in
at
le
pat
r,†
ers
Un
10,
An
mal
prov
(MN
463–
mai
land
volu
proc
war
cou
then
mm
sing
assi
three procedures to perform the automated anatom-
ical labeling of functional studies are proposed: (1)
labe
nate
the
coor
to e
ter.
Map
ton,
as a
com
men
area
lair
kno
ee
ase
sti
O
ie
ee
uce
ns
al
ati
gi
from different individuals. This can be achieved using,
in particular, the procedure implemented in the Sta-
1
T
dress
levar
470 2
NeuroImage 15, 273–289 (2002)
doi:1 omling of an extremum defined by a set of coordi-
s, (2) percentage of voxels belonging to each of
AVOI intersected by a sphere centered by a set of
dinates, and (3) percentage of voxels belonging
ach of the AVOI intersected by an activated clus-
An interface with the Statistical Parametric
ping package (SPM, J. Ashburner and K. J. Fris-
1999, Hum. Brain Mapp. 7, 254–266) is provided
freeware to researchers of the neuroimaging
munity. We believe that this tool is an improve-
t for the macroscopical labeling of activated
compared to labeling assessed using the Ta-
ach atlas brain in which deformations are well
wn. However, this tool does not alleviate the
tistical Parametric Mapping package (SPM, Friston et
al., 1995b; Ashburner and Friston, 1999). In the
SPM99 package the target brain for spatial normaliza-
tion consists in the average of 152 brains (MNI aver-
age), supplied by the Montreal Neurological Institute
(MNI) (Collins et al., 1994). However, because of the
lack of detailed anatomical features on the average
anatomical image, functional imaging results are usu-
ally overlaid on a high-resolution single-subject T1-
weighted MR, termed the MNI single subject in this
work. Since this brain is not anatomically labeled, most
researchers prefer the use of the Talairach atlas (Ta-
lairach and Tournoux, 1988) to report the localization
of the activations detected in functional imaging stud-
ies, because it offers a detailed anatomical brain de-TECHNICAL
Automated Anatomical Label
Using a Macroscopic An
of the MNI MRI Sing
N. Tzourio-Mazoyer, B. Landeau,* D. Pa
N. Delcroix, B. Mazoye
Groupe d’Imagerie Neurofonctionnelle, UMR 6095 CNRS CEA, Univ
and †Unite´ IRM, CHU de Caen, and Institut
Received April
anatomical parcellation of the spatially nor-
ized single-subject high-resolution T1 volume
ided by the Montreal Neurological Institute
I) (D. L. Collins et al., 1998, Trans. Med. Imag. 17,
468) was performed. The MNI single-subject
n sulci were first delineated and further used as
marks for the 3D definition of 45 anatomical
mes of interest (AVOI) in each hemisphere. This
edure was performed using a dedicated soft-
e which allowed a 3D following of the sulci
rse on the edited brain. Regions of interest were
drawn manually with the same software every 2
on the axial slices of the high-resolution MNI
le subject. The 90 AVOI were reconstructed and
gned a label. Using this parcellation method,
n
b
li
sc
tw
d
co
ic
v
re
0.1006/nimg.2001.0978, available online at http://www.idealibrary.cscrip
man
coor
Tala
o whom correspondence and reprint requests should be ad-
ed at Groupe d’Imagerie Neurofonctionnelle, GIP Cyceron, Bou-
d Becquerel, BP 5229, 14074 Caen Cedex, France. Fax:133 231
22. E-mail: joliot@cyceron.fr.
273NOTE
g of Activations in SPM
omical Parcellation
-Subject Brain
hanassiou, F. Crivello, O. Etard,
and M. Joliot
1
ite´ de Caen, Universite´ de Paris 5; *INSERM U320, Caen;
iversitaire de France, Caen, France
2001
d for more sophisticated labeling strategies
d on anatomical or cytoarchitectonic probabi-
c maps. © 2002 Elsevier Science
INTRODUCTION
ne of the major goals of modern human neuro-
nce research is to establish the relationships be-
n brain structures and their functions and to re-
the anatomical and functional variability which is
iderable between subjects. To reduce the anatom-
variability, the first step of group analysis in acti-
on studies consists of applying tools for spatial
stration and normalization of brain images taken
ontion within the stereotaxic space, including Brod-
n’s areas (BA). However, assigning a set of
dinates to an anatomical label using the paper
irach atlas is quite inaccurate for several reasons:
1053-8119/02 $35.00
© 2002 Elsevier Science
All rights reserved.

FI
subje
cbu.c
(D) O
sulcu
grap
Talai
274 TZOURIO-MAZOYER ET AL.G. 1. (A) Overlay of the Talairach atlas on the T1 average of 152 subjects. (B) Overlay of the Talairach atlas on the T1 average of 152
cts after the nonlinear correction proposed by the MRC Cognition and Brain Sciences Unit (Cambridge, England, http://www.mrc-
am.ac.uk/Imaging/mnispace.html). (C) Overlay of the Talairach atlas on the T1 single-subject template after the nonlinear correction.
verlay of the cortical gray matter of the T1 single-subject template on the T1 average of 152 subjects. (Cortical strip, blue; Rolando
s anterior cortical bank, light orange; subcortical structures, yellow; cerebellum, orange). The Talairach right hemisphere was computer
hically created as the mirror image of the left hemisphere. The red dot indicates a region labeled as Broca’s area with reference to
rach atlas and insula on the MNI single subject.