ArticlePDF Available

Abstract and Figures

The hypothalamus and its subdivisions are involved in many neuropsychiatric conditions such as affective disorders, schizophrenia, or narcolepsy, but parcellations of hypothalamic subnuclei have hitherto been feasible only with histological techniques in postmortem brains. In an attempt to map subdivisions of the hypothalamus in vivo, we analyzed the directionality information from high-resolution diffusion-weighted magnetic resonance images of healthy volunteers. We acquired T1-weighted and diffusion-weighted scans in ten healthy subjects at 3 T. In the T1-weighted images, we manually delineated an individual mask of the hypothalamus in each subject and computed in the co-registered diffusion-weighted images the similarity of the principal diffusion direction for each pair of mask voxels. By clustering the similarity matrix into three regions with a k-means algorithm, we obtained an anatomically coherent arrangement of subdivisions across hemispheres and subjects. In each hypothalamus mask, we found an anterior region with dorsoventral principal diffusion direction, a posteromedial region with rostro-caudal direction, and a lateral region with mediolateral direction. A comparative analysis with microstructural hypothalamus parcellations from the literature reveals that each of these regions corresponds to a specific group of hypothalamic subnuclei as defined in postmortem brains. This is to our best knowledge the first in vivo study that attempts a delineation of hypothalamic subdivisions by clustering diffusion-weighted magnetic resonance imaging data. When applied in a larger sample of neuropsychiatric patients, a structural analysis of hypothalamic subnuclei should contribute to a better understanding of the pathogenesis of neuropsychiatric conditions such as affective disorders.
Content may be subject to copyright.
ORIGINAL PAPER
Diffusion imaging-based subdivision of the human hypothalamus:
a magnetic resonance study with clinical implications
Peter Scho
¨nknecht Alfred Anwander Friederike Petzold
Stephanie Schindler Thomas R. Kno
¨sche Harald E. Mo
¨ller
Ulrich Hegerl Robert Turner Stefan Geyer
Received: 2 November 2012 / Accepted: 18 December 2012 / Published online: 4 January 2013
ÓSpringer-Verlag Berlin Heidelberg 2013
Abstract The hypothalamus and its subdivisions are
involved in many neuropsychiatric conditions such as
affective disorders, schizophrenia, or narcolepsy, but par-
cellations of hypothalamic subnuclei have hitherto been
feasible only with histological techniques in postmortem
brains. In an attempt to map subdivisions of the hypo-
thalamus in vivo, we analyzed the directionality informa-
tion from high-resolution diffusion-weighted magnetic
resonance images of healthy volunteers. We acquired
T1-weighted and diffusion-weighted scans in ten healthy
subjects at 3 T. In the T1-weighted images, we manually
delineated an individual mask of the hypothalamus in each
subject and computed in the co-registered diffusion-
weighted images the similarity of the principal diffusion
direction for each pair of mask voxels. By clustering the
similarity matrix into three regions with a k-means algo-
rithm, we obtained an anatomically coherent arrangement
of subdivisions across hemispheres and subjects. In each
hypothalamus mask, we found an anterior region with
dorsoventral principal diffusion direction, a posteromedial
region with rostro-caudal direction, and a lateral region
with mediolateral direction. A comparative analysis with
microstructural hypothalamus parcellations from the liter-
ature reveals that each of these regions corresponds to a
specific group of hypothalamic subnuclei as defined in
postmortem brains. This is to our best knowledge the first
in vivo study that attempts a delineation of hypothalamic
subdivisions by clustering diffusion-weighted magnetic
resonance imaging data. When applied in a larger sample
of neuropsychiatric patients, a structural analysis of
hypothalamic subnuclei should contribute to a better
understanding of the pathogenesis of neuropsychiatric
conditions such as affective disorders.
Keywords Hypothalamus Subdivision Diffusion
MRI In vivo
Introduction
The hypothalamus is a gray matter brain region of about
4cm
3
neuronal tissue [57] situated on both sides of the
third ventricle, but recent histological or neuroimaging
studies reported it to be no more than half this size. The
variability of the reported volumes might be caused by
different methodical approaches as well as different ana-
tomical definitions of the hypothalamic region.
As part of the limbic system, it is involved in neuro-
psychiatric conditions, such as affective disorders [2,7,16,
47], schizophrenia [6,12], or narcolepsy [45]. Evidence of
associated volumetric or structural hypothalamic changes
exists [9,15,20], but it remains open whether these
abnormalities are restricted to specific subregions of the
hypothalamus.
To date, substructuring of the hypothalamus has only
been feasible postmortem based on differences in micro-
anatomy. Between 9 [33] and 17 [45] distinct nuclei and
areas have been identified. For descriptive convenience,
they are usually grouped into three or four compartments
with either mediolateral or anteroposterior orientation.
P. Scho
¨nknecht (&)F. Petzold S. Schindler U. Hegerl
Department of Psychiatry and Psychotherapy, University
Hospital Leipzig, Semmelweisstr. 10, 04103 Leipzig, Germany
e-mail: Peter.Schoenknecht@medizin.uni-leipzig.de
A. Anwander T. R. Kno
¨sche H. E. Mo
¨ller R. Turner
S. Geyer
Max Planck Institute for Human Cognitive and Brain Sciences,
Stephanstr. 1a, 04103 Leipzig, Germany
123
Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508
DOI 10.1007/s00406-012-0389-5
Crosby and Woodburne [13] distinguished three medi-
olateral zones, namely the periventricular, medial, and
lateral hypothalamus. In contrast, Le Gros Clark [34],
Braak and Braak [10,11], and Hofman and Swaab [23]
described three anteroposterior zones of the hypothalamus:
the preoptic (or supraoptic or chiasmatic) region, the tub-
eral region (or tuber cinereum), and the stronger myelin-
ated mamillary region. In an earlier work, Le Gros Clark
[33, p. 204] had suggested that the hypothalamus might be
‘conveniently subdivided’’ into the pars supraoptica, pars
infundibularis, retroinfundibular part, and pars mamillaris
hypothalami. Such a four-compartment model with anter-
oposterior orientation has also been applied in recent
studies using magnetic resonance imaging (MRI) where the
boundaries were defined by macroanatomical landmarks
[3,20].
Beyond the disputable criterion of convenient macro-
structure, the underlying microanatomy generally supports
the three-compartment model, both with mediolateral and
anteroposterior orientation. Saper [51], Koutcherov et al.
[30], and Toni et al. [59] argue that cell phenotype, protein
expression, and neurogenesis support the mediolateral
subdivision of the hypothalamus suggested by Crosby and
Woodburne [13]. In addition, Saper [51] and Toni et al.
[59] review evidence that each of these zones can be fur-
ther subdivided into three myeloarchitectonically and
functionally distinct compartments with anteroposterior
orientation according to Le Gros Clark [34], Braak and
Braak [10,11], and Hofman and Swaab [23].
Interestingly, based on a three-dimensional (3-D)
reconstruction of the human hypothalamus, Young and
Stanton [62, p. 326] also observed three ‘‘clearly discern-
ible clusters’’ when viewed from above. Unfortunately, the
authors did not elaborate on this, but the pictorial 3-D
reconstructions provided in the article show an anterolat-
eral, a medial, and a posteroperiventricular cluster, which
are in line with the mediolateral as well as with the
anteroposterior models.
Due to the small size and intricate structural organiza-
tion of the hypothalamus, it has been a challenge so far to
map its internal architecture with MRI. The first studies
have focused on white matter tracts within the hypothala-
mus which are more readily accessible to MRI technology,
such as the fornix [50] and the stria terminalis [31]. More
recent studies have ventured into the hypothalamic
gray matter in an attempt to chart its organization into
subnuclei. Lemaire et al. [36] defined several hypothalamic
compartments in an indirect way, that is, based on a pro-
portional grid system anchored to easily identifiable
extrahypothalamic landmarks (e.g., anterior commissure or
mamillary bodies) and individually scaled to the anatomy
of each single hypothalamus. Baroncini et al. [3] used a
battery of different MR sequences to provide qualitative
evidence for several intrahypothalamic substructures
such as the paraventricular, ventromedial, and infundibular
nucleus.
Recently, a more refined MRI-based parcellation of
human gray matter structures has become possible by dif-
fusion-weighted magnetic resonance imaging (dMRI)
which exploits the apparent diffusion of water. Diffusion
results from random motion of particles and is isotropic for
a uniform sample without heterogeneity, that is, the parti-
cles travel similarly fast along any direction resulting in
spherical propagation. In brain tissue, however, diffusion
of water molecules is anisotropic due to restriction by ax-
ons or cell membranes [5,14,42]. In an axon, diffusion of
water is six times faster along the orientation of the axon,
than perpendicular to it [32] which results in strong
anisotropy of water in white matter as compared to a
considerably lower anisotropy in gray matter structures [5].
With high-quality dMRI, even slight diffusion differences,
as typical for gray matter, can be detected. This has already
been successfully employed to segment subcortical gray
matter such as the thalamus [27,61] and the amygdala
[56] but, until now, has never been applied in the human
hypothalamus.
Therefore, it is the goal of the present study to exploit
the great potential of dMRI and investigate whether sub-
structures of the human hypothalamus can be distinguished
in vivo. Based on the concepts of hypothalamic subnuclei
outlined above, we hypothesized that dMRI data of the
hypothalamus could be clustered into three reproducible
entities with distinct diffusion orientations.
To this end, we interactively delineated a hypothalamus
region of interest (ROI) in T1- and T2-weighted scans from
ten healthy subjects based on a set of meticulously defined
landmarks. In a second step, we clustered these ROIs in
co-registered diffusion-weighted scans into three repro-
ducible entities that might correspond to microanatomi-
cally defined subnuclei of the hypothalamus in postmortem
brains.
Methods
Data acquisition
Ten healthy young subjects (5 males, age 26.4 ±2.5 year,
5 females, age 24.5 ±5.7 year) without history of neuro-
logical or psychiatric diseases participated in the experi-
ment. All subjects had given written informed consent, and
the study protocol was consistent with guidelines of the
Ethics Committee of the University of Leipzig, Germany.
We scanned our subjects with a 3-T TIM Trio scanner
(Siemens, Erlangen, Germany) and a 12-channel head
matrix coil. We acquired T1-weighted whole-brain images
498 Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508
123
with a 3-D magnetization-prepared rapid gradient echo
[3-D MP RAGE; [43]] sequence with selective water
excitation and linear phase encoding (inversion time
(TI) =650 ms, repetition time (TR) =1,300 ms, echo
time (TE) =3.93 ms, flip angle =10°, bandwidth =
130 Hz/pixel, image matrix =256 9240, field of view
(FOV) =256 9240 mm
2
, slab thickness =192 mm, 128
partitions, 95 % slice resolution, sagittal orientation, 2
acquisitions). To avoid aliasing, oversampling was per-
formed in the read direction (head–foot). Reconstructed
images were obtained after zero padding with a nominal
resolution of 1 9191mm
3
. We also acquired structural
T2-weighted images of the same nominal spatial resolution
with a 3-D Turbo spin echo sequence generating a variable
flip angle pulse train [1,35] with TR =3200 ms,
TE =447 ms, and generalized autocalibrating partially
parallel acquisitions [GRAPPA; [21]] with acceleration
factor 2. Finally, we acquired diffusion-weighted images
with a twice-refocused spin echo (TRSE) echo-planar-
imaging (EPI) sequence [48] with TE =100 ms,
TR =12 s, image matrix =128 9128, FOV =220 9
220 mm
2
, GRAPPA with acceleration factor 2, 60 diffu-
sion-encoding gradient directions, and a b-value of
1,000 s/mm
2
(no cardiac gating). We obtained seven ima-
ges without diffusion weighting (b&0) as anatomical
reference for offline motion correction: one image at the
beginning of the sequence and one after each block of ten
diffusion-weighted images. The interleaved measurement
of 72 axial slices with 1.7 mm thickness (no gap) covered
the entire brain. We improved the signal-to-noise ratio by
averaging 3 acquisitions of the same protocol, resulting in
an acquisition time of about 45 min.
Preprocessing and co-registration of T1-weighted
and diffusion-weighted MR images
We skull-stripped the T1-weighted scans and co-registered
them into Talairach space [58] together with the
T2-weighted images with rigid-body transformations [24],
implemented in FSL [25;http://www.fmrib.ox.ac.uk/fsl/].
We used the images without diffusion weighting to esti-
mate motion correction parameters with rigid-body trans-
formations computed as above. We combined motion
correction for the 180 diffusion-weighted images with an
additional global registration to the T1 anatomy. We cor-
rected the gradient direction for each volume with the
rotation parameters, interpolated the registered images to
the new reference frame with an isotropic voxel resolution
of 1 mm, and averaged the three corresponding acquisi-
tions and gradient directions. Finally, we fitted a diffusion
tensor to the data for each voxel and computed from the
tensor the principal diffusion direction and fractional
anisotropy.
Hypothalamus segmentation
We interactively traced the hypothalamus with FSLView
3.1 in the registered T1- and T2-weighted scans separately
in each subject’s left and right hemisphere with reference
to a high-resolution anatomical brain atlas [39]. Except for
some adjustments, we defined the relevant anatomical
landmarks and boundaries (Table 1) according to criteria
established by Goldstein et al. [20]. Since in our sample the
infundibular stalk was situated anterior to the tuberal
hypothalamus, we defined it as the inferior border of the
anterior hypothalamus, whereas the brain’s external surface
marked the inferior border of the tuberal hypothalamus.
Additionally, to improve reliability, we replaced rather
diffuse landmarks with more clearly defined boundaries
(e.g., lateral extent of optic chiasm instead of substantia
innominata). We traced the hypothalamus masks in coronal
slices and validated them in the horizontal and sagittal
plane. The superior, inferior, and lateral borders of the
masks were traced on T1-weighted images, whereas T2-
weighted images were used for delineating the third ven-
tricle. In general, we applied conservative decision criteria,
that is, in case of doubt, the voxel was classified as not
hypothalamic. To assess the degree of intrarater reliability,
the same investigator (F. P.) traced each hypothalamus
mask twice under identical experimental conditions
(monitor settings, ambient illumination) but in separate
tracing sessions.
For quantitative analysis, we determined the volume of
each hypothalamus mask in mm
3
(separately for each
hemisphere). We analyzed test–retest reliability by calcu-
lating Pearson’s correlation coefficient rbetween mask
volumes of the first and second tracing session and tested
for significant hemispheric differences in mask volumes
with a paired ttest (p\0.05 was considered significant).
Diffusion direction-based hypothalamus parcellation
We projected the hypothalamus masks onto the co-regis-
tered diffusion-weighted images, computed the similarity
of the principal diffusion direction for each pair of voxels
within the mask, and clustered the similarity matrix in three
regions using a k-means algorithm. The directional infor-
mation from the left and right hypothalamus masks were
combined assuming symmetry of the hypothalamus
microstructural directions with respect to the mid-sagittal
plane. In the analysis, we assembled the diffusion tensor
data from all voxels within the hypothalamus masks of all
subjects. The data from the right hypothalamus masks were
mirrored with respect to the symmetry plane. As similarity
measure of the diffusion directions, we computed the angle
between the principal diffusion directions of any pair of
voxels within the assembled dataset containing the left and
Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508 499
123
right hypothalamus masks of all participants. The angle
(ranging from 0 to p/2) was represented in a matrix
showing in each column the angular difference between a
specific voxel and all other voxels in one of the masks. This
multidimensional column vectors were provided as input
for the k-means clustering.
We specifically focused on a model of three clusters
based on the logic outlined in the introduction. We ren-
dered the clusters in 3-D and compared their topography
with data from the literature in order to assess each clus-
ter’s neuroanatomical validity.
Results
Segmentation of the hypothalamus mask
We were able to reliably segment the left and right hypo-
thalamus based on the predefined anatomical landmarks in
all 10 participants. The volumes of the hypothalamus
masks were normally distributed in both hemispheres
and tracing sessions (Table 2). We found male subjects to
have significantly greater left hypothalamic volumes
than females (males: 652.80 ±50.10 mm
3
; females:
561.80 ±6.58 mm
3
) but no right side difference (males:
656.60 ±46.30 mm
3
; females: 586.80 ±92.63 mm
3
). For
each hemisphere, we found a high degree of intrarater test–
retest reliability (r=0.86 left, r=0.85 right) and voxel
overlap (89.5 % left, 93.3 % right). The intraclass corre-
lation coefficient (ICC) to test–retest reliability (ICC) was
0.858 for the right and 0.917 for the left hemisphere.
A significant hemispheric difference in mask volumes
emerged in the second tracing session (p=0.002). It
disappeared again when the overlapping voxels (test \
retest) were tested for hemispheric difference (p=0.203,
Table 3).
Subdivision of the hypothalamus
Figure 1shows the hypothalamus mask outline of one
subject overlaid on T1-weighted images and direction-
encoded fractional anisotropy images. Figure 2depicts a
3-D rendering of the segmented mask of one subject. Each
mask voxel is color-coded according to its principal dif-
fusion direction. Even without clustering, the mask shows
a clear subdivision into three entities: an anteromedial
region with dorsoventral diffusion direction (blue), a
Table 1 Anatomical landmarks and boundaries of the hypothalamus
Superior Inferior Medial Lateral
Preoptic hypothalamus (coronal
level of anterior commissure
and optic chiasm)
Anterior commissure Optic chiasm Third ventricle Lateral extent of optic
chiasm
Anterior hypothalamus (coronal
level of interventricular
foramen)
Column of the fornix and more posterior
medial pole of the internal capsule
Junction of optic
tract and
infundibular
stalk
Third ventricle Genu of the internal
capsule, lateral edge of
optic tract
Tuberal hypothalamus
(anteroventral/medial
thalamus)
Medial pole of the genu of the internal
capsule
External surface Third ventricle Internal capsule, globus
pallidus
Posterior hypothalamus
(mamillary body)
White matter fibers (mamillothalamic
tract) above mamillary body, axial level
of anterior commissure
Hemispheric
margin
Third
ventricle,
hemispheric
midline
Globus pallidus
including substantia
nigra, cerebral
peduncle
Table 2 Descriptive statistics of the hypothalamus masks
Experimental condition Mean
volume
(n=10)
[mm
3
]
Standard
deviation
Test (1st session), left hemisphere 607.3 66.18
Test (1st session), right hemisphere 621.7 78.21
Retest (2nd session), left hemisphere 642.0 66.01
Retest (2nd session), right
hemisphere
610.9 69.52
Voxel overlap (test \retest),
left hemisphere
559.0 60.90
Voxel overlap (test \retest),
right hemisphere
575.3 70.06
Table 3 Hemispheric differences in mask volumes
Experimental condition Mean
difference
(n=10)
(mm
3
)
Tp
Test, left versus right hemisphere -14.4 -0.93 0.379
Retest, left versus right hemisphere 31.1 4.44 0.002
Voxel overlap, left versus right
hemisphere
-16.3 -1.37 0.203
500 Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508
123
posteromedial region with rostro-caudal direction (green),
and a lateral region with mediolateral direction (red). This
pattern becomes even more evident when a k-means clus-
tering algorithm is applied to each subject’s mask voxels.
We obtained a reproducible and anatomically coherent
arrangement of clusters across hemispheres and subjects. In
each mask, we found an anterior (blue), a posteromedial
(green), and a lateral (red) subdivision with consistent
principal diffusion directions across subjects (Fig. 3).
Discussion
The main finding of our study is that, by clustering a
landmark-based ROI of the hypothalamus in diffusion-
weighted MR scans, we obtained an arrangement of sub-
divisions that is reproducible and anatomically coherent
across hemispheres and subjects. The question now arises
how these clusters can be interpreted in terms of
hypothalamic subnuclei. Over the years, various parcella-
tions of the hypothalamus have been proposed, based on
structural criteria (e.g., cytoarchitecture, myeloarchitecture,
neurochemistry, fiber connections) and functional aspects
(e.g., microstimulation or lesion experiments). These
diverse classifications and nomenclatures can be summa-
rized under a common structural framework [45] that dis-
tinguishes four rostro-caudal levels: a preoptic, anterior,
tuberal, and mamillary region. Each of these regions can be
further subdivided into three mediolaterally arranged
zones: a periventricular zone close to the midline, a medial,
and a lateral zone. This ‘4 93 matrix’’ contains the fol-
lowing major hypothalamic subdivisions: Preoptic region:
periventricular zone (periventricular preoptic nucleus)—
medial zone (medial preoptic nucleus, #6 in Figs. 4,5)—
lateral zone (lateral preoptic nucleus, #12). Anterior region:
periventricular zone (paraventricular nucleus, #2 and sup-
rachiasmatic nucleus, #14)—medial zone (anterior hypo-
thalamic nucleus, #7)—lateral zone (lateral hypothalamic
Fig. 1 Outline of the
hypothalamus mask on coronal
(a) and horizontal (b) MR scans
of one subject: T1-weighted
images (top row), color-coded
fractional anisotropy (FA)
images (middle row),
T1-weighted images with
clustering result (bottom row).
Left ?right column:
rostral ?caudal level in A and
ventral ?dorsal level in B.
Colors in FA images and
clustering results correspond to
principal diffusion directions
(red medial–lateral, green
rostral-caudal, blue dorsal–
ventral). Dashed line indicates
midline of the brain (same
convention for all images).
Ccaudal, Ddorsal, Rrostral,
Vventral
Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508 501
123
area, #3 and supraoptic nucleus, #13). Tuberal region:
periventricular zone (infundibular nucleus, #15)—medial
zone (dorsomedial nucleus, #8 and ventromedial nucleus,
#9)—lateral zone (lateral hypothalamic area, #3). Mamil-
lary region: periventricular zone (no entry)—medial zone
(posterior hypothalamic nucleus, #4 and medial and lateral
mamillary nucleus, #11)—lateral zone (lateral hypotha-
lamic area, #3). Embedded in these gray matter nuclei are
several fiber systems, for example, the hypothalamo-
hypophyseal pathways, medial forebrain bundle, dorsal
longitudinal fascicle, fornix, and mamillothalamic tract
[45]. The hypothalamo-hypophyseal pathways arise from
neurons in the paraventricular, supraoptic, and infundibular
nucleus that give rise to descending axons that terminate in
the pituitary. The medial forebrain bundle is an arrange-
ment of loosely packed, mostly thin fibers that extend from
the septal region to the tegmentum of the midbrain. The
dorsal longitudinal fascicle runs in a periventricular posi-
tion from the posterior part of the hypothalamus to the
Fig. 2 3-D rendering of the principal diffusion direction (red medial–
lateral, green rostral-caudal, blue dorsal–ventral) of each mask voxel
of one subject (both hemispheres are shown, rostral-left-dorsal
direction of view)
Fig. 3 3-D rendering of the
hypothalamus parcellation into
an anterior (blue),
posteromedial (green), and
lateral (red) subdivision for six
subjects (both hemispheres are
shown, rostral-left-dorsal
direction of view, cf. Fig. 2).
Colors correspond to principal
diffusion directions (red
medial–lateral, green rostral-
caudal, blue dorsal–ventral)
502 Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508
123
caudal medulla oblongata. The fornix is a large fiber
bundle that extends from the hippocampus in an arch-like
fashion to the septal region and hypothalamus. The final
segment (column of the fornix, #1 in Figs. 4,5) descends
through the hypothalamus and terminates largely in the
mamillary body. The mamillothalamic tract contains part
of the efferents of the mamillary body. It terminates in the
anterior thalamic nucleus.
Table 4shows the efferents and afferents of each
hypothalamic subnucleus. Note that these data are based on
invasive anterograde and retrograde tract tracing studies in
laboratory animals which are still the gold standard for
mapping anatomical connections in the brain. The key
point is that the connections of all subnuclei form a spa-
tially largely overlapping (and not disparate!) network of
fibers running rostrally (e.g., to the septal nuclei), caudally
(e.g., to a wide variety of nuclei in the brain stem), medi-
ally (e.g., intrahypothalamic connections), laterally (e.g.,
to the amygdala), dorsally (e.g., to the thalamus and
neocortex), and ventrally (e.g., to the pituitary gland).
Figures 10.2 (on page 295) and 10.3 (on page 297) in
Nieuwenhuys et al. [45] provide a very clear graphical
presentation of this densely interconnecting and overlap-
ping network of afferent and efferent connections. It is
plausible to assume that the 3-D orientation of these fibers
leaving or entering a given subnucleus is also preserved
within the nucleus itself (where the cells of origin and
terminations, respectively, of these fibers are actually
located). This, however, means that structurally separating
(e.g., with a clustering algorithm) these subnuclei based on
solely direction-dependent aspects of the microstructure
(and this is the only information extracted by diffusion-
weighted imaging) is extremely challenging. The situation
is much easier elsewhere in the brain where adjacent brain
structures with disparate connectivity profiles exist. There,
this technique can produce very convincing results. A good
example is the supplementary motor area (SMA) proper
and area pre-SMA. Both areas lie next to each other on the
mesial aspect of the frontal lobe rostral to the primary
motor cortex. Invasive tract tracing studies in macaque
monkeys have shown that SMA proper and pre-SMA
markedly differ from each other in their cortical and sub-
cortical connections [38,41,49]. Not surprising that their
presumed homologs in the human mesial frontal cortex
(human SMA proper and pre-SMA) can be very clearly
separated from each other with a diffusion-weighted
imaging–based clustering approach [26,29,44]. In the
hypothalamus, however, with its largely overlapping fiber
architecture, it is quite unlikely that a 1:1 mapping of each
subnucleus will be feasible with diffusion-weighted imag-
ing alone in the near future—neither with our relatively
simplistic single-tensor model nor with more sophisticated
modeling approaches commonly referred to as high angular
resolution diffusion imaging [HARDI; cf. 28]. A promising
Fig. 4 Subnuclei of the
hypothalamus (reproduced from
[45], p. 292, Fig. 10.1, with
kind permission from Springer
Science ?Business Media
B.V.). Ccaudal, Ddorsal,
Rrostral, Vventral
Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508 503
123
way to improve this in the future would be to increase the
dimensionality of the mapping strategy, for example, by
combining diffusion-weighted imaging with MR sequences
that extract further quantitative information on gray matter
microstructure such as a modified magnetization-prepared
rapid gradient echo (so-called MP2RAGE) sequence [40]
that reflects structural differences in myeloarchitecture
[19]. It also has to be emphasized that a vulnerable co-
registration of MRI and DTI data could not be excluded by
the present method. Therefore, further studies should apply
combined DTI-based clustering and tractography to even
better discriminate the region of interest [8,37].
Despite these limitations, however, we managed to
cluster our hypothalamic ROI into an arrangement of
subdivisions that is (i) reproducible and (ii) anatomically
coherent across hemispheres and subjects, even if the
present study does not allow to determine the exact number
of hypothalamic subregions in humans.
Figure 5shows a schematic superimposition of the ROI
(borders marked in yellow) and its three subdivisions
(transparent polygons) onto the microanatomical map of
the hypothalamus shown in Fig. 4. Since the anatomical
map is drawn in a pseudo-3-D format, it can be combined
with the parcellation result both in a paramedian section
(i.e., a sagittal section close to the midline, Fig. 5, top
panel) and in a sagittal section further lateral (Fig. 5, bot-
tom panel). Combining our data and findings from Lemaire
et al. [36], we suggest the following interpretation: The
Fig. 5 Borders of the
hypothalamus mask (yellow
lines) and hypothalamic
subdivisions (blue,green, and
red compartments) in a
paramedian (top) and more
lateral (bottom) sagittal plane
schematically superimposed on
the anatomical drawing shown
in Fig. 4.Arrows mark principal
diffusion direction within each
subdivision. Ccaudal, Ddorsal,
Rrostral, Vventral
504 Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508
123
anterior cluster (blue with dorsoventral principal diffusion
direction) overlaps with the paraventricular, anterior, and
dorsomedial hypothalamic nucleus and partly with the
lateral hypothalamic area. A significant contribution
toward the dorsoventral principal diffusion direction comes
from major fiber tracts within this compartment that run
Table 4 Hypothalamic subnuclei: efferents and afferents (based on
tract tracing data from the animal literature, numbers in parentheses
refer to the corresponding structures in Figs. 4and 5, table rearranged
from [45])
Efferents to Afferents from
Medial preoptic nucleus (#6)
Lateral septal nucleus Lateral septal nucleus
Thalamus Amygdala
Zona incerta Periaqueductal gray
Periaqueductal gray Parabrachial nucleus
Ventral tegmental area Dorsal raphe nuclei
Dorsal raphe nuclei Noradrenergic brain stem cell
groups
Mesencephalic locomotor region
Lateral preoptic nucleus (#12)
Nucleus of the diagonal band Hippocampal subicular cortex
Dorsal raphe nuclei Bed nucleus of stria terminalis
Dorsal raphe nuclei
Paraventricular nucleus (#2)
Neurohypophysis Bed nucleus of stria terminalis
Parabrachial nucleus Parabrachial nucleus
Dorsal vagal complex Dorsal raphe nuclei
Locus coeruleus Noradrenergic brain stem cell
groups
Ventrolateral medullary reticular
formation
Adrenergic brain stem cell
groups
Spinal cord Nucleus of solitary tract
Ventrolateral medullary reticular
formation
Spinal nucleus of trigeminal
nerve
Suprachiasmatic nucleus (#14)
– Retina
Dorsal raphe nuclei
Spinal nucleus of trigeminal
nerve
Anterior hypothalamic nucleus (#7)
Lateral septal nucleus Neocortex
Thalamus Lateral septal nucleus
Ventrolateral medullary reticular
formation
Amygdala
Periaqueductal gray
Dorsal raphe nuclei
Lateral hypothalamic area (#3)
Neocortex Neocortex
Medial septal nucleus Hippocampal subicular cortex
Amygdala Amygdala
Nucleus of the diagonal band Bed nucleus of stria terminalis
Substantia innominata Nucleus accumbens
Thalamus Periaqueductal gray
Periaqueductal gray Parabrachial nucleus
Parabrachial nucleus Dorsal raphe nuclei
Table 4 continued
Efferents to Afferents from
Dorsal vagal complex Noradrenergic brain stem cell
groups
Dorsal raphe nuclei Nucleus of solitary tract
Rhombencephalic reticular
formation
Spinal nucleus of trigeminal
nerve
Ventrolateral medullary reticular
formation
Spinal cord
Supraoptic nucleus (#13)
Neurohypophysis Bed nucleus of stria terminalis
Noradrenergic brain stem cell
groups
Adrenergic brain stem cell
groups
Ventrolateral medullary reticular
formation
Infundibular nucleus (#15)
Median eminence
Dorsomedial nucleus (#8)
Periaqueductal gray Bed nucleus of stria terminalis
Ventral tegmental area Periaqueductal gray
Locus coeruleus Parabrachial nucleus
Spinal cord Nucleus of solitary tract
Spinal nucleus of trigeminal
nerve
Ventromedial nucleus (#9)
Lateral septal nucleus Neocortex
Amygdala Lateral septal nucleus
Bed nucleus of stria terminalis Amygdala
Substantia innominata Periaqueductal gray
Thalamus Parabrachial nucleus
Zona incerta
Periaqueductal gray
Mesencephalic reticular
formation
Posterior hypothalamic nucleus (#4)
Neocortex Neocortex
Hippocampus Periaqueductal gray
Dorsal vagal complex Parabrachial nucleus
Dorsal raphe nuclei Noradrenergic brain stem cell
groups
Ventrolateral medullary reticular
formation
Spinal nucleus of trigeminal
nerve
Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508 505
123
in the same direction: the hypothalamo-hypophyseal pro-
jection, column of the fornix, and parts of the stria termi-
nalis. The lateral cluster (red with mediolateral principal
diffusion direction) overlaps with the ventromedial and
supraoptic nucleus and partly with the lateral hypothalamic
area. Fiber tracts contributing to the mediolateral diffusion
direction are the supraoptic commissures and the sublen-
ticular system. The posteromedial cluster (green with ro-
stro-caudal principal diffusion direction) overlaps with the
suprachiasmatic, infundibular, ventromedial, posterior
hypothalamic, and medial and lateral mamillary nucleus.
Fiber tracts contributing to the rostro-caudal diffusion
direction are the medial forebrain bundle and dorsal lon-
gitudinal fascicle.
The findings of our study provide a good basis for fur-
ther neuropsychiatric research on structure and volume of
distinct diencephalic regions which have already been
shown to be of functional importance for affect regulation
[18,60]. Indirect findings of the neighboring third ventricle
already suggested a structural impairment of the hypo-
thalamus in mood disorders [4]. Seeking macrostructural
changes of the human hypothalamus, Bielau et al. [9]
determined postmortem volumes in patients with uni- and
bipolar depression. Group comparisons showed the stron-
gest effect (f=0.56; -15.5 %) for the hypothalamic
region in bipolar I disorder. The major depressive group
also showed a trend toward volume decrease of -9.5 %
(f=0.36, medium effect), but this was not significant due
to the small sample size. In contrast, the only two in vivo
studies of the human hypothalamus in mood disorders
published so far do not allow definitive conclusions to be
drawn due to inherent methodological problems. The first
study by Dupont et al. [17] defined a region of interest
encompassing the hypothalamic gray matter, mamillary
bodies, but also the septal nuclei. The second study by
Pinilla [46], using an exploratory whole-brain voxel-wise
analysis in healthy controls and patients with major
depression, claimed to have found a significant volume
decrease in the patient group with peak differences in the
hypothalamus (coordinates (x,y,z)-6, 2, -9 in the left
and 9, 2, -9 in the right hemisphere). A reanalysis of the
data (reviewed in [52]), however, found the peak difference
to lie in the white matter with the lentiform nucleus as the
nearest gray matter structure.
Volumetric changes in the hypothalamus exist in affective
disorders, but it is unclear whether these abnormalities are
restricted to specific hypothalamic subregions since, up to
now, a fine-grained morphological analysis of the hypothal-
amus in vivo has not been feasible. However, in patients with
mood disorders, distinct neurocognitive dysfunctions such as
dysregulation of vigilance or wakefulness may be hypothe-
sized to correlate with the volume of specific hypothalamic
subdivisions [22,55]. This assumption is based on previous
work on impaired orexinergic neurotransmission which is
restricted to the lateral hypothalamus as demonstrated in
narcolepsy. Since disruption of the sleep and wake cycle is a
potential major feature of mood disorders, future studies are
warranted. Moreover, volume changes in hypothalamic sub-
divisions may be investigated with respect to neurochemical
markers such as cerebrospinal fluid orexin levels, recently
studied in patients with manic episodes [53,54].
In summary, thisis to our best knowledge the first study that
attempts a delineation of hypothalamic subdivisions by clus-
tering diffusion-weighted MRI data. Although the definition
of the hypothalamus mask is subjective, a high degree of
reliability can be achieved by adhering to a carefully selected
list of predefined anatomical landmarks. Once applied in
a larger sample of neuropsychiatric patients, a structural
analysis of the hypothalamus should contribute to a better
understanding of the pathogenesis of affective disorders.
Acknowledgments Part of thiswork was supported by theFET project
CONNECT of the European Union (http://www.brain-connect.eu).
Conflict of interest None.
References
1. Alsop DC (1997) The sensitivity of low flip angle RARE imag-
ing. Magn Reson Med 37:176–184
2. Bao AM, Meynen G, Swaab DF (2008) The stress system in
depression and neurodegeneration: focus on the human hypo-
thalamus. Brain Res Rev 57:531–553
3. Baroncini M, Jissendi P, Balland E, Besson P, Pruvo JP, Francke
JP, Dewailly D, Blond S, Prevot V (2012) MRI atlas of the
human hypothalamus. Neuroimage 59:168–180
4. Baumann B, Bogerts B (2001) Neuroanatomical studies on
bipolar disorder. Br J Psychiatry 178:S142–S147
5. Beaulieu C (2002) The basis of anisotropic water diffusion in the
nervous system: a technical review. NMR Biomed 15:435–455
6. Bernstein HG, Keilhoff G, Steiner J, Dobrowolny H, Bogerts B
(2010) The hypothalamus in schizophrenia research: no longer a
wallflower existence. Open Neuroendocrinol J 3:59–67
7. Bernstein HG, Klix M, Dobrowolny H, Brisch R, Steiner J,
Bielau H, Gos T, Bogerts B (2012) A postmortem assessment of
mammillary body volume, neuronal number and densities, and
fornix volume in subjects with mood disorders. Eur Arch Psy-
chiatry Clin Neurosci 262:637–646
8. Besseling RMH, Jansen JFA, Overvliet GM, Vaessen MJ, Bra-
akman HMH, Hofman PAM, Aldenkamp AP, Backes WH (2012)
Tract specific reproducibility of tractography based morphology
and diffusion metrics. PLoS One 7:e34125
9. Bielau H, Tru
¨bner K, Krell D, Agelink MW, Bernstein HG, Stauch R,
Mawrin C, Danos P, Gerhard L, Bogerts B, Baumann B (2005)
Volume deficits of subcortical nuclei in mood disorders: a postmortem
study. Eur Arch Psychiatry Clin Neurosci 255:401–412
10. Braak H, Braak E (1987) The hypothalamus of the human adult:
chiasmatic region. Anat Embryol (Berl) 175:315–330
11. Braak H, Braak E (1992) Anatomy of the human hypothalamus
(chiasmatic and tuberal region). In: Swaab DF, Hofman MA,
Mirmiran M, Ravid R, Van Leeuwen FW (eds) Progress in brain
research. Elsevier, Amsterdam, pp 3–16
506 Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508
123
12. Busse S, Bernstein HG, Busse M, Bielau H, Brisch R, Mawrin C,
Mu
¨ller S, Sarnyai Z, Gos T, Bogerts B, Steiner J (2012) Reduced
density of hypothalamic VGF-immunoreactive neurons in
schizophrenia: a potential link to impaired growth factor signal-
ing and energy homeostasis. Eur Arch Psychiatry Clin Neurosci
262:365–374
13. Crosby EC, Woodburne RT (1939) The comparative anatomy of
the preoptic area and the hypothalamus. Res Publ Assoc Res Nerv
Ment Dis 20:52–169
14. Doran M, Hajnal JV, Van Bruggen N, King MD, Young IR,
Bydder GM (1990) Normal and abnormal white matter tracts
shown by MR imaging using directional diffusion weighted
sequences. J Comput Assist Tomogr 14:865–873
15. Draganski B, Geisler P, Hajak G, Schuierer G, Bogdahn U,
Winkler J, May A (2002) Hypothalamic gray matter changes in
narcoleptic patients. Nat Med 8:1186–1188
16. Drevets WC, Price JL, Furey ML (2008) Brain structural and
functional abnormalities in mood disorders: implications for
neurocircuitry models of depression. Brain Struct Funct 213:93–
118
17. Dupont RM, Jernigan TL, Heindel W, Butters N, Shafer K,
Wilson T, Hesselink J, Gillin JC (1995) Magnetic resonance
imaging and mood disorders: localization of white matter and
other subcortical abnormalities. Arch Gen Psychiatry 52:747–755
18. Georgiadis JR, Farrell MJ, Boessen R, Denton DA, Gavrilescu M,
Kortekaas R, Renken RJ, Hoogduin JM, Egan GF (2010)
Dynamic subcortical blood flow during male sexual activity with
ecological validity: a perfusion fMRI study. Neuroimage 50:208–
216
19. Geyer S, Weiss M, Reimann K, Lohmann G, Turner R (2011)
Microstructural parcellation of the human cerebral cortex: from
Brodmann’s post-mortem map to in vivo mapping with high-field
magnetic resonance imaging. Front Hum Neurosci 5:19
20. Goldstein JM, Seidman LJ, Makris N, Ahern T, O’Brien LM,
Caviness VS, Kennedy DN, Faraone SV, Tsuang MT (2007)
Hypothalamic abnormalities in schizophrenia: sex effects and
genetic vulnerability. Biol Psychiatry 61:935–945
21. Griswold MA, Jakob PM, Heidemann RM, Nittka M, Jellus V,
Wang J, Kiefer B, Haase A (2002) Generalized autocalibrating
partially parallel acquisitions (GRAPPA). Magn Reson Med
47:1202–1210
22. Hegerl U, Wilk K, Olbrich S, Schoenknecht P, Sander C (2012)
Hyperstable regulation of vigilance in patients with major
depressive disorder. World J Biol Psychiatry 13:436–446
23. Hofman MA, Swaab DF (1992) The human hypothalamus: com-
parative morphometry and photoperiodic influences. In: Swaab DF,
Hofman MA, Mirmiran M, Ravid R, Van Leeuwen FW (eds)
Progress in brain research. Elsevier, Amsterdam, pp 133–149
24. Jenkinson M, Bannister P, Brady M, Smith S (2002) Improved
optimization for the robust and accurate linear registration and
motion correction of brain images. Neuroimage 17:825–841
25. Jenkinson M, Beckmann CF, Behrens TEJ, Woolrich MW, Smith
SM (2012) FSL. Neuroimage 62:782–790
26. Johansen-Berg H, Behrens TEJ, Robson MD, Drobnjak I, Rush-
worth MFS, Brady JM, Smith SM, Higham DJ, Matthews PM
(2004) Changes in connectivity profiles define functionally dis-
tinct regions in human medial frontal cortex. Proc Natl Acad Sci
USA 101:13335–13340
27. Johansen-Berg H, Behrens TEJ, Sillery E, Ciccarelli O,
Thompson AJ, Smith SM, Matthews PM (2005) Functional-
anatomical validation and individual variation of diffusion trac-
tography-based segmentation of the human thalamus. Cereb
Cortex 15:31–39
28. Jones DK, Kno
¨sche TR, Turner R (2012) White matter integrity,
fiber count, and other fallacies: the do’s and don’ts of diffusion
MRI. Neuroimage. doi:10.1016/j.neuroimage.2012.06.081
29. Klein JC, Behrens TEJ, Robson MD, Mackay CE, Higham DJ,
Johansen-Berg H (2007) Connectivity-based parcellation of
human cortex using diffusion MRI: establishing reproducibility,
validity and observer independence in BA 44/45 and SMA/pre-
SMA. Neuroimage 34:204–211
30. Koutcherov Y, Mai JK, Paxinos G (2003) Hypothalamus of the
human fetus. J Chem Neuroanat 26:253–270
31. Kwon HG, Byun WM, Ahn SH, Son SM, Jang SH (2011) The ana-
tomical characteristics of the stria terminalis in the human brain: a
diffusion tensor tractography study. Neurosci Lett 500:99–102
32. Le Bihan D (2003) Looking into the functional architecture of the
brain with diffusion MRI. Nat Rev Neurosci 4:469–480
33. Le Gros Clark WE (1936) The topography and homologies of the
hypothalamic nuclei in man. J Anat 70:203–214
34. Le Gros Clark WE (1938) Morphological aspects of the hypo-
thalamus. In: Le Gros Clark WE, Beattie J, Riddoch G, Dott NM
(eds) The hypothalamus: morphological, functional, clinical and
surgical aspects. Oliver and Boyd, Edinburgh, pp 1–68
35. Le Roux P, Hinks RS (1993) Stabilization of echo amplitudes in
FSE sequences. Magn Reson Med 30:183–190
36. Lemaire JJ, Frew AJ, McArthur D, Gorgulho AA, Alger JR,
Salomon N, Chen C, Behnke EJ, De Salles AAF (2011)
White matter connectivity of human hypothalamus. Brain Res
1371:43–64
37. Li M, He HG, Shi W, Li J, Lv B, Wang CH, Miao QW, Wang
ZC, Wang NL, Walter M, Sabel BA (2012) Quantification of the
human lateral geniculate nucleus in vivo using MR imaging
based on morphometry: volume loss with age. Am J Neuroradiol
33:915–921
38. Luppino G, Matelli M, Camarda R, Rizzolatti G (1993) Corti-
cocortical connections of area F3 (SMA-proper) and area F6 (pre-
SMA) in the macaque monkey. J Comp Neurol 338:114–140
39. Mai JK, Paxinos G, Voss T (2008) Atlas of the human brain, 3rd
edn. Elsevier, Academic Press, Amsterdam
40. Marques JP, Kober T, Krueger G, van der Zwaag W, van de
Moortele PF, Gruetter R (2010) MP2RAGE, a self bias-field
corrected sequence for improved segmentation and T
1
-mapping
at high field. Neuroimage 49:1271–1281
41. Matelli M, Luppino G (1996) Thalamic input to mesial and
superior area 6 in the macaque monkey. J Comp Neurol
372:59–87
42. Moseley ME, Cohen Y, Kucharczyk J, Mintorovitch J, Asgari
HS, Wendland MF, Tsuruda J, Norman D (1990) Diffusion-
weighted MR imaging of anisotropic water diffusion in cat cen-
tral nervous system. Radiology 176:439–445
43. Mugler JP, Brookeman JR (1990) Three-dimensional magneti-
zation-prepared rapid gradient-echo imaging (3D MP RAGE).
Magn Reson Med 15:152–157
44. Nanetti L, Cerliani L, Gazzola V, Renken R, Keysers C (2009)
Group analyses of connectivity-based cortical parcellation using
repeated k-means clustering. Neuroimage 47:1666–1677
45. Nieuwenhuys R, Voogd J, van Huijzen C (2008) The human
central nervous system, 4th edn. Springer, Berlin
46. Pinilla BP (2009) Auswirkungen der unipolar depressiven
Sto
¨rung auf strukturelle Gehirnvera
¨nderungen in der Voxel-
based-NMR-Morphometry und auf ‘‘hippocampusspezifische’’
kognitive Leistungen. Dissertation, Charite
´, Berlin
47. Price JL, Drevets WC (2010) Neurocircuitry of mood disorders.
Neuropsychopharmacology 35:192–216
48. Reese TG, Heid O, Weisskoff RM, Wedeen VJ (2003) Reduction
of eddy-current-induced distortion in diffusion MRI using a
twice-refocused spin echo. Magn Reson Med 49:177–182
49. Rizzolatti G, Luppino G, Matelli M (1996) The classic supple-
mentary motor area is formed by two independent areas. In:
Lu
¨ders HO (ed) Supplementary sensorimotor area. Lippincott-
Raven, Philadelphia, pp 45–56
Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508 507
123
50. Saeki N, Kansaku K, Higuchi Y, Kawano K, Iijima T, Inoue N,
Yamaura A (2001) Demonstration of the post commissural fibres
of the fornix in short-inversion time inversion-recovery imaging
on a high-field system. Neuroradiology 43:547–550
51. Saper CB (2012) Hypothalamus. In: Mai JK, Paxinos G (eds) The
human nervous system, 3rd edn. Elsevier, Academic Press,
Amsterdam, pp 548–583
52. Schindler S, Geyer S, Strauß M, Anwander A, Hegerl U, Turner
R, Scho
¨nknecht P (2012) Structural studies of the hypothalamus
and its nuclei in mood disorders. Psychiatry Res 201:1–9
53. Schmidt FM, Bru
¨gel M, Kratzsch J, Strauß M, Sander C, Baum P,
Thiery J, Hegerl U, Scho
¨nknecht P (2010) Cerebrospinal fluid
hypocretin-1 (orexin A) levels in mania compared to unipolar
depression and healthy controls. Neurosci Lett 483:20–22
54. Schmidt FM, Arendt E, Steinmetzer A, Bruegel M, Kratzsch J,
Strauß M, Baum P, Hegerl U, Scho
¨nknecht P (2011) CSF-
hypocretin-1 levels in patients with major depressive disorder
compared to healthy controls. Psychiatry Res 190:240–243
55. Schoenknecht P, Olbrich S, Sander C, Spindler P, Hegerl U
(2010) Treatment of acute mania with modafinil monotherapy.
Biol Psychiatry 67:e55–e57
56. Solano-Castiella E, Anwander A, Lohmann G, Weiss M,
Docherty C, Geyer S, Reimer E, Friederici AD, Turner R (2010)
Diffusion tensor imaging segments the human amygdala in vivo.
Neuroimage 49:2958–2965
57. Stephan H, Frahm H, Baron G (1981) New and revised data on
volumes of brain structures in insectivores and primates. Folia
Primatol (Basel) 35:1–29
58. Talairach J, Tournoux P (1988) Co-planar stereotaxic atlas of the
human brain. 3-Dimensional proportional system: an approach to
cerebral imaging. Thieme, Stuttgart
59. Toni R, Malaguti A, Benfenati F, Martini L (2004) The human
hypothalamus: a morpho-functional perspective. J Endocrinol
Invest 27:73–94
60. Walter M, Bermpohl F, Mouras H, Schiltz K, Tempelmann C,
Rotte M, Heinze HJ, Bogerts B, Northoff G (2008) Distinguish-
ing specific sexual and general emotional effects in fMRI: sub-
cortical and cortical arousal during erotic picture viewing.
Neuroimage 40:1482–1494
61. Wiegell MR, Tuch DS, Larsson HBW, Wedeen VJ (2003)
Automatic segmentation of thalamic nuclei from diffusion tensor
magnetic resonance imaging. Neuroimage 19:391–401
62. Young JK, Stanton GB (1994) A three-dimensional reconstruc-
tion of the human hypothalamus. Brain Res Bull 35:323–327
508 Eur Arch Psychiatry Clin Neurosci (2013) 263:497–508
123
... However, sample sizes were small and applying fixed ROIs could be misleading due to partial volume effects and the heterogenous appearance of the hypothalamus. In addition, the direction of effects was partly contradictory and a limited resolution and multiple sources of image artifacts limit interpretability 14,15,[17][18][19] . ...
... Reliability analysis revealed acceptable to excellent intra-rater and inter-rater reliabilities of hypothalamus volume and spatial overlap of resulting masks using this method. This highlights the sensitivity and specificity of our procedure and compares to previous high-quality segmentation protocols implemented in smaller sample sizes 13,19,36 . ...
Article
Full-text available
Animal studies suggest that obesity-related diets induce structural changes in the hypothalamus, a key brain area involved in energy homeostasis. Whether this translates to humans is however largely unknown. Using a novel multimodal approach with manual segmentation, we here show that a higher body mass index (BMI) selectively predicted higher proton diffusivity within the hypothalamus, indicative of compromised microstructure in the underlying tissue, in a well-characterized population-based cohort (n1 = 338, 48% females, age 21–78 years, BMI 18–43 kg/m²). Results were independent from confounders and confirmed in another independent sample (n2 = 236). In addition, while hypothalamic volume was not associated with obesity, we identified a sexual dimorphism and larger hypothalamic volumes in the left compared to the right hemisphere. Using two large samples of the general population, we showed that a higher BMI specifically relates to altered microstructure in the hypothalamus, independent from confounders such as age, sex and obesity-associated co-morbidities. This points to persisting microstructural changes in a key regulatory area of energy homeostasis occurring with excessive weight. Our findings may help to better understand the pathomechanisms of obesity and other eating-related disorders.
... However, sample sizes were small and applying fixed ROIs could be misleading due to partial volume effects and the heterogenous appearance of the hypothalamus. In addition, the direction of effects was partly contradictory and a limited resolution and multiple sources of image artifacts limit interpretability 14,15,[17][18][19] . ...
... Reliability analysis revealed acceptable to excellent intra-rater and inter-rater reliabilities of hypothalamus volume and spatial overlap of resulting masks using this method. This highlights the sensitivity and specificity of our procedure and compares to previous high-quality segmentation protocols implemented in smaller sample sizes 13,19,36 . ...
Preprint
Full-text available
Obesity is a multifactorial disorder driven by sustained energy imbalance. The hypothalamus is an important regulator of energy homeostasis and therefore likely involved in obesity pathophysiology. Animal studies suggest that obesity-related diets induce structural changes in the hypothalamus through inflammation-like processes. Whether this translates to humans is however largely unknown. Therefore, we aimed to assess obesity-related differences in hypothalamic macro- and microstructure based on a multimodal approach using T1-weighted and diffusion-weighted magnetic resonance imaging (MRI) acquired at 3 Tesla in a large well-characterized sample of the Leipzig Research Center for Civilization Diseases (LIFE) cohort ( n 1 = 338, 48% females, age 21-78 years, BMI 18-43 kg/m ² ). We found that higher body mass index (BMI) selectively predicted higher mean proton diffusivity (MD) within the hypothalamus, indicative of compromised microstructure in the underlying tissue. Results were independent from confounders and confirmed in another independent sample (n 2 = 236). In addition, while hypothalamic volume was not associated with obesity, we identified a sexual dimorphism and larger hypothalamic volumes in the left compared to the right hemisphere. Using two large samples of the general population, we showed that a higher BMI specifically relates to altered microstructure in the hypothalamus, independent from confounders such as age, sex and obesity-associated co-morbidities. This points to persisting microstructural changes in a key regulatory area of energy homeostasis occurring with excessive weight. These findings may help to better understand the pathomechanisms of obesity and other eating-related disorders.
... Figure 5 summarizes the structural extra-hypothalamic networks of food intake discussed above. Using color coded directional fractional anisotropy (FA) maps of the HT, three different subregions of the HT were defined: an anteromedial region with dorsoventral diffusion direction, a posteromedial region with rostro-caudal direction, and a lateral region with mediolateral direction (86). The anteromedial region was assigned to the PVN, anterior, and dorsomedial hypothalamic nucleus and part of the LHA, the lateral region to the VMH and supraoptic nucleus and finally the posteromedial region to the suprachiasmatic, infundibular, VMH, posterior hypothalamic, medial and lateral mammillary nucleus. ...
Article
Full-text available
Hypothalamus (HT), this small structure often perceived through the prism of neuroimaging as morphologically and functionally homogeneous, plays a key role in the primitive act of feeding. The current paper aims at reviewing the contribution of magnetic resonance imaging (MRI) in the study of the role of the HT in food intake regulation. It focuses on the different MRI techniques that have been used to describe structurally and functionally the Human HT. The latest advances in HT parcellation as well as perspectives in this field are presented. The value of MRI in the study of eating disorders such as anorexia nervosa (AN) and obesity are also highlighted.
... Unfortunately, resting-state functional connectivity studies with standard whole-brain fMRI protocols cannot target specific nuclei of the hypothalamus due to the relatively low spatial resolution with whole brain coverage and the spatial smoothing required during the preprocessing of the data. Moreover, there are no clear anatomical borders between these nuclei using standard structural MRI, and the spatial resolution of most fMRI studies is limited to 2-3 mm, while the human hypothalamus measures around 700 mm 3 in total volume with many specialized cell groups (Saper, 1990;Schindler et al., 2013;Schonknecht et al., 2013). For this reason, it is difficult to target different nuclei unequivocally and reliably in a whole brain fMRI protocol implemented in RSFC analysis. ...
Chapter
Communication pathways of the hypothalamus with other brain regions and the periphery are critical to successfully control key physiological and psychological processes. With advanced functional magnetic resonance imaging (fMRI) techniques, it is possible to target hypothalamic function and infer discrete hypothalamus networks. Resting-state functional connectivity (RSFC) is a promising tool to study the functional organization of the brain and may act as a marker of individual differences and dysfunctions. Based on recent fMRI findings, the hypothalamus is mostly connected to parts of the striatum, midbrain, thalamus, insula, frontal, cingulate, and temporal cortices and the cerebellum. There is a strong interplay of the hypothalamus with these regions in response to different metabolic, hormonal, and nutritional states. In a state of hunger, hypothalamus RSFC increases with a strong shift to reward-related brain regions, especially in person with excessive weight. Nutrient signals and hormones, as insulin, act on these same connections conveying reward and internal signals to regulate homeostatic control. Moreover, dysfunctional hypothalamus communication has been documented in persons with neurological and psychiatric diseases. The results implicate that patients with depression, epilepsy, and neurodegenerative diseases show mostly a reduction in hypothalamus RSFC, whereas patients with migraine and headache display predominantly increased hypothalamus RSFC. The extent of these changes and regions affected depend on the disorder and symptom severity. Whether hypothalamus RSFC can serve as a marker for disease states or is a prodromal neurobiological feature still needs to be investigated.
... During the past years, advances in MR imaging enabled the development of methods for automated parcellations using local diffusion properties. A first approach to automated hypothalamic parcellation based on diffusion-weighted imaging (DWI) was introduced by Schönknecht et al. 16 . The authors applied k-means clustering on the principal diffusion direction information in each voxel of a manually segmented hypothalamus mask. ...
Article
Full-text available
The hypothalamus is a small, yet highly versatile structure mainly involved in bodily functions such as control of food intake and endocrine activity. Functional anatomy of different hypothalamic areas is mainly investigated using structural MRI, validated by ex-vivo histological studies. Based on diffusion-weighted imaging (DWI), recent automated clustering methods provide robust tools for parcellation. Using data of 100 healthy adults provided by the Human Connectome Project Database, we applied DWI-based automated clustering to the hypothalamus and related microstructural properties in these hypothalamic compartments to obesity. Our results suggest that the hypothalamus can be reliably partitioned into four clusters in each hemisphere using diffusion-based parcellation. These correspond to an anterior-superior, anterior-inferior, intermediate, and posterior cluster. Obesity was predicted by mean diffusivity of the anterior-superior cluster, suggesting altered inhibition of food intake. The proposed method provides an automated hypothalamic parcellation technique based on DWI data to explore anatomy and function of hypothalamic subunits in vivo in humans.
Article
Full-text available
Since the hypothalamus is involved in many neuroendocrine, metabolic, and affective disorders, detailed hypothalamic imaging has become of major interest to better characterize disease-induced tissue damages and abnormalities. Still, image contrast of conventional anatomical magnetic resonance imaging lacks morphological detail, thus complicating complete and precise segmentation of the hypothalamus. The hypothalamus' position lateral to the third ventricle and close proximity to white matter tracts including the optic tract, fornix, and mammillothalamic tract display one of the remaining shortcomings of hypothalamic segmentation, as reliable exclusion of white matter is not yet possible. Recent studies found that quantitative magnetic resonance imaging (qMRI), a method to create maps of different standardized tissue contents, improved segmentation of cortical and subcortical brain regions. So far, this has not been tested for the hypothalamus. Therefore, in this study, we investigated the usability of qMRI and diffusion MRI for the purpose of detailed and reproducible manual segmentation of the hypothalamus and data-driven white matter extraction and compared our results to recent state-of-the-art segmentations. Our results show that qMRI presents good contrast for delineation of the hypothalamus and white matter, and that the properties of these images differ between subunits, such that they can be used to reliably exclude white matter from hypothalamic tissue. We propose that qMRI poses a useful addition to detailed hypothalamic segmentation and volumetry.
Preprint
Full-text available
Since the hypothalamus is involved in many neuroendocrine, metabolic, and affective disorders, detailed hypothalamic imaging has become of major interest to better characterize disease-induced tissue damages and abnormalities. Still, image contrast of conventional anatomical magnetic resonance imaging lacks morphological detail, thus complicating complete and precise segmentation of the hypothalamus. The hypothalamus’ position lateral to the third ventricle and close proximity to white matter tracts including the optic tract, fornix, and mammillothalamic tract display one of the remaining shortcomings of hypothalamic segmentation, as reliable exclusion of white matter is not yet possible. Recent studies found that quantitative magnetic resonance imaging (qMRI), a method to create maps of different standardized tissue contents, improved segmentation of cortical and subcortical brain regions (Carey et al., 2017). So far, this has not been tested for the hypothalamus. Therefore, in this study, we investigated the usability of qMRI and diffusion MRI for the purpose of detailed manual segmentation and data-driven parcellation of the hypothalamus and compared our results to recent state-of-the-art segmentations. Our results show that qMRI and diffusion parameters indeed differ between hypothalamic subunits, and that qMRI is helpful for hypothalamic segmentation. In addition, we provide a data-driven clustering algorithm to reliably exclude white matter from hypothalamic tissue. We propose that qMRI poses a useful addition to detailed hypothalamic segmentation and volumetry.
Article
Full-text available
Despite their critical roles in autonomic functions, individual hypothalamic nuclei have not been extensively investigated in humans using functional magnetic resonance imaging, partly due to the difficulty in resolving individual nuclei contained in the small structure of the hypothalamus. Areal parcellation analyses enable discrimination of individual hypothalamic nuclei but require a higher spatial resolution, which necessitates long scanning time or large amounts of data to compensate for the low signal-to-noise ratio in 3T or 1.5T scanners. In this study, we present analytic procedures to estimate likely locations of individual nuclei in the standard 2-mm resolution based on our higher resolution dataset. The spatial profiles of functional connectivity with the cerebral cortex for each nucleus in the medial hypothalamus were calculated using our higher resolution dataset. Voxels in the hypothalamus in standard resolution images from the Human Connectome Project (HCP) database that predominantly shared connectivity profiles with the same nucleus were subsequently identified. Voxels representing individual nuclei, as identified with the analytic procedures, were reproducible across 20 HCP datasets of 20 subjects each. Furthermore, the identified voxels were spatially separate. These results suggest that these analytic procedures are capable of refining voxels that represent individual hypothalamic nuclei in standard resolution. Our results highlight the potential utility of these procedures in various settings such as patient studies, where lengthy scans are infeasible.
Article
The hypothalamus consists of numerous nuclei, and is regarded as the highest center for various autonomic functions. Although each hypothalamic nucleus implements a distinct function, it remains difficult to investigate the human hypothalamus at the nucleus level. In the present high-resolution functional MRI study, we utilized areal parcellation to discriminate individual nuclei in the human hypothalamus based on areal profiles of resting-state functional connectivity. The areal parcellation detected ten foci that were expected to represent hypothalamic nuclei, and the locations of the foci were consistent with those of the hypothalamic nuclei identified in previous histological studies. Regions of interest (ROI) analyses revealed contrasting brain activity changes following glucose ingestion: decrease in the ventromedial hypothalamic nucleus and increase in the lateral hypothalamic area in parallel with blood glucose increase. Moreover, decreased brain activity in the arcuate nucleus predicted future elevation of blood insulin during the first 10 min after glucose ingestion. These results suggest that the hypothalamic nuclei can putatively be determined using areal parcellation, and that the ROI analysis of the human hypothalamic nuclei is useful for future scientific and clinical investigations into the autonomic functions.
Article
Background: Impairment of sleep-wake cycles and circadian rhythm are found in human narcolepsy which is characterized by deficiency of hypocretin (hcrt) or its receptors. A disturbed electroencephalography (EEG) based vigilance regulation is also found in affective disorders such as major depressive disorder (MDD) and mania. For the first time, in the present study hcrt levels were investigated in patients with a manic episode and compared with age-matched patients with MDD and controls. Methods: 15 subjects were enrolled in the study after admission to hospital: 5 manic (mean YMRS 15.6 ± 2.9) and 5 age- matched patients with MDD (mean HDRS 11.6 ± 8.0), and 5 age-matched controls without any neurological or psychiatric disorder. Cerebrospinal fluid (CSF) hcrt levels were measured in all three groups using a fluorescence immunoassay (FIA). Results: Mean hcrt-1 level in manic patients (77.3 ± 20.7 pg/ml) did not differ significantly compared to patients with MDD (75.6 ± 15.7 pg/ml MDD) or controls (74.9 ± 19.3 pg/ml). Hcrt-levels and severity of disease did not show a significant association. Conclusion: In the present study, for the first time hcrt-1 levels in manic patients were investigated but did not reveal significant differences neither compared to age-matched patients with MDD nor healthy controls without any psychiatric or neurological disorder. Schmidt FM et al., Neuroscience Letters 2010; 483: 20–22
Article
The classical CPMG sequence and its extension as an imaging sequence, fast spin echo (FSE, based on RARE), suffer from signal magnitude variations in the early echoes when the refocusing pulses are not set exactly to 180-degrees. It has been suggested that by varying the value of the nutation angle of each refocusing pulse the signal magnitude could be made constant. This article describes an algorithm permitting the generation of sequences of nutation angles yielding series of echoes with constant signal magnitudes. This result is then used to design selective pulses for the FSE imaging technique.
Article
Modern neurosurgical concepts call for not only "seeing" but also for "localizing" structures in three-dimensional space in relationship to each other. Hence there is a need for a reference system. This book aims to put this notion into practice by means of anatomical and MRI sections with the same stereotaxic orientation. The purpose is to display the fundamental distribution of structures in three-dimensional space and their spatial evolution within the brain as a whole, while facilitating their identification; to make comparative studies of cortico-subcortical lesions possible on a basis of an equivalent reference system; to exploit the anatomo-functional data such as those furnished by SEEG in epilepsy and to enable the localization of special regions such as the SMA in three-dimensional space; and to apply the anatomical correlations of this reference system to neurophysiological investigations lacking sufficient anatomical back-up (including PET scan).
Article
Water diffusion magnetic resonance imaging (dMRI) allows tissue structure to be probed and imaged on a microscopic scale, providing unique clues to the fine architecture of neural tissues and to changes associated with various physiological and pathological states, such as acute brain ischaemia. Because diffusion is anisotropic in brain white matter, reflecting its organization in bundles of fibres running in parallel, dMRI can also be used to map the orientation in space of the white matter tracks in the brain, opening a new window on brain connectivity and brain maturation studies. This article provides an introduction to the key physical concepts that underlie dMRI, and reviews its potential applications in the neurosciences and associated clinical fields.
Article
This chapter focuses on the functional organization of the hypothalamus. The functions of the hypothalamus are so basic that its plan and organizational patterns have been, to a great degree, conserved throughout the mammalian spectrum. As a result, even when data in humans or monkeys are still sketchy, it is useful to view in a functional context that is derived in large part from other sources. The functional knowledge about the hypothalamus has relied on morphological approaches because of the extraordinarily small size and functional complexity of this region. The hypothalamus contains the integrative systems that support life, including such activities as fluid and electrolyte balance, food ingestion and energy metabolism, thermoregulation and immune response, and, of course, emotional expression and reproduction. Furthermore, the chapter reviews the overall plan of the cytoarchitecture and the major fiber pathways of the hypothalamus. This is followed by the functional organization of the hypothalamus.
Article
Background: Recent reports in the literature document an association between focal white matter abnormalities in bipolar as well as unipolar mood disorder. The importance of this finding and other associated anatomic differences is uncertain. Methods: We examined the volume of abnormal white matter and other brain volumes using quantitative magnetic resonance imaging analysis. We explored the relationship of these variables with diagnosis, cognitive function, and clinical variables in 36 patients with bipolar disorder, 30 patients with unipolar disorder, and 26 control subjects who were free from significant medical and neurologic illness. Results: Younger patients with bipolar disorder (but not similarly aged patients with unipolar disorder or controls) have an increased volume of abnormal white matter. Data also indicate that the total volume of abnormal white matter may be associated with increased cognitive impairment, increased rate of psychiatric illness in the family, and onset after adolescence. Conclusion: Patients with bipolar disorder demonstrate a pattern of subcortical brain morphologic abnormalities and cognitive impairment.