mapped and adjacent areas results in a thresh-
old value of about 35%. Under the assumption
that this threshold also applies to MPM borders
to yet unmapped areas, this threshold is used
tocuttheMPMtowardunmappedregions.
The visualization of neighboring areas dem-
onstrates that gyri may be occupied by one
or more areas that differ in cytoarchitecture
and function. Conversely,singleareasmaybe
foundonmorethanonegyrus;examplesare
given for the auditory cortex in fig. S7. The
manifold of relationships between areas and
sulci/gyri illustrates the advantage and higher
precision of cytoarchitectonic probabilistic maps
as compared to macroscopic brain maps.
At present, about 70% of the cortical surface
has been covered by completed and published
mapping projects. However, there are still
areas that have not been mapped and repre-
sent projects for future research. To provide
whole-brain coverage for the cortex (fig. S8),
we have combined parts of the cortex that have
not yet been charted into several“gap maps,”
pooling these uncharted areas in a given brain
region (see supplementary materials and fig.
S9). The distributions were modeled so that
probabilistic gap maps were computed in anal-
ogy to the other maps. As mapping progresses,
new maps are continuously replacing gap maps
while the process is captured and documented
by provenance tracking. Consequently, the atlas
is not static (as is, e.g., Brodmann’s map) but
rather represents a“living map”—aconcept
that is known, for example, from geography for
navigating complex spaces.
Gap maps allow computation of a parcella-
tioncoveringtheentirecorticalsurfaceand
the unambiguous assignment of each position
to a cortical region. Together with the increas-
ing number of probabilistic maps of subcortical
nuclei, gap maps contribute to a whole-brain
human atlas. Maps in Julich-Brain can be com-
bined with findings in other atlases and maps;
one example involves the study of microstruc-
tural correlates of activations from neuroim-
aging studies of healthy subjects and patients
(Fig. 3). Moreover, Julich-Brain contributes
to brain modeling and simulation through
informing the model by a functionally sound
microstructural parcellation. It is expected
that this will open new avenues to generate
models of brain activity such as those used in
the treatment of epilepsy, where personalized
brain models are used to predict the propaga-
tion of seizures ( 18 ).
The modular, flexible, and extensible work-
flows cover a broad range of steps from image
acquisition to 3D reconstruction and the gen-
eration of probabilistic maps, which can be
found in several areas of research. The me-
thodical framework (or parts of it) can be ex-
tended to brains of other species, and it can be
used to process section images labeled by other
techniques (e.g., immunohistology). New mod-
ules can be added to the workflows for ap-
plications such as mapping brain areas on the
basis of deep learning ( 19 ).
The Julich-Brain atlas is a freely available re-
source (www.julich-brain.org). Maps have been
made available through different tools and web-
sites such as the SPM anatomy toolbox (www.fz-
juelich.de/inm/inm-7/JuelichAnatomyToolbox),
FSL (https://fsl.fmrib.ox.ac.uk), FreeSurfer (https://
surfer.nmr.mgh.harvard.edu), and the EBRAINS
research infrastructure of the HBP (https://
ebrains.eu/services/atlases). The maps can be
linked to diffusion tensor imaging (DTI)–based
connectivity data (Fig. 3C) and to gene expres-
sion data provided by the Allen Institute for
Brain Science (https://alleninstitute.org/what-
we-do/brain-science) through the JuGEx tool
( 20 ) to enable a multimodal perspective on hu-
man brain organization (Fig. 3B).
Julich-Brain represents a new kind of human
brain atlas that is (i) cytoarchitectonic to reflect
a basic principle of the brain’smicrostructural
parcellation; (ii) whole-brain,to cover both the
cerebral cortex and subcortical nuclei; (iii)
3D-probabilistic, to consider variations be-
tween individual brains in stereotaxic space;
(iv) dynamic, a living atlas, to be supplemented
by maps of new areas or subdivision of existing
maps of areas (e.g., when new studies suggest
a finer or new parcellation); (v) flexible, to
allow for modifications of modules in the
workflows for other data modalities, organs,
or species; (vi) open-access and based on FAIR
principles, to contribute to studies by other
researchers addressing structure-function re-
lationships and network organization; and (vi)
interoperable, to link it to other atlases and
resources that provide complementary informa-
tion about brain organization.REFERENCES AND NOTES- A. Goulas, K. Zilles, C. C. Hilgetag,Trends Neurosci. 41 ,
775 – 788 (2018). - K. Amunts, K. Zilles,Neuron 88 , 1086–1107 (2015).
- D. C. Van Essenet al.,Proc. Natl. Acad. Sci. U.S.A. 116 ,
26173 – 26180 (2019). - B. Fischl, M. I. Sereno,Neuroimage 182 , 219–231 (2018).
5. R. Nieuwenhuys, C. A. J. Broere,Brain Struct. Funct. 222 ,
465 – 480 (2016).
6. M. F. Glasseret al.,Nature 536 , 171–178 (2016).
7. J. K. Maiet al.,Atlas of the Human Brain(Academic Press, ed. 4, 2015).
8. K. Zilles, K. Amunts,Nat. Rev. Neurosci. 11 ,139–145 (2010).
9. A. C. Evans, A. L. Janke, D. L. Collins, S. Baillet,Neuroimage 62 ,
911 – 922 (2012).
10. K. Amuntset al.,J. Neurosci. 27 , 1356–1364 (2007).
11. K. Amuntset al.,Science 340 , 1472–1475 (2013).
12. H. Mohlberget al., inBrain-Inspired Computing: Second
International Workshop, BrainComp 2015, K. Amuntset al., Eds.
(Springer, 2016), pp. 15–27.
13. J. Pichat, J. E. Iglesias, T. Yousry, S. Ourselin, M. Modat,
Med. Image Anal. 46 ,73–105 (2018).
14. H. Damasio,Human Brain Anatomy in Computerized Images
(Oxford Univ. Press, ed. 2, 2005).
15. S. L. Dinget al.,J. Comp. Neurol. 524 , 3127– 3481 (2016).
16. A. Schleicheret al.,Anat. Embryol. 210 , 373–386 (2005).
17. S. B. Eickhoffet al.,Neuroimage 25 , 1325–1335 (2005).
18. T. Proix, F. Bartolomei, M. Guye, V. K. Jirsa,Brain 140 ,641–654 (2017).
19. H. Spitzeret al., inMedical Image Computing and Computer
Assisted Intervention—MICCAI 2018, Part III, A. F. Frangiet al.,
Eds. (Springer, 2018), pp. 663–671.
20. S. Bludauet al.,Brain Struct. Funct. 223 ,2335–2342 (2018).
ACKNOWLEDGMENTS
We thank more than 41 postdocs, doctoral students, guests,
and colleagues who contributed to the mapping of 248
cytoarchitectonic areas and nuclei; the technical assistants of the
C. and O. Vogt Institute of Heinrich Heine University Düsseldorf
and Research Centre Jülich; S. Eickhoff, R. Hübbers, P. Pieperhoff,
N. Palomero-Gallagher, S. Caspers, T. Dickscheid, and A. C. Evans
for intensive discussions; A. Schleicher for developing the original
observer-independent mapping approach; our partners of the
International Consortium for Human Brain Mapping who stimulated
and actively supported this research; and our colleagues at the Jülich
Supercomputing Centre, in particular B. Tweddell and T. Lippert.
Funding:Supported by the Portfolio Theme“Supercomputing and
Modeling for the Human Brain”of the Helmholtz Association, the
European Union Seventh Framework Programme (FP7/2007-2013,
HBP), and the European Union’s Horizon 2020 Research and
Innovation Programme, grant agreements 604102 (HBP SGA1),
785907 (HBP SGA2), and 945539 (HBP SGA3).Author
contributions:K.A. and K.Z. developed the concept of 3D
probabilistic cytoarchitectonic brain mapping and atlas. K.A. is
overseeing the atlas projects, contributed to methodological
developments, and drafted the manuscript. H.M. developed and
adapted the methodology and the software for data processing
and atlas building, and drafted the manuscript. S.B. developed new
tools for analysis and mapping of cytoarchitectonic areas. All authors
contributed to the writing of the manuscript.Competing interests:
The authors declare no competing interests.Data and materials
availability:All data are available in the main text or the
supplementary materials. Various data modalities of the already
published maps of brain areas and the complete atlas are available
online at http://www.julich-brain.org and via EBRAINS (https://ebrains.eu/)
(DOI: 10.25493/TAKY-64D). Previously developed parts of the
workflows were published earlier; new scripts for computing the
Julich-Brain Atlas and for analyzing contour lines and gap maps
(https://doi.org/10.5281/zenodo.3906413) are provided via the git
repository, https://github.com/JulichBrainAtlas.
SUPPLEMENTARY MATERIALS
science.sciencemag.org/content/369/6506/988/suppl/DC1
Materials and Methods
Figs. S1 to S9
Tables S1 and S2
References ( 21 – 56 )
27 February 2020; accepted 24 June 2020
Published online 30 July 2020
10.1126/science.abb4588Amuntset al.,Science 369 , 988–992 (2020) 21 August 2020 5of5
RESEARCH | REPORT