Three-Dimensional MRI Atlas of the Human Cerebellum in ... · ﬁssure on every slice in all three...

uaTtfatTmgmtslob

pp

itohetolgaimoct

NeuroImage 10, 233–260 (1999)Article ID nimg.1999.0459, available online at http://www.idealibrary.com on

Three-Dimensional MRI Atlas of the Human Cerebellum in ProportionalStereotaxic Space

Jeremy D. Schmahmann,* Julien Doyon,†,‡ David McDonald,‡ Colin Holmes,§ Karyne Lavoie,†Amy S. Hurwitz,* Noor Kabani,‡ Arthur Toga,§ Alan Evans,‡ Michael Petrides‡,¶

*Department of Neurology, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts; †Department ofPsychology, and Rehabilitation Research Group, Francois-Charon Center, Laval University, Quebec City, Canada; ‡McConnell BrainImaging Center, and ¶Cognitive Neuroscience Unit, Montreal Neurological Institute, and McGill University, Montreal, Canada; and

§Laboratory of Neuroimaging, Department of Neurology, University of California School of Medicine, Los Angeles, California

Received February 11, 1999

vis

bra1trlsccars

cdfuiclaeboistliipca

We have prepared an atlas of the human cerebellumsing high-resolution magnetic resonance-derived im-ges warped into the proportional stereotaxic space ofalairach and Tournoux. Software that permits simul-aneous visualization of the three cardinal planesacilitated the identification of the cerebellar fissuresnd lobules. A revised version of the Larsell nomencla-ure facilitated a simple description of the cerebellum.his atlas derived from a single individual was instru-ental in addressing longstanding debates about the

ross morphologic organization of the cerebellum. Itay serve as the template for more precise identifica-

ion of cerebellar topography in functional imagingtudies in normals, for investigating clinical–patho-ogic correlations in patients, and for the developmentf future probabilistic maps of the human cere-ellum. r 1999 Academic Press

Key Words: cerebellum; atlas; anatomy; human; pro-ortional stereotaxic (Talairach) space; brain map-ing.

As clinical and functional magnetic resonance imag-ng (MRI) technology continues to improve, the abilityo differentiate cerebellar lobules and folia on all planesf section has become quite precise. There is a problem,owever, in correlating this technology with the knowl-dge of cerebellar anatomy because there is no atlashat can be reliably used for the accurate identificationf the different cerebellar regions. Terms such as medial/ateral, anterior/posterior, or vermis/hemisphere areenerally used in clinical studies, but these terms arenatomically vague. In functional imaging studies us-ng positron emission tomography (PET) and functional

agnetic resonance imaging (fMRI) when the locationsf cerebellar activations are presented as Talairachoordinates, it has generally not been possible to iden-
ify precisely which cerebellar lobule or folium is in-
233

olved. The atlas of Talairach and Tournoux (1988)tself has no anatomic structures depicted within theketched outlines of the cerebellum.Available cerebellar atlases depict major landmarks

ut they are generally limited to gross morphologicelationships and they cannot localize brain structuresctivated in PET and fMRI studies (e.g., Crosby et al.,962; Carpenter, 1976; DeArmond et al., 1976; Wadding-on, 1984; Roberts et al., 1987; Kretshmann and Wein-ich, 1992). In these atlases the individual cerebellarobules are generally not labeled, and only limitedections are available in either one or two of theardinal planes. More recent attempts to address theseoncerns using MRI (Courchesne et al., 1989; Press etl., 1989, 1990) are useful, but they are limited by theelatively small number of sections presented and thepatial resolution.The most detailed available atlas of the human

erebellum is that of Angevine et al. (1961). Thisissection in three planes reveals the structure of theolia and deep nuclei. There are inherent difficulties insing this atlas for contemporary purposes, however. It

s applicable only in the parasagittal plane, as theoronal and axial planes of sections shown are dissimi-ar from those used in clinical and functional neuroim-ging studies. There are a limited number of sections inach of the cardinal planes, and there are large gapsetween the sections. A further critical limiting factorf the Angevine and other atlases is that the cerebellums not viewed within the widely used proportionaltereotaxic space of Talairach and Tournoux. It isherefore not possible to determine the anatomic corre-ate of a Talairach coordinate derived from a functionalmaging experiment. Even coregistration of a PETmage with an MRI template of the cerebellum may notroduce precise structure–function correlations be-ause there is no atlas that can accurately identify thectivated anatomic structure.
An additional consideration that limits the useful-
1053-8119/99 $30.00Copyright r 1999 by Academic Press

All rights of reproduction in any form reserved.

234 SCHMAHMANN ET AL.

niotmAF1

1111tB1a

c

2353-D MRI ATLAS OF HUMAN CEREBELLUM

ess of available atlases is that the terminology used todentify the fissures and lobules is not uniform and isften contradictory. There have been many revisions ofhe nomenclature of cerebellum of both human andonkey through the years (Malacarne, 1776; Vicq d’zyr, 1786; Meckel, 1838; His, 1895; Stroud, 1895;latau and Jacobsohn, 1899; Dejerine, 1901; Smith,902; Bradley, 1904; Bolk, 1906; Edinger, 1909; Ingvar,

FIG. 2. Right lateral view of the

FIG. 1. This figure illustrates the Display software used in this st
erebellum as well as on a 3-D reconstruction of the surface.
918, 1928; Langelaan, 1919; Jakob, 1928; Riley, 1929,930; Anatomical Society of Great Britain and Ireland,933; Ziehen, 1934; Ariens Kappers et al., 1936; Dow,942; Loyning and Jansen, 1955; International Ana-omical Nomenclature Committee, 1956; Jansen androdal, 1958; Larsell 1934, 1937, 1947, 1953, 1958,962, 1970; Larsell and Jansen, 1972). The number ofttempts to understand the gross morphologic organiza-

ures with cerebellar tissue removed.

y. The cursor is visible on the sagittal, coronal, and axial views of the

fiss

ud

tTtiwhracc(

cihqothfopocf

uottsatiospa

M

wtio1v((cdvpsa

t31ataibuevdrrov1

I

oSptcviifisvcavaatelta

imifiifmssvt

ss


ion of the cerebellum reflects the difficulty of the task.here is considerable variation and conflict within theerminologies previously adopted. The activation seenn functional imaging studies and the precision withhich MRI can determine the site of a lesion in theuman cerebellum make it imperative now that eachegion be identified by a uniform nomenclature. It islso important to correlate hemispheric and vermalomponents of each lobule and to compare humanerebellar morphology with that of lower speciesLarsell, 1970; Madigan and Carpenter, 1971).

A detailed atlas developed specifically for use inonjunction with anatomic and functional neuroimag-ng would be a valuable tool for understanding theuman cerebellum. This is essential in order to addressuestions of functional localization, the predictabilityf location of cerebellar fissures, lobules, and folia, andhe possibility of predictable asymmetry across theemispheres. The difficulty in correlating the details ofocal activations in functional imaging studies withnly vague notions of cerebellar anatomy has ham-ered the understanding of the human cerebellum andf how it is incorporated into the distributed neuralircuits that subserve different domains of neurologicunction.

We have therefore prepared a human cerebellar atlassing high-resolution MRI images that satisfies manyf the mandates of a useful atlas. It is presented in thehree cardinal planes in a universally understood sys-em of proportional stereotaxic space with sectionshown at 2-mm intervals, displaying high resolution,nd utilizing a revised and simplified nomenclaturehat avoids the terminological confusion that character-zed earlier attempts. This paper presents the method-logy used, selected sections of the atlas, and 3-dimen-ional depictions of the lobules and fissures. We alsoresent a revised nomenclature developed from thistlas.

METHODS

agnetic Resonance Imaging

A signal-enhanced magnetic resonance (MR) scanas used to permit the identification of brain struc-

ures in as noise-free data as possible. To produce thesemages, a series of 27 T1-weighted MR scans wasbtained from a single subject (CJH) using a Phillips.5 Tesla MR. The scan parameters were: 3D sagittalolume composed of 140 1-mm slices, FOV 256 mmSI) 3 204 mm (AP), acquired with a spoiled GRASSTR/TE 5 18/10 ms, flip angle 30°, NSA (NEX) 1, flowompensation on for a total scan time of 10 h 50 min,uring 27 scans. The 27 scans consisted of 20 1.0-mm3

ol and 7 0.78-mm3 vol. The scans were acquired over aeriod of 3 months, with the maximum duration of anycanning session being 3 h. Each scan was registered
nd transformed into a standard proportional stereo- i
axic space, the Montreal Neurologic Institute (MNI)05 brain averaged-space (Talairach and Tournoux,988), using an automatic registration tool (Collins etl., 1994) and resampled to 0.5 mm3 before normaliza-ion and intensity averaging. The effect of the intensityveraging of the 27 vol was a gain of approximately 4.5n the signal-to-noise ratio. This effect was accentuatedy the reduction of partial volume effects. Due to thenavoidable small motions of the head between scans,ach region of space was imaged in slightly differentoxels in each scan. As the final point sample wasrawn from many volumes, all slightly displaced withespect to one another, the final subsampled voxelseduce the overall partial volume effect. Further detailsf the production of the intensity average of theseolumes have been reported previously (Holmes et al.,998).

dentification of Cerebellar Fissures and Lobules

The brain slices were viewed using in-house softwaref the McConnell Brain Imaging Center running onilicon Graphics workstations (Fig. 1). The interactiveortion of the software, Display, provided several simul-aneous views of imaging data, including sagittal,oronal, and horizontal cross-sections. Within eachiew a variety of standard color-coding algorithms wasnvoked to assign colors to the numerical values of themages, and each view could be independently magni-ed and translated. The current positions of the cross-ectional planes were chosen by selecting a point on anyiew that caused the positions of the other views tohange accordingly. This permitted cross-correlation ofgiven locus between the different planes. Individual

oxels in the image volume were labeled with anrbitrary integer label by using the mouse to sweep outcircular brush across an image plane. In addition, a

hree-dimensional view of reconstructed surface geom-try was available. The boundaries in an image orabeling of an image were created by an isosurfaceriangulation algorithm (Lorenson and Cline, 1987)nd displayed in the three-dimensional window.The atlas of Angevine et al. (1961) was used to

dentify fissures that could reliably be seen on theidsagittal plane. The cerebellar fissures were then

dentified by labeling empty voxels in the center of eachssure on every slice in all three planes, and this

nformation was used to parcellate the lobules andolia. For those fissures that were optimally seen onore lateral sections (e.g., the superior posterior fis-

ure), comparison was made with corresponding para-agittal sections in the atlas of Angevine et al. andalidated using the coronal and horizontal images ofhis atlas.

The voxellated fissures were subjected to a least-quares curve-fitting smoothing algorithm, which as-isted in the final stage of presentation of the data. The
mages were rendered with in-house ray-tracing soft-

wplncb

sngtlggoslaecMdc

ti

T

tmuE5w1vficawtfil

M

aafetecnuaitTtftscfiqchwlpsbt

tf


are with appropriate ruled indicators delineatingositions in Talairach space. Finally, the lobules wereabeled according to our revision of the cerebellaromenclature that respects the vermal–hemisphericorrelation outlined in Table 1 and explained furtherelow.This investigation was not specifically designed to

tudy the detailed anatomy of the deep cerebellaruclei. The outline of the nuclei characterized by theray-white matter differentiation proved indistinct onhese images. It was therefore difficult to define theirimits with certainty. This was true for the nuclei as aroup, as well as for the delineation of the fastigial,lobose, emboliform, and dentate nuclei from eachther. This limitation was compensated for by compari-on with postmortem cryosections of another cerebel-um. This cryosectioned brain was prepared, digitized,nd then warped into Talairach space as describedlsewhere (Toga et al., 1997). Equivalent sections in theardinal planes were selected for comparison with theRI atlas. The anatomy of the deep nuclei is well

efined in these cryosections, and their location andonfiguration can be compared with the predicted loca-

TABLE 1

Relationship of the Fissures to the Cerebellar Lobules in theVermis and Hemispheres

p

ion within the cerebellar parenchyma on the MRImages.

hree-Dimensional Reconstructions

The fissures marked on the serial sections wereransformed into smooth sheets coursing through 3-di-ensional space with the cerebellar tissue removed,

sing image blurring and the isosurface algorithm.ach set of labeled voxels was first blurred with a-mm-width box filter, then the isosurface algorithmas applied to the blurred image (Lorenson and Cline,987). The resulting triangulated surface constituted aisually smooth representation of the volumes of thessures. These fissures suspended in the invisibleerebellum are shown as viewed from the right lateralspect (Fig. 2). The external surface of the cerebellumas reconstructed by applying the isosurface algorithm

o the average MR image. The surface markings of thessures were identified on these views, and the cerebel-

um thus parcellated into lobules (Fig. 3).

ethodological Limitations

Single subject. There are important limitations thatrise from the use of single brain to develop a cerebellartlas. A number of variables are by definition excludedrom consideration. It is not possible, for example, tovaluate consistency of hemispheric asymmetry andhe individual variation of the size, shape, and constitu-nt elements of an individual lobule. Some possibleonfounding variables such as gender, age, and handed-ess are also not factored into this evaluation. Thetility of the approach used in this in-depth analysis ofsingle brain, however, is that the primary step is the

dentification of the major fissures, and the determina-ion of the lobules and folia follows from this first step.he anatomic location of major fissures can be ascer-

ained in other cerebella by reference to the anatomiceatures of this atlas, even if there is some variability inheir location within standard stereotaxic space andome divergence from the exact shape of this particularerebellum. This atlas can be used as the template foruture population-based probabilistic studies that exam-ne large numbers of cerebella, and that can addressuestions of individual variability determined by, oroncomitant with, such considerations as gender andandedness. Morphometric studies of the cerebellumill also be facilitated by reference to the fissures,

obules, and folia demonstrated in this atlas. It is notossible to proceed to the next stage of morphologictudies of the cerebellum without first identifying theoundaries and structures of a single cerebellum usinghe MRI technology.

Talairach coordinate space. It could be argued thathe use of Talairach proportional space is suboptimalor an atlas of the cerebellum because the anterior–
osterior commissure line is far removed from the


FIG

.3.

3-di

men

sion

alre

con

stru

ctio

ns

ofth

esu

peri

or(a

bove

)an

dpo

ster

ior

(bel

ow)s

urf

aces

ofth

ece

rebe

llu

mpr

esen

ted

asst

ereo

-pai

rs.

pm(ubt

ptsut


osterior fossa. The concern, therefore, is that theargin of error in reporting the location of structures

or structure–function correlations) in cerebellum isnacceptably great. This concern may have validity,ut the Talairach system is widely used, and activa-

TAB

Comparison of Earlier Nomenclature Systems with

ions on functional imaging scans are commonly re- d

orted in both cerebral and cerebellar structures usinghe same sagittal/coronal/axial images. It seems pos-ible to develop a system of coordinates specifically forse with the posterior fossa, but the overall utility ofhis approach may be less generally applicable. The

2

Designation Used in This Atlas (Based on Larsell)

LE

the

egree to which the location of the cerebellar folia are


TA

BL

E2—

Con

tin

ued

ctf

bvtce

N

taceW

siVambairsptptt(a

mT


onsistent from one brain to the next using this stereo-axic proportional space will need to be determined inuture investigations.

Cerebellar nuclei. The absence of clearly visibleoundaries of the cerebellar nuclei on the MRI pre-ented the confident incorporation of the nuclei intohis atlas. This limitation was partially overcome byomparison with digitized autopsy cryosections of cer-bellum warped into Talairach space.

omenclature

What began as a straightforward project of mappinghe fissures and the nomenclature used by Angevine etl. (1961) onto the MRI images turned out to be moreomplicated. The nomenclature systems used by botharly and contemporary investigators are confusing.e addressed this issue by reviewing the evidence in

FIGS. 4–14. Sagittal images of the right cerebellum (Figs. 5–9) anm. The midsagittal plane (x 5 0) is shown in Fig. 4. In this and subse

able 1.

upport of the different terminologies presented by thenvestigators mentioned above, starting with that ofincenzo Malacarne in 1776. The atlas of Angevine etl. also helped us revise the nomenclature and deter-ine the relationship between the fissures and lobules,

ecause their atlas included an analysis, comparison,nd synthesis of the findings of many of the earliernvestigators. By using the current technology (MRIeconstructions and Display software that permitsimultaneous cross-referencing between the threelanes) we were able to explain the origins of someerminological difficulties that have arisen over theast two centuries and then resolve them. We amendedhe terminology of Larsell and compared this with theerminologies introduced by Malacarne (1776), Bolk1906), Ziehen (1934), and others. In so doing wedopted a nomenclature that is anatomically correct,

ft cerebellum (Figs. 10–14), progressing laterally by increments of 10ent figures, the fissures are color coded according to the convention in

d lequ

c
FIG. 5. Top: Sagittal MRI image of the cerebellum 10 mm to the right of midline. Bottom: A postmortem cryosection of another human
erebellum in the sagittal plane prepared according to the method of Toga et al. (1997), corresponding to x 5 110 mm. D, dentate nucleus.242

e corresponding to x 5 120 mm. D, dentate nucleus.


FIG. 6. Bottom: A cryosection in the sagittal plan


FIGURES 7 AND 8

ebb

nitsshcimJtsbwtMfcfmc

T

at(stlieo‘sipbc(osttm(


asily understandable, and melds with the neuronamesrain hierarchy (Bowden and Martin, 1995) acceptedy the International Consortium for Brain Mapping.The method employed for determining the simplified

omenclature was derived from the use of the atlastself. Vermal lobules were identified with the assis-ance of the midsagittal section of the Angevine atlas,ubsequent to the identification of the fissures andulci. The lobules that could be readily discerned in theemisphere were also demarcated, such as crus I andrus II that are separated by the horizontal fissure thats prominent in lateral sections. By applying a slightly

odified version of the Larsell (1970; Larsell andansen, 1972) terminology, and maintaining consis-ency through all planes in continuous adjacent 0.5-mmections, it was possible to determine which foliumelonged to which lobule and which Larsell designationas appropriate. The descriptions and illustrations of

he rhesus monkey cerebellum by Larsell (1970) andadigan and Carpenter (1971) were also helpful in

urthering our understanding of the folia of the humanerebellum. Some of the small subfolia buried at theundus of the fissures or sulci were troublesome, butost could be parcellated by use of the 3-dimensional

FIG

orrelations. fi

he Problem of the Vermis

There is no true ‘‘vermis’’ in the anterior lobe. Thepplication of this term to the paramedian sectors ofhe anterior lobe is an extension of the Latin termmeaning worm) used by Malacarne to denote thetructure visible in the posterior and inferior aspect ofhe cerebellum. The vermis (as such) is present fromobules VI through X. The use of the term vermis tondicate ‘‘midline’’ has become well entrenched, how-ver, and in so doing it has brought with it the problemf defining what is the lateral extent of the anterior lobe

‘vermis.’’ It has been suggested that the paravermianulcus limits the vermis laterally. In many brains theres no paravermian sulcus and where one appears to beresent, it may simply reflect the indentation producedy the course of the medial branch of the superiorerebellar artery. On coronal section the anterior lobelobules I through V) has cortex buried beneath theverlying midline cortex, separated from the hemi-pheral cortex by a superior white matter extension ofhe medullary core (see, e.g., coronal sections y 5 242hrough y 5 258). Even where there may be a paraver-ian sulcus this is most likely the preculminate fissure

see coronal section y 5 242) or the intraculminate

E 9
UR
ssure (y 5 248) that is not necessarily constant from


FIG. 10. Bottom: A cryosection in the sagittal pla
ne corresponding to x 5 210 mm. D, dentate nucleus.


FIG. 11. Bottom: A cryosection in the sagittal plane corresponding to x 5 220 mm. D, dentate nucleus.


FIGURES 12 AND 13

olmbes(edvtlltt

stcdtIvw

sritfitwctsptwumlmhpir


ne section to the next and does not demarcate theateral extent of the hidden cortex that truly occupies a

idline position between the hemispheres. In thisrain the anterior lobe ‘‘vermis’’ occupies a mediolateralxtent that varies from 14 to 20 mm depending on theection and the definition of the lateral boundarysurface markings of the ‘‘paramedian sulcus’’; lateralxtent of the buried cortex; lateral aspect of the parame-ian white matter laminae). Each brain will no doubtary to some extent, and the functional significance ofhis distinction is not known. The determination of theateral extent of the ‘‘vermis proper’’ (in the posteriorobe) is not a problem, as this is a gross morphologicerm originally intended to define a structure visible onhe surface of the brain.

Larsell distinguished the vermis from the hemi-pheral compartments of the different lobules by usinghe prefix ‘‘H’’ for the hemispheres. As the bulk of theerebellum is composed of hemisphere, and the ‘‘H’’esignation adds a cumbersome element to the descrip-ion of the lobules, we dispensed with this designation.nstead, when referring to a midline structure the termermal area (anterior lobe) or vermis (posterior lobe)

FIGU

as added. c

RESULTS

The technique of averaging multiple scans of theame brain produced MR images that were of highesolution and permitted detailed evaluation. The coreg-stration of the images in sagittal, coronal, and horizon-al planes was an essential tool in the definition of bothssures and folia and facilitated the characterization ofhese structures with confidence. Eleven major fissuresere identified, and these were used to subdivide the

erebellum into lobules. The 3-dimensional reconstruc-ion of the surface of the MRI cerebellum with theuperimposed externally visible fissures provided inde-endent confirmation of the accuracy of the labeling ofhe fissures. The differentiation of folia and subfoliaas clear in most instances. The cerebellar cortex couldsually be distinguished from the medullary core andedullary rays extending into the larger folia. Cerebel-

ar nuclei on the MRI images were visible as grayatter structures situated within the medullary core;

owever, they were not sufficiently well defined toermit accurate delineation. The nuclei were readilydentifiable, however, on the equivalent sections de-ived from the cryosections of the postmortem human

E 14
R
erebellum.

N

rtccltcfitfwtcata

V

tassficap1p

T

pt


omenclature

The revised nomenclature proved to be simple andeliable. The names of the fissures were not changed, ashey are anatomically consistent, have historic signifi-ance, and are still widely used and understood. Inontrast, we dispensed with the Latin names of theobules, and used Larsell’s Roman numeral designa-ions with some revisions. Table 1 presents the nomen-lature used in this atlas to describe the cerebellarssures and lobules. Table 2 presents a comparison ofhe earlier nomenclature systems with that derivedrom this atlas (based on Larsell). This table is helpfulhen drawing anatomic correlations between studies

hat employ different nomenclatures. The detailed dis-ussion of the various nomenclatures through the years,nd the way that this atlas, and particularly the use ofhe Display software, allowed for the resolution of thepparent conflicts is beyond the scope of this paper.

FIGU

FIGS. 15–19. Axial images of the cerebellum commencing at z 5 215

iews of the Cerebellum in the Sagittal, Coronal, andHorizontal Planes

Selected sections of the cerebellum are presented inhe parasagittal (Figs. 4–14), horizontal (Figs. 15–20),nd coronal planes (Figs. 21–27), showing the progres-ion of the fissures and lobules from midline to lateral;uperior to inferior; and anterior to posterior. Thessures are color coded, according to the table. Selectedryosections of postmortem brain show the locationsnd anatomic structures of the deep nuclei in thearasagittal (Figs. 5 (bottom), 6 (bottom), 10 (bottom),1 (bottom)), horizontal (Fig. 17 (bottom)), and coronallanes (Figs. 21 (bottom), 22 (bottom)).

hree-Dimensional Images

The three-dimensional reconstruction of the fissuresrovides a new view of the fissures as they coursehrough the cerebellum and subdivide it into lobules

E 15
R
and progressing inferiorly to z 5 255.

(asfii

D

aaetfpthb

lsw

af

l

tdiiwfrstscttt


Fig. 2). The 3-dimensional renditions of the superiornd posterior surface views of the cerebellum (shown astereo-pairs in Fig. 3) reveal the external location of thessures and the parcellation of the cerebellar surface

nto lobules.

escription of the Fissures and Lobules

Vermal lobules I and II were indistinguishable andbut the anterior medullary velum, which lies anteriornd inferior to the vermis. There is no hemisphericxtension of vermal lobule I. Hemispheric lobule II ishe lateral extension of I and II. Lobule I is distinctrom lobule II in some human cerebella and in lowerrimates. This distinction is thus maintained, evenhough it is not always apparent. The vermal andemispheric components of lobule I/II have togethereen termed the lingula.The precentral fissure separates lobules I and II from

obule III (centralis), both in the vermis and the hemi-pheres. Lobule III is partially obscured by lobule IV

FIGU

hen viewed from the superior aspect, but it is size- t

ble, has a semilunar form, and is the first distinctolium not attached to the superior medullary velum.

The preculminate fissure separates lobule III fromobule IV.

The use of the older term ‘‘culmen’’ (including bothhe vermis and the hemisphere) obscures the fact of theivision of the culmen into lobules IV and V by thentraculminate fissure. This fissure is readily discern-ble on the hemisphere, but at the midline is continuousith a small fissure situated between the branches of a

olium and thus can be difficult to localize with accu-acy in the absence of adjacent sagittal sections orurface reconstructions. In this brain it was possible torace the intraculminate fissure from the right hemi-phere across the midline to the left side. By using theoronal images (see y 5 246) we determined that thehree major folia of the ‘‘culmen’’ are subdivided suchhat lobule IV contains one folium, and lobule V con-ains two.

The primary fissure distinguishes the anterior lobe of

E 16
R
he cerebellum (lobules I through V) from the posterior

FIG. 17. A cryosection in the axial plane corresponding to z 5 235. D, dentate nucleus.

252

FIGURES 18 & 19

253

lrtwtauom

otbawmhrct

f(

s

ofbhtts

llctrtalabh

vfit


obe (lobules VI through IX), and specifically it sepa-ates lobule V from lobule VI, both in the vermis andhe hemisphere. The primary fissure and other fissuresithin the anterior lobe are progressively more difficult

o discern on parasagittal sections as one moves later-lly away from the midline. The primary fissure isnmistakable on midsagittal section but it is continu-us with an undistinguished small fissure in the inter-ediate sectors of the hemispheres.The superior posterior fissure appears early in devel-

pment and is visible in this brain on the 3-D reconstruc-ion of the superior surface (Fig. 3). Lobule VI liesetween the primary and superior posterior fissuresnd is therefore sizeable and contains two large foliaith medullary rays that take their origin off theedullary core (e.g., see y 5 258). The vermal andemispheric components of lobule VI have previouslyeceived different names. Lobule VI in the vermis wasalled the declive, and in the hemisphere it has receivedhe name lobulus simplex among others (see Table 2).

The superior posterior fissure separates lobule VIrom lobule VII in the vermis and lobule VI from crus Iof the ansiform lobule) in the hemisphere.

Vermal lobule VII is complex and its lateral exten-

FIGS. 20–25. Coronal images of the cerebellum comm

ion into the ansiform lobule of Bolk forms a large part l

f the hemisphere in the human cerebellum. The termsolium and tuber have been used to denote vermis VII,ut this fails to convey the complexity of the equivalentemispheric extensions. Further, VIIA (in the Larsellerminology) is divided by the horizontal fissure intowo sectors, with important differences in their hemi-pheric extensions.Vermal lobule VIIAf (identical to the folium) expands

aterally to form crus I (of the ansiform lobule). Vermalobule VIIAt (previously the rostral half of the tuber)orresponds in the hemisphere with crus II. The horizon-al fissure separates VIIAf from VIIAt in the vermalegion, and crus I from crus II in the hemispheres. Inhis brain, VIIAf is not present in the midline. Itppears at 6 mm to the right of midline and 4 mm to theeft of midline. For this reason, in the midsagittal planend just lateral to it, the superior posterior fissure isounded caudally by the ansoparamedian and not theorizontal fissure.The ansoparamedian fissure is submerged on the

entral surface of the ‘‘tuber,’’ separating lobules VIIAtrom VIIB (previously termed the paramedian or grac-le lobule). The precise delineation of this fissure andhe division of crus II from VIIB in the hemisphere has

cing at y 5 238 and progressing caudally to y 5 288.
en
ong been the subject of controversy. We were able to


FIG. 21. Bottom: A cryosection in the coronal plane corresponding to y 5 248. D, dentate nucleus.


FIG. 22. Bottom: A cryosection in the coronal plane corresponding to y 5 258. D, dentate nucleus; E, emboliform nucleus.

FIGURES 23 & 24

ddwbdssptrotIrhVeZttcsm

f

Vfib‘t

v(o

tflth

bdMstE


elineate this fissure, but in following the ansoparame-ian fissure from the midline laterally, an asymmetryas revealed in the delineation of crus II. This occurredecause on the left side the horizontal and ansoparame-ian fissures meet on the posterior inferior cerebellarurface. Crus II on the left therefore is a triangular lobeituated on the superior and the posterior aspects of theosterior lobe. On the right side, however, the horizon-al and ansoparamedian fissures do not meet, thusevealing that crus II continues laterally and forwardnto the anterior surface of the cerebellum. The effect ofhis asymmetry is that on lateral sagittal sections, crusis adjacent to lobule VIIB on the left, whereas on theight crus II is adjacent to VIIB. This asymmetry mayelp explain some of the earlier debate as to whetherIIB is distinct from the inferior semilunar lobule (see,.g., Basle Nomina Anatomica, 1895; Smith, 1902;iehen, 1934; Larsell and Jansen, 1972). Coronal sec-ions in this brain (see y 5 254 through 274) confirmhat despite the fact that the folia at the lateral inferiorerebellar surface are asymmetric across the two hemi-pheres, they are derived from medullary rays that areorphologically similar.The prepyramidal/prebiventer fissure separates VIIB

FIGU

rom VIIIA. It is named according to whether the s

IIB/VIIIA distinction is in the vermis (prepyramidalssure) or hemisphere (prebiventer fissure). The intra-iventer fissure separates VIIIA from VIIIB. (The term

‘intrapyramidal’’was omitted for purposes of simplifica-ion.)

The secondary fissure is interposed both at theermis and in the hemispheres between VIIIB and IXin the older terminology—uvula at the vermis; tonsilr ventral and accessory paraflocculus at the hemisphere).The posterolateral fissure forms the boundary be-

ween the posterior lobe of the cerebellum and theocculonodular lobe, separating IX from X (in the oldererminology—nodulus at the vermis; flocculus at theemisphere).

DISCUSSION

The understanding of the human cerebellum haseen substantially advanced by the availability ofetailed anatomic images revealed by conventionalRI and by cerebellar activation with PET and fMRI

tudies of motor, sensory, and cognitive/emotional func-ions (see, e.g., Desmond et al., 1998; Doyon, 1997;lliott and Dolan, 1998; Fiez and Raichle, 1997; Par-

E 25
R
ons and Fox, 1997). Until now, determining the precise

lcdh

inpteoecsacpaicaa

ntpmtbtbtttaadodcoptc

bwtcnneitficu

na

icdpsatvf

MpaAHNao

A

A

A

B

B

BC

C

C

C

D

D

D

D


ocation of lesions or sites of activation within theerebellum has been limited by the lack of a sufficientlyetailed and practical atlas. The MRI atlas that weave developed addresses this need.This atlas has relied upon high-resolution anatomic

mages and Display software that permits simulta-eous visualization of a cursor in the three cardinallanes within the cerebellum and 3-D reconstruction ofhe data. The resulting atlas of a human cerebellumxtends the understanding of the gross morphology andrganization of the cerebellum to a previously impen-trable level. It clarifies the anatomy of the humanerebellum, provides correlations of vermal with hemi-pheric structures, and defines the cerebellar lobulesnd folia. It demonstrates the details of the cerebellarortex within the three cardinal planes in Talairachroportional stereotaxic space and makes this materialvailable and useful to clinicians and investigatorsnterested in the study of the cerebellum. An immediateonsequence of this new understanding of cerebellarnatomy is a revised and more contemporary andccessible cerebellar nomenclature.It is hoped that this atlas will facilitate the study of a

umber of anatomical–functional–pathological ques-ions that could not previously be addressed. Theossibility of predictable cerebellar hemispheric asym-etries and the constancy of the location and pattern of

he cerebellar fissures and lobules may now be studiedy comparing identifiable cerebellar landmarks be-ween different individuals. This atlas may serve as theasis for future morphometric analyses and probabilis-ic maps of the human cerebellum and permit quantita-ive estimations and comparisons to be made regardinghe size of different lobules and folia. Morphometricnalysis will be necessary to determine the changes, ifny, that occur with normal aging and in cerebellariseases and for accurately monitoring rate and degreef atrophy once treatment options become available foregenerative diseases of the cerebellum. We have re-ently used this atlas to develop a semiflattened mapnto which we charted the results of a meta-analysis ofublished functional imaging studies, suggesting thathere are distinct motor and cognitive maps in theerebellum (Schmahmann et al., 1998).The imprecise definition of the deep cerebellar nuclei

y even this high resolution MRI technique was some-hat disappointing and is most likely a reflection of

heir anatomic organization. The dentate nucleus isomposed of a relatively thin and convoluted layer ofeurons; the fastigial nucleus is a flattened plate ofeurons adjacent to the midline; and the globose andmboliform nuclei interposed between them have ratherndistinct borders. This methodological limitation ofhe current MRI technology was partially compensatedor by comparing the MRI sections with digitizedmages warped into Talairach space of postmortemryosections of another brain (Toga et al., 1997). The
se of these two complimentary techniques provided
ecessary additional detail of the cerebellar nuclearnatomy.In summary, we have used high resolution MR

mages and software that permits simultaneous cross-orrelation in the three cardinal planes to develop aetailed atlas of a human cerebellum in stereotaxicroportional space including three-dimensional repre-entations of the external surface and fissures. Thispproach substantially advances our understanding ofhe anatomy of the human cerebellum and should provealuable in future studies of its organization andunction.

ACKNOWLEDGMENTS

This work was supported in part by the National Library ofedicine (J.D.S.). Nikos Makris, Ph.D., and David Kennedy, Ph.D.,

rovided valuable input in the early stages of this project. Thessistance of librarians in the Rare Book Collections of the New Yorkcademy of Medicine and in the Countway Library of Medicine atarvard Medical School is appreciated. The assistance of Marygraceeal M.Ed. in the preparation of this manuscript is gratefullycknowledged. Presented in part at the 2nd International Conferencen Functional Mapping of the Human Brain, June 1996, Boston, MA.

REFERENCES

natomical Society of Great Britain and Ireland. 1933. Final reportof the committee appointed by the Anatomical Society of GreatBritain and Ireland on June 22, 1928. Robert Maclehose and Co.,Ltd., University Press, Glasgow. (Birmingham Revision of theBasle Nomina Anatomica; BNA-BR.).

ngevine, J. B., Mancall, E. L., and Yakovlev, P. I. 1961. The HumanCerebellum: An Atlas of Gross Topography in Serial Sections.Little, Brown and Co., Boston.riens Kappers, C. U., Huber, G. C., and Crosby, E. C. 1936.Comparative Anatomy of the Nervous System of Vertebrates, Includ-ing Man, Vol. 1. Macmillan, New York.

owden, D. M., and Martin, R. F. 1995. Neuronames brain hierarchy.Neuroimage 2:63–83.

radley, O. C. 1904. The mammalian cerebellum: Its lobes andfissures, part 1. J. Anat. Physiol. 38:448–475.

olk, L. 1906. Das Cerebellum der Saeugetiere. Bohn and Fischer, Jena.arpenter, M. B. 1976. Human Neuroanatomy. Williams & Wilkins,Baltimore, MD.ollins, D. L., Neelin, P., Peters, T. M., and Evans, A. C. 1994.Automatic 3D intersubject registration of MR volumetric data instandardized Talairach space. J. Comput. Assist. Tomogr. 18:192–205.ourchesne, E., Press, G. A., Murakami, J., Berthoty, D., Grafe, M.,Wiley, C. A., and Hesselink, J. R. 1989. The cerebellum in sagittalplane—Anatomic–MR correlation: 1. The Vermis. Am. J. Nerol.Res. 10:659–665.

rosby, E. C., Humphrey, T., and Lauer, E. W. 1962. CorrelativeAnatomy of the Nervous System, pp. 188–192. Macmillan, New York.eArmond, S. J., Fusco, M. M., and Devey, M. M. 1976. A Photo-graphic Atlas: Structure of the Human Brain. Oxford Univ. Press,New York.ejerine, J. 1901. Anatomie des Centres Nerveux, Tome 2. J. Rueff etCie, Paris.esmond, J. E., Gabrieli, J. D., and Glover, G. H. 1998. Dissociation offrontal and cerebellar activity in a cognitive task: Evidence for adistinction between selection and search. Neuroimage 7:368–376.ow, R. S. 1942. The evolution and anatomy of the cerebellum. Biol.
Rev. 17:179–220.

D

E

E

F

F

H

H

H

I

I

I

J

J

K

K

L

L

L

L

L

L

L

L

L

L

L

M

M

M

P

P

P

R

R

R

S

S

S

S

S

S

T

T

V

W

Z


oyon, J. 1997. Skill learning. In The Cerebellum and Cognition.(J. D. Schmahmann, Ed.) Int. Rev. Neurobiol. 41:273–296. Aca-demic Press, San Diego.

dinger, L. 1909. Uber die Enteilung des Cerebellums. Anatomischer.Anzeiger 35:319–323.

lliott, R., and Dolan, R. J. 1998. Activation of different anteriorcingulate foci in association with hypothesis testing and responseselection. Neuroimage 8:17–29.

iez, J. A., and Raichle, M. E. 1997. Linguistic processing. In TheCerebellum and Cognition (J. D. Schmahmann, Ed.) Int. Rev.Neurobiol. 41:233–254. Academic Press, San Diego.

latau, E., and Jacobsohn, L. 1899. Handbuch der Anatomie undvergleichende Anatomie des Centralnervensystems der Saeugetiere.I. Makroskopischer Teil. Karger, Berlin.enle, J. 1901. Grundriss der Anatomie des Menschen, Neu bearbeitetvon F. Merkel, 4th Aufl. Atlas. Friedrich Vieweg & Sohn, Braun-schweig.is, W. 1895. Die anatomische Nomenklatur. Nomina anatomica.Separat-Abzug aus Archiv fur Anatomie und Physiologie. Anato-mische Abteilung, Supplement-Band (Basle Nomina Anatomica of1895).olmes, C. J., Hoge, R., Collins, L., Woods, R., Toga, A. W., and Evans,A. C. 1998. Enhancement of MR images using registration forsignal averaging. J. Comput. Assist. Tomogr. 22:324–33.

ngvar, S. 1918. Zur Phylo- und Ontogenesae des Kleinhirns. FoliaNeuro-Biologica 11:205–495.

ngvar, S. 1928. The phylogenetic continuity of the central nervoussystem. In Studies in Neurology, Vol. I. Bull. Johns Hopkins Hosp.43:315–337.

nternational Anatomical Nomenclature Committee. 1956. NominaAnatomica. Revised by the I.A.N.C. appointed by the Fifth Interna-tional Congress of Anatomists held at Oxford in 1950. Submitted tothe Sixth International Congress of Anatomists, held in Paris, July1955. Williams & Wilkins, Baltimore.

akob, A. Das Kleinhirn. 1928. In (v. Mollendorff ’s) Handbuch dermikroskopischen Antomie des Menschen, Vol. 4, pp. 674–916.Springer, Berlin.

ansen, J., and Brodal, A. 1958. Das Kleinhirn. In (v. Mollendorff ’s)Handbuch der mikroskopischen Antomie des Menschen, Vol. 4, pp.1–323. Springer, Berlin.retschmann, H. J., and Weinrich, W. 1992. Cranial neuroimagingand clinical neuroanatomy. Thieme, New York.uithan, W. 1894. Die Entwicklung des Kleinhirns von Saugetieren,unter Ausschluß der Histogenese. Sitzungsbericht der Gesellschaftfur Morphologie und Physiologie, Munchen 10:89–128.

angelaan, J. W. 1919. On the development of the external form ofthe human cerebellum. Brain 42:130–170.

arsell, O. 1934. Morphogenesis and evolution of the cerebellum.Arch. Neurol. Psychiatry 31:373–395.

arsell, O. 1937. The cerebellum: A review and interpretation. Arch.Neurol. Psychiatry 38:580–607.

arsell, O. 1947. The development of the cerebellum in man inrelation to its comparative anatomy. J. Comp. Neurol. 87:85–129.

arsell, O. 1953. The anterior lobe of the mammalian and the humancerebellum. Anat. Rec. 115:341.

arsell, O. 1958. Lobules of the mammalian and human cerebellum.Anat. Rec. 130:329–330.

arsell, O. 1967. The Comparative Anatomy and Histology of theCerebellum from Myxinoids through Birds. (J. Jansen, Ed.) Univ.Minnesota Press, Minneapolis.

arsell, O. 1970. The Comparative Anatomy and Histology of theCerebellum from Monotremes through Apes. (J. Jansen, Ed.) Univ.Minnesota Press, Minneapolis.

arsell, O., and Jansen, J. 1972. The Comparative Anatomy and
Histology of the Cerebellum. The Human Cerebellum, Cerebellar
Connections, and Cerebellar Cortex. Univ. Minnesota Press, Minne-apolis.

orenson, W. E., and Cline, H. E. 1987. Marching cubes: A highresolution surface reconstruction algorithm. Comput. Graphics4:163–169.

oyning, Y., and Jansen, J. 1955. A note on the morphology of thehuman cerebellum. Acta Anatomica 25:309–318.adigan, J. C., and Carpenter, M. B. 1971. Cerebellum of the RhesusMonkey: Atlas of Lobules, Laminae, and Folia, in Sections. Univer-sity Park Press, Baltimore.alacarne, M. V. G. 1776. Nuova esposizione della vera struttura delcerveletto umano. Briolo, Torino.eckel, J. F. 1838. Manual of Descriptive and Pathological Anatomy.Volume II (Translated from German into French by A. J. L. Jourdanand G. Breschet. Translated from French into English by A. S.Doane.) E. Henderson, Old Bailey, London.

arsons, L. M., and Fox, P. T. 1997. Sensory and cognitive functions.In The Cerebellum and Cognition (J. D. Schmahmann, Ed.) Int.Rev. Neurobiol. 41:255–272. Academic Press, San Diego.

ress, G. A., Murakami, J., Courchesne, E., Berthoty, D., Grafe, M.,Wiley, C. A., and Hesselink, J. R. 1989. The cerebellum in sagittalplane—Anatomic–MR correlation: 2. The cerebellar hemispheres.Am. J. Neurol. Res. 10:667–676.

ress, G. A., Murakami, J., Courchesne, E., Grafe, M., and Hesselink,J. R. 1990. The cerebellum: 3. Anatomic–MR correlation in thecoronal plane. Am. J. Neurol. Res. 11:41–50.iley, H. A. 1929. The mammalian cerebellum. A comparative studyof the arbor vitae and folial pattern. Res. Publ. Assoc. Res. NervousMental Dis. 6:37–192.

iley, H. A. 1930. The lobules of the mammalian cerebellum andcerebellar nomenclature. Arch. Neurol. Psychiatry 24:227–256.

oberts, M. P., Hanaway, J., and Morest, D. K. 1987. Atlas of theHuman Brain in Sections. Lea & Febiger, Philadelphia.

chafer, E. A., and Symington, J. Neurology. 1908. In Quain’sElements of Anatomy, 11th ed., pp. 167–201. Longmans, Green &Co, London.

chmahmann, J. D., Hurwitz, A. S., Loeber, R. T., and Marjani, J. L.1998. A semi-flattened map of the human cerebellum. A newapproach to visualizing the cerebellar cortex in 2-dimensionalspace. Soc. Neurosci. Abstr. 24:1409.

chmahmann, J. D., Loeber, R. T., Marjani, J., and Hurwitz, A. S.1998. Topographic organization of cognitive functions in the humancerebellum. A meta-analysis of functional imaging studies. Neuro-Image 7:S721.

chwalbe, G. A. 1881. Lehrbuch der Neurologie. Eduard Besold,Erlangen.

mith, G. E. 1902. The primary subdivision of the mammaliancerebellum. J. Anat. Physiol. 36:381–385.

troud, B. B. 1895. The mammalian cerebellum. J. Comp. Neurol.5:71–118.

alairach, J., Tournoux, P. 1988. Co-Planar Stereotaxic Atlas of theHuman Brain. 3-Dimensional Proportional System: An Approachto Cerebral Imaging (Translated by Mark Rayport.) Thieme, NewYork.

oga, A. W., Goldkorn, A., Ambach, K., Chao, K., Quinn, B. C., andYao, P. 1997. Postmortem cryosectioning as an anatomic referencefor human brain mapping. Comput. Med. Imaging Graph.21:131–41.

icq-d’Azyr, F. 1786. Traite d’ anatomie et de physiologie, avec desplanches coloriees, Tome premier. Francois Ambrose Didot l’aıne,Paris.addington, M. M. 1984. Atlas of Human Intracranial Anatomy.Academic Books, Rutland, VT.

iehen, T. 1934. Centralnervensystem. Handbuch der Anatomie, pp.
1230–1289. Verlag von Gustav Fischer, Jena.

Three-Dimensional MRI Atlas of the Human Cerebellum in ... · ﬁssure on every slice in all three...

Documents

Transcript of Three-Dimensional MRI Atlas of the Human Cerebellum in ... · ﬁssure on every slice in all three...