of subjects whohad depressive symptoms but did not meet criteria for the major depressive syndrome are re-viewed in the text.
bBF and metabolism generally increased in unmedicated, primary depressives versus controls(Table 1).
cComparison between depressed and nondepressed subjects with the same illness.dBF and metabolism generally increased in unmedicated subjects with primary obsessive-
compulsive disorder relative to controls (3, 32).eSubjects with secondary obsessive-compulsive syndrome compared to healthy controls.fComparison between Tourette’s subjects who manifest obsessions and compulsions vs.
Tourette’s subjects without such features.
sive syndromes. This observation provides further support for a circuitrymodel in which mood disorders are associated with dysfunctional interactionsbetween multiple structures, rather than increased or decreased activity withina single structure (1).
Other neuropsychiatric syndromes that occur comorbidly with MDD mayalso involve the same neural circuits. For example, the orbital-striatal-pallidal-thalamic circuits have also been implicated in the pathophysiology ofobsessive-compulsive disorder (3, 9, 32), potentially providing insight into thecomorbidity between depressive and obsessional syndromes (1). The differ-ences found in the ventral PFC between primary and secondary depressionparallel the findings in primary and secondary obsessive-compulsive syn-dromes, in which ventral PFC metabolism is increased in the former but de-creased or unchanged in the latter (Table 3).
FUNCTIONAL CORRELATES OF NEUROIMAGINGABNORMALITIES IN DEPRESSION
Regional BF and metabolic abnormalities in depressed subjects may reflect (a)the pathophysiological changes that predispose individuals to recurrent, ab-normal mood episodes; (b) the physiological concomitants of signs and symp-toms of depression; (c) the physiological correlates of neurotransmitter systemdysfunction; (d) the compensatory mechanisms invoked to modulate or inhibitpathological processes; or (e) the areas that have been deactivated during on-going emotional processing. Hypotheses regarding the functional significanceof BF and metabolic correlates of depression thus depend on converging infor-mation from functional imaging studies performed in healthy controls or in pa-tients who have related diseases and from electrophysiological and lesionanalysis studies of humans and monkeys.
The Dorsal Prefrontal Cortex and NeuropsychologicalImpairment
The areas of reduced BF and metabolism found in the dorsolateral and the dor-
somedial PFC in depression have been linked to the neuropsychological mani-
festations of depression (2, 65, 66). Regional BF does not increase in these re-
gions during experimentally induced emotional states (2, 12). Instead, in the
vicinity of the dorsolateral prefrontal areas where BF and metabolism are de-
creased in depression, blood flow increases in healthy subjects imaged as they
perform tasks involving visuospatial memory, pictorial memory for visual ob-
jects, maze navigation, or manipulation of memorized verbal information (2).
Dolan et al (65) proposed that this abnormality reflects nonspecific slowing of
cognitive processing, based upon their findings that BF is also decreased in
NEUROIMAGING STUDIES OF DEPRESSION 355
this area in subjects who have schizophrenia, and that the BF reduction corre-
lates with ratings of impoverished speech both in depressed and in schizo-
phrenic subjects. In contrast, Dolan et al (66) more specifically linked abnor-
mal BF in the dorsomedial PFC (which includes the dorsal anterior cingulate
gyrus) to impaired performance on neuropsychological testing.
The Orbital Cortex: Potential Role in Correcting Perseverative,Unreinforced Behavior
The abnormal elevations of BF and metabolism in the ventrolateral and orbital
areas of the PFC and the pregenual anterior cingulate cortex in depression are
thought to reflect physiological activity related to ongoing emotional process-
ing and/or obsessive ruminations (2, 9, 37, 12, 32, 69). Regional BF increases
in these areas during experimentally induced sadness or anxiety in healthy sub-
jects and during induced anxiety and/or obsessional states in subjects who
have obsessive-compulsive disorder, panic disorder, post-traumatic stress dis-
order, and simple animal phobia (e.g. 2, 8, 9, 10, 68). Lesion analyses in hu-
mans and experimental animals; single-unit recording studies in monkeys; and
autoradiographic studies in experimental animals exposed to stressors, threats,
or appetitive stimuli also support a role for these areas in emotional behavior
(12, 50, 67). Finally, these prefrontal areas share the most extensive anatomi-
cal connections with subcortical areas implicated in emotional behavior, such
as the amygdala, the ventral striatum, and the lateral hypothalamus. Consistent
with their proposed role in ongoing emotional processing during depressive
episodes, BF and metabolism decrease in the ventrolateral PFC and the orbital
cortex as depressive symptoms remit during treatment (Table 2).Parts of the orbital cortex may be activated in depression to mediate at-
tempts to modulate or inhibit emotional responses (69). Although flow and
metabolism are abnormally elevated in the lateral and posterior orbital cortices
during depression, the magnitude of the physiological activity in these regions
correlates inversely with ratings of depression severity and negative thought
frequency (37, 38). Similarly, in obsessive-compulsive or animal phobic sub-
jects scanned during exposure to phobic stimuli and in healthy subjects imaged
during induced sadness, BF in the posteromedial orbital cortex increases, yet
the magnitude of the BF change correlates inversely with ratings of obsessive
thinking, anxiety, and sadness, respectively (9, 10, 68). These findings are
consistent with evidence from electrophysiological studies and lesion analyses
indicating that the orbital cortex plays a role in correcting behavioral or emo-
tional responses that become inappropriate as reinforcement contingencies
change, and in inhibiting emotion-like responses elicited by electrical stimula-
tion of the amygdala (70–72). Thus, in primary major depression, activation of
356 DREVETS
the posterior orbital cortex may reflect endogenous attempts to break per-
severative patterns of negative thought and emotion. In contrast, in some sec-
ondary depressive syndromes, a perseverative emotional state may result from
disruption of orbital cortex function either by lesions within the PFC or by le-
sions of the basal ganglia that interrupt neural interactions between the orbital
cortex and the striatum (1, 69).
The Amygdala and Emotional Behavior
The amygdala is the only structure in which regional BF and glucose metabo-lism consistently correlate positively with depression severity (37, 38, 52).The preliminary observation that amygdala metabolism remains abnormallyelevated during sleep implies that amygdala hypermetabolism reflects a patho-logical process rather than a physiological correlate of conscious behaviors orthoughts during scanning (42). In addition to being elevated in the depressedstate, BF and metabolism in the left amygdala appear abnormally elevated(though to a lesser extent) in asymptomatic (i.e. between depressive episodes),familial depressives who are not receiving treatment (37). Conversely, duringantidepressant drug treatment that both ameliorates the depressive symptomsand helps prevent relapse, amygdala metabolism decreases toward normal(48).
Consistent with these observations, antidepressant drug–treated, remitted
subjects with MDD who relapse when given a tryptophan-free diet (which
putatively depletes CNS serotonin levels) have higher baseline amygdala me-
tabolism (prior to depletion) than similar subjects who did not relapse (73).
These data are thus compatible with the hypothesis that abnormally elevated
amygdala activity may confer susceptibility to relapse and recurrence of de-
pressive episodes (1, 74). Finally, abnormal resting metabolism in the amyg-
dala has not been reported in other conditions, suggesting that this finding may
be specific to primary mood disorders.The amygdala has been implicated by single-unit and lesion studies in as-
signing emotional significance to experiential stimuli and in organizing the be-
havioral, autonomic, and neuroendocrine manifestations of emotional expres-
sion (12, 51). For example, amygdala activity stimulates corticotropin releas-
ing factor (CRF) containing neurons in the paraventricular nucleus of the hy-
pothalamus to release CRF via both direct and indirect projections (75). Dur-
ing depressive episodes, amygdala metabolism correlates positively with
plasma cortisol levels (52). Because amygdala hypermetabolism and cortisol
release may be linked in depression, the failure to normalize glucocorticoid
function during antidepressant treatment may be associated with an increased
risk of depressive relapse (76).
NEUROIMAGING STUDIES OF DEPRESSION 357
CONCLUDING REMARKS
The future of functional neuroimaging research appears bright, as advances inPET and functional MRI (fMRI) technologies and the availability of newradioligands for receptor imaging progressively increase the sophistication ofthe experimental questions that can be addressed using imaging techniques. Indepression research, PET and SPET will be used increasingly to quantify neu-roreceptor binding and neurotransmitter function, permitting fuller characteri-zation of the neurochemical abnormalities suggested by studies of body fluids,postmortem tissue, and neuroendocrine function. In addition, PET and fMRIwill be used to assess hemodynamic changes during the performance of neu-ropsychological and emotional tasks to examine the functional responses ofspecific brain structures in depression. The integration of such data with thoseobtained using lesion analysis and postmortem histopathological approacheswill increasingly refine our understanding of the anatomical correlates of mel-ancholia.
Visit the Annual Reviews home page at
http://www.AnnualReviews.org.
358 DREVETS
Literature Cited
1. Drevets WC, Todd RD. 1997. Depres-sion, mania and related disorders. In AdultPsychiatry, ed. SB Guze, 8:99–141. St.Louis, MO: Mosby
2. Drevets WC, Raichle ME. 1997. Recipro-cal suppression of regional cerebral bloodflow during emotional versus higher cog-nitive processes: implications for interac-tions between emotion and cognition.Cogn. Emot. In press
3. Drevets WC, Botteron K. 1997. Neuroi-maging in psychiatry. In Adult Psychia-try, ed. SB Guze, 6:53–81. St. Louis, MO:Mosby
4. Raichle ME. 1987. Circulatory and meta-bolic correlates of brain function in nor-mal humans. In Handbook of Physiol-ogy—The Nervous System V, ed. VBMountcastle, F Plum, SR Geiger,16:643–74. Baltimore, MD: Williams &Wilkins
5. DiRocco RJ, Kageyama GH, Wong-RileyMTT. 1989. The relationship betweenCNS metabolism and cytoarchitecture: areview of 14C-deoxyglucose studies withcorrelation to cytochrome oxidase histo-
chemistry. Comp. Med. Imag. Graph. 13:81–92
6. Mazziotta JC, Phelps ME, Plummer D, etal. 1981. Quantitation in positron emis-sion computed tomography. 5. Physical-anatomical effects. J. Comput. Assist. To-mogr. 5:734–43
7. Marangell LB, Ketter TA, George MS, etal. 1997. Inverse relationship of periph-eral thyrotropin-stimulating hormone lev-els to brain activity in mood disorders. AmJ. Psychiat. 154: 224-30
8. George MS, Ketter TA, Parekh PI, et al.1995. Brain activity during transient sad-ness and happiness in healthy women.Am. J. Psychiatr. 152:341–51
9. Rauch SL, Jenike MA, Alpert NM, et al.1994. Regional cerebral blood flow meas-ured during symptom provocation inobsessive-compulsive disorder usingoxygen 15-labeled carbon dioxide andpositron emission tomography. Arch.Gen. Psychiatr. 51:62–70
10. Schneider F, Gur RE, Alavi A, et al. 1995.Mood effects on limbic blood flow corre-late with emotion self-rating: a PET study
NEUROIMAGING STUDIES OF DEPRESSION 359
with oxygen-15 labeled water. Psychiatr.Res. Neuroimag. 61:265–83
11. Talairach J, Tournoux P. 1988. Co-Planar Stereotaxic Atlas of the HumanBrain. Stuttgart, Germany: Thieme
12. Price JL, Carmichael ST, Drevets WC.1996. Networks related to the orbital andmedial prefrontal cortex: a substrate foremotional behavior? Prog. Brain Res.107:523–36
13. Drevets WC. 1994. Geriatric depression:brain imaging correlates and pharma-cologic considerations. J. Clin. Psychiatr.55(9, Suppl. A):74
14. Drevets WC, Price JL, Simpson JR, et al.1997. Subgenual prefrontal cortex abnor-malities in mood disorders. Nature 386:824–27
15. Wooten GF, Collins RC. 1981. Metaboliceffects of unilateral lesion of the substan-tia nigra. J. Neurosci. l:285–91
16. Silfverskiöld P, Risberg J. 1989. Regionalblood flow in depression and mania.Arch. Gen. Psychiatr. 46: 253–59
17. Maes M, Dierckx R, Meltzer H, et al.1993. Regional cerebral BF in unipolardepression measured with Tc-99m-HMPAO SPECT: negative findings. Psy-chiatr. Res. Neuroimag. 50:77–88
18. Krishnan KR, Hays JC, Blazer DG. 1997.MRI-defined vascular depression. Am J.Psychiatr. 154:497–50
19. Awad IA, Johnson PC, Spetzler RJ, et al.1986. Incidental subcortical lesions iden-tified on magnetic resonance imaging inthe elderly. II: Postmortem pathologicalcorrelations. Stroke 17:1090–97
20. Chimowitz MI, Estes ML, Furlan AJ, etal. 1992. Further observations on the pa-thology of subcortical lesions identifiedon magnetic resonance imaging. Arch.Neurol. 49:747–52
21. Lesser IM, Mena I, Boone KB, et al. 1994.Reduction of cerebral blood flow in olderdepressed patients. Arch. Gen. Psychiatr.51:677–86
22. Baxter LR, Phelps ME, Mazziotta JC, etal. 1985. Cerebral metabolic rates for glu-cose in mood disorders. Arch. Gen. Psy-chiatr. 42:441–47
23. Deleted in proof24. Goldman-Rakic PS. 1977. Circuitry of
primate prefrontal cortex and regulationof behavior by representational memory.In Handbook of Physiology—The Nerv-ous System V, ed. VB Mountcastle, FPlum, SR Geiger, 16:373–417. Balti-more, MD: Williams & Wilkins
25. Petersen SE, Fox PT, Posner MI, et al.1989. Positron emission tomographic
studies of the processing of single words.J. Cogn. Neurosci. 1:153–70
26. Shulman GL, Corbetta M, Buckner RL, etal. 1997. Common blood flow changesacross visual tasks. II. Decreases in cere-bral cortex. J. Cogn. Neurosci. 9(5):647–62
27. Drevets WC, Burton H, Simpson JR, et al.1995. Blood flow changes in human so-matosensory cortex during anticipatedstimulation. Nature 373:249–52
28. Baxter LR, Schwartz JM, Phelps ME, etal. 1989. Reduction of prefrontal cortexglucose metabolism common to threetypes of depression. Arch. Gen. Psychi-atr. 46:243–50
29. Bonne O, Krausz Y, Shapira B, et al.1996. Increased cerebral blood flow in de-pressed patients responding to electro-convulsive therapy. J. Nuclear Med. 37:1075–80
30. Nobler MS, Sackeim HA, Prohovnik I, etal. 1994. Regional cerebral blood flow inmood disorders. III. Treatment and clini-cal response. Arch. Gen. Psychiatr. 51:884–97
31. Rubin E, Sackeim HA, Nobler MS, et al.1994. Brain imaging studies of antide-pressant treatments. Psychiatr. Ann.24(12):653–58
32. Baxter LR, Phelps ME, Mazziotta JC, etal. 1987. Local cerebral glucose meta-bolic rates in obsessive-compulsive disor-der—A comparison with rates in unipolardepression and in normal controls. Arch.Gen. Psychiatr. 44:211–18
33. Bench CJ, Friston KJ, Brown RG, et al.1992. The anatomy of melancholia—fo-cal abnormalities of cerebral blood flowin major depression. Psychiatr. Med. 22:607–15
34. Biver F, Goldman S, Delvenne V, et al.1994. Frontal and parietal metabolic dis-turbances in unipolar depression. Biol.Psychiatr. 36:381–88
35. Buchsbaum MS, DeLisi LE, Holcomb H,et al. l986. Frontal cortex and basal gan-glia metabolic rates assessed by PET with[18F]2-deoxyglucose in affective illness.J. Affect. Dis. 10:137–52
36. Cohen RM, Gross M, Nordahl TE, et al.1992. Preliminary data on the metabolicbrain pattern of patients with winter sea-sonal affective disorder. Arch. Gen. Psy-chiatr. 49:545– 52
37. Drevets WC, Videen TO, Price JL, et al.l992. A functional anatomical study ofunipolar depression. J. Neurosci. 12:3628–41
38. Drevets WC, Spitznagel E, Raichle ME.
360 DREVETS
1995. Functional anatomical differencesbetween major depressive subtypes. J.Cereb. B. F. Metab. 15(1):S93
39. Ebert D, Feistel H, Barocka A. 1991. Ef-fects of sleep deprivation on the limbicsystem and the frontal lobes in affectivedisorders: a study with Tc-99m-HMPAOSPECT. Psychiatr. Res. Neuroimag. 40:247–51
40. Mayberg HS, Brannan SK, Mahurin RK,et al. 1997. Cingulate function in depres-sion: a potential predictor of treatment re-sponse. NeuroReport 8(4):1057–61
41. Mentis MH, Krasuski J, Pietrini P, et al.1995. Cerebral glucose metabolism in lateonset depression without cognitive im-pairment. Soc. Neurosci. 21(3):1736(Abstr.)
42. Nofzinger E, Keshevan M, Buysse D, etal. The neurobiology of sleep in relationto mental disorder. In The Neurobiologi-cal Foundation of Mental Illness, ed. DSCharney, EJ Nestler, BS Bunney. NewYork: Oxford University Press. In press
43. Trivedi MH, Devous MD, Rush AJ. 1994.RCBF findings in major depression: pre-and post-treatment with fluoxetine.American College of Neuropsychophar-macology Abstracts 149 (Abstr.)
44. Uytdenhoef P, Portelange P, Jacquy J, etal. 1983. Regional cerebral blood flowand lateralized hemispheric dysfunctionin depression. Br. J. Psychiatr. 143:128–32
45. Wu JC, Gillin JC, Buchsbaum MS, et al.1992. Effect of sleep deprivation on brainmetabolism of depressed patients. Am. J.Psychiatr. 149:538–43
46. Goodwin GM, Austin MP, Dougall N, etal. 1993. State changes in brain activityshown by the uptake of 99mTc-exa-metazine with SPET in major depressionbefore and after treatment. J. Affect. Dis.29:243–53
47. Drevets WC, Raichle ME. l992. Neuroa-natomical circuits in depression: implica-tions for treatment mechanisms. Psycho-pharmacol. Bull. 28:26l–74
48. Drevets WC, Price JL, Simpson JS, et al.1996. State- and trait-like neuroimagingabnormalities in depression: effects of an-tidepressant treatment. Soc. Neurosci.22(1):266 (Abstr.)
49. George MS. 1996. Repetitive transcra-nial magnetic stimulation: a novel probeof mood. Presented at 35th Annu. Meet.Am. Coll. Neuropsychopharmacol., SanJuan, Puerto Rico
50. LeDoux JE. 1977. Emotion. In Handbookof Physiology—The Nervous System V,
ed. VB Mountcastle, F Plum, SR Geiger,16:419–59. Baltimore, MD: Williams &Wilkins.
51. Drevets WC, Price JL, Todd RD, et al.1997. PET measures of amygdala me-tabolism in bipolar and unipolar depres-sion: correlation with plasma cortisol.Soc. Neurosci. 23(2):1407(Abstr.)
52. Abercrombie HC, Larson CL, Ward RT,et al. 1996. Metabolic rate in the amyg-dala predicts negative affect and depres-sion severity in depressed patients: anFDG-PET study. Neuroimage 3:S217(Abstr.)
53. Schwartz JM, Baxter LR, Mazziotta JC, etal. 1987. The differential diagnosis of de-pression: relevance of positron emissiontomography studies of cerebral glucosemetabolism to the bipolar-unipolar di-chotomy. J. Am. Med. Assoc. 258:1368–73
54. Philpot MP, Banerjee S, Needham-Bennett H, et al. 1993. 99mTc-HMPAOsingle photon emission tomography inlate life depression: a pilot study of re-gional cerebral blood flow at rest and dur-ing a verbal fluency task. J. Affect. Dis.28:233–40
55. Reiman EM, Raichle ME, Robins E, et al.1989. Neuroanatomical correlates of alactate-induced anxiety attack. Arch.Gen. Psychiatr. 46:493–500
56. Pearlson GD, Barta PE, Powers RE, et al.1997. Medial and superior temporal gyralvolumes and cerebral asymmetry inschizophrenia versus bipolar disorder.Biol. Psychiatr. 41:1–14
57. Mayberg HS, Starkstein SE, Sadzot B, etal. 1990. Selective hypometabolism in theinferior frontal lobe in depressed patientswith Parkinson's disease. Ann. Neurol.28:57–64
58. Mayberg HS, Starkstein SE, Peyser CE, etal. 1992. Paralimbic frontal lobe hypome-tabolism in depression associated withHuntington's disease. Neurology 42:1791–97
59. Mayberg HS, Starkstein SE, Morris PL, etal. 1991. Remote cortical hypometabo-lism following focal basal ganglia injury:relationship to secondary changes inmood. Neurology 41:266 (Suppl.)
60. Ring HA, Bench CJ, Trimble MR, et al.1994. Depression in Parkinson's Disease:a positron emission study. Br. J. Psychi-atr. 165:333–39
61. LaPlane D, Levasseur M, Pillon B, et al.1989. Obsessive-compulsive and otherbehavioral changes with bilateral basalganglia lesions. Brain 112:699–725
NEUROIMAGING STUDIES OF DEPRESSION 361
62. George MS, Trimble MR, Costa DC, et al.1992. Elevated frontal cerebral bloodflow in Gilles de la Tourette Syndrome: a99Tcm-HMPAO SPECT study. Psychiatr.Res. 45:143–51
63. Andreason PJ, Altemus MS, ZametkinAJ, et al. 1992. Regional cerebral glucosemetabolism in bulimia nervosa. Am. J.Psychiatr. 149:1506–13
64. Volkow ND, Fowler JS, Wolf AP, et al.1991. Changes in brain glucose metabo-lism in cocaine dependence and with-drawal. Am. J. Psychiatr. 148:621–66
65. Dolan RJ, Bench CJ, Liddle PF, et al.1993. Dorsolateral prefrontal cortex dys-function in the major psychoses: symp-tom or disease specificity? J. Neurol.Neurosurg. Psychiatr. 56:1290–94
66. Dolan RJ, Bench CJ, Brown RG, et al.1994. Neuropsychological dysfunction indepression: the relationship to regionalcerebral BF. Psychiatr. Med. 24:849–57
67. Damasio AR. 1994. Descarte’s Error:Emotion, Reason, and the Human Brain,pp. 3–253. London/ New York: MacMil-lan
68. Drevets WC, Simpson JR, Raichle ME.1995. Regional blood flow changes in re-sponse to phobic anxiety and habituation.J. Cereb. B.F. Metab. 15(1):S856
69. Drevets WC. 1995. PET and the func-tional anatomy of major depression. InEmotion, Memory and Behavior-Study ofHuman and Nonhuman Primates. TNakajima and T Ono, ed. 4:43–62. To-
kyo: Japan Scientific Societies70. Morgan MA, Romanski LM, LeDoux JE.
1993. Extinction of emotional learning:contribution of medial prefrontal cortex.Neurosci. Lett. 163:109–13
71. Rolls ET. 1995. A theory of emotion andconsciousness, and its application to un-derstanding the neural basis of emotion.In The Cognitive Neurosciences, ed. MGazzaniga, 72:1091–106. Cambridge,MA: MIT Press
72. Timms RJ. 1977. Cortical inhibition andfacilitation of the defense reaction. J.Physiol. Lond. 266:98–99P
73. Bremner JD, Innis RB, Salomon RM, etal. 1997. Positron emission tomographymeasurement of cerebral metabolic corre-lates of tryptophan depletion-induced de-pressive relapse. Arch. Gen. Psychiatr.54:346–74
74. Post RM. 1992. Transduction of psycho-social stress into the neurobiology of re-current affective disorder. Am. J. Psychi-atr. 149:999–1010
75. McEwen BS. 1995. Stressful experience,brain, and emotions: developmental, ge-netic and hormonal influences. In TheCognitive Neurosciences, ed. M Gazza-niga, 74:1117–35. Cambridge, MA: MITPress
76. Ribeiro SCM, Tandon R, Grunhaus L, etal. 1993. The DST as a predictor of out-come in depression: a meta-analysis. Am.J. Psychiatr. 150:1618–29