Thoracic Human Vertebrae
Quantitative Three-Dimensional Anatomy
MANOHAR M. PANJABI, PhD,* KOICHIRO TAKATA, MD,t VIJAY GOEL, PhD;
DALE FEDERICO, MD,” THOMAS OXLAND, MASc,* JOANNE DURANCEAU, MS,*
and MARTIN KRAG, MD§
This study details the quantitative three-dimensional sur-
face anatomy of thoracic vertebrae based on a study of
1144 vertebrae. The thoracic spine was found tohave three
distinct regions: upper, middle, and lower segments. The
‘twoend segments appear to be transitional zones toward
‘cervical and lumbar regions. The middle zone (T3 to T9) is
Of utmost importance due to the presence of the combi-
nation of narrow spinal canal and critical vascular supply.
Means and standard errors of the means for linear, angu-
lar, and area dimensions of vertebral bodies, spinal car
pedicle, pars articularis, spinous and transverse pro-
esses, and rib articulations are provided for all thoracic
vertebrae. This information is necessary for constructing
accurate mathematical models of the human spine, It will
also provide a better understanding of the spine, and
allow for a more precise clinical and surgical mar
‘ment of spinal problems. [Key words: thoracic spine,
vertebrae, anatomy, vertebral dimensions, spinal canal,
pedicles}
‘vertebrae inthe various regions of the spine." However, often
these are aris’s drawings based on visual inspection of
cadavers." No quantitative geometric infoemation is provided, Such
information i necessary to build an anatomic database, which may be
used, among othe things, forthe development of accurate and valid
‘mathematiesl modes of the human sine. Mathematical models depic-
ing the thoracic spine have been used to study etiology, development
and teatment of soliosis, and the mechanical esponse ofthe spine to
external forces.*""84 The geometric data conceming vertebrae
‘sed these spinal models are generally of much lower quality than the
ication ofthe models themselves. As the models"
3 depends on the quality of the incorporated data, it
is essential that high-quality data sets should be availabe.
‘There are other important uses for a quantitative verbal geometric
data base. Physical properties of vertebrae are related to their size and
shape. The sizeof the spinal canal bas been found tobe related to the
incidence of low-back pain,® and size is affected by physiologic
stresses in childhood. The shape ofa lumbar vertebme—specifically,
the concavity inthe horizontal plane of its posterior walls been
Postulated to affect the frequency of disc herniation, The size and
‘orientation of pedicles are important dimensions fr the design of
transpedicular fixation devices, as well as for associated surgical
procedures. Several studies have been conducted in the past few Years
T= ‘ARE EXCELLENT TEXTS describing the shapes and ses of
om th ‘Depron of One and Rahbiaon, Yale Uae
Sool of idee, New Haven, Connect, te fUniety of Tocesimn,
“Tashi, lp, Depart of Dione Engeeing, Unter at
{twa lon iy on, nde become of Orbos te Ree
‘Sos, Uns of Vaan Balepton Verve
Sopot ng y Ni ast ANGIE
‘Ag frites 1990
with the primary goal of documenting the pedicle dimensions for
intemal fixation device application.‘ "695 These studies, however,
documented relatively few thoracic vertebral levels, and not all verte:
tral cimensions
‘The purpose ofthis paper is to report quantitative thee-dimensional
surface geomety ofthe thoracic spine, including clinically important
dimensions (e, linear, angular, and area measurements) from Tl to
‘T12 veriebrae. This is the first pape ina series of thee, the others
covering the cervical and lumbar regions.
MATERIALS AND METHODS
Fresh autopsy wholespine specimens, onsisting of C205, were
obtained The specimens were ideale by age, sx, beght, wei,
and cause of death (Table) Each vertebra wastage indicat spine
Specinen number andthe vertebral lvl. All muscles and thie
attachments were removed. The ligaments diss, and remaining sot
tases were softened by boiling. They were then subsequently i=
| mersedin leah (0. hou) end 75% eano (L hou), which cleaned
oft all he sft issue and divided the spine into individual vertebra.
“wo pointed, teaded rods, 3.0mm inameer nd 3.2em in length,
were threaded into the verbal Body ofeach vertebra, The rods were
cvientedso that one ly approximately in the onal plane and he other
in the mid-sagital plan (Figure 1). Ameropesterioe and ata aio
epi were taken iden the tis ofthe pins inthe focal anatomic
coordinate systemafthe vetebca with origin at pointC, the cetera the
‘pper endplate Figure 1).
Each anatomic prt was quaniied in ems of st of points, citer
Spanning a suf or forming a contour (igus 2). For exampl, each
ribarculation was represented by upto 2 points (Ave rows thes t9
Tour points), andthe pedicle was dained by pins located cramer
ently. The numberof points chosen in ast was such that the surface
seometry was defined reasonably well by the saigh ns joining the
ace points. Each vertebra was marked withthe poins tobe ste.
“he ee dimeesionl coordinates ofl the points were obtied with
the use ofthe morphometer an instument described below. The
vertebral parts whose gometc characteristics Were studied are sted
in Table 2, ogetier with the omenclatr used ident tem.
‘The Three-Dimensional Morphometer. Ths specially designed
ingrumeat consisted of one linear variable differential transformer
(LYDT) and wo angular rotary variable differential transformer
(RVDT) displacement transducers." Tho transducers were soaranged
that thelr snes, meeting at a point, established a spherical coordinate
system. The movement of the morphometer pointer tip produced
voltage changes proportional to i location in space. The instrument
‘was connected to = computer (Digital Equipment Company, Boston,
‘Massachusetts; 11/03) through a 12-bit A/D converter to record the
coordinate dat
"To measure the thre-dimensional location ofa pont on a vertebra,
the following procedure was adopted, with the specimen held inthe
‘morphomete stand. Fist, the thre tps of the threaded rods were
touched ina given order, recording their spatial coordinates. The the
point tobe digitized was touched and its coordinates recorded. Com-
puter programs were developed to conver these voltages into CartesianTHORACIC HUMAN VERTEBRAE + PANJABI ET AL
‘Table 1. Spine Specimen Data
‘Specimen No. Age. Sex Weight (kg) Height (om) (Cause of death
3 8 M 70 178 Coronary artery disease
84 @ F 0 187 Breast carcinoma
a 53 M 54 167 Esophageal carcinoma
59 19 F 83 188 ‘cule lymphocytic leukemia
810 “4 F 6 181 Fight tung carennoma
313 56 M 1 187 onal fallure
314 56 M 4 7 Pancreatic carcinoma
515 59 M 34 im Myocardial infarction
820 49 F 7 1ST Third degree burns
522 53 M a8 172 Carcinomatosie
823 57 M 5 170 Myocardial infarction
S24 20 M 85 170 Subdural hemormage
‘coordinates with respect toan axis sytem defined by the tps ofthe three
pins. The largest error ofthe instrument in locating 2 point on the
Verebra was 0.5 mm.
Data Collection and Analysis. The vertebra to be studied was
rigidly secured in the morphometr stand. The three tps of the threaded
pins and each of the poins on the vertebra were touched with the
‘morphometer pointer n'a predetermined sequence, and thle positions
recorded, Som ofthe anatomie points (eg, the lower endplate) could
‘not be recorded from the intial position of the vertebra. In such @
situation, the vertebra was tumed upside-down and fixed in a new
position. Inaddltion othe anatomic points, the ips ofthe threaded pins
‘were redigtized to allow for proper spatial orientation, The coordinate
teansformation, determined from the wo positions of the tips of the
three threaded pins, made it possible o transform he coordinates ofthe
anatomic points in the second postion to the coordinates of the
anatomic points obtained inthe first position.
‘The first step in data analysis was to transform the digitized points for
cach vertebra to their own local anatomie coordinate sytem. Points A,
B, D, and E define respectively the right superior, the left superior, the
right inferior, and the left inferior edges of the posterior wall of the
veribral body (Figure 1). These four points were chosen as the
reference points because they were consistent and showed minimum
variation in the presence of osteophytes. Using the least-squares
method, the bestfit plane passing through these points was determined.
‘This defined the frontal plane. Point C was the centroid ofthe superior
endplate, and formed the origin of the local anatomic coorinate
system. A line perpendicular tothe fronal plane formed the z-axis. A
line paalet to line AB and pointing tothe left formed the x-axis.
Finally, the y-axis was defined a a line perpendicular othe z-x plane.
‘The positive diteztons of the thee axes ae shown in Figure I. Three
types of anatomic parameters (linear, angular, and surface area mea
surements) were computed.
Linear Parameters. The necessary computer software was devel-
opedto proces the coordinate datato yield well defined linear ansiomic
mensions. For example, the endplate was characterized by its width
(EPW) and depth (EPD). The esl ofall the specimens were pooledto
caleulate the mean and standard error ofthe mean foreach parameter.
Angular Parameters. Orientation of plane was determined either
by a unit vector perpendicular to the plane or by angles made with
respect o each ofthe three anatomic planes. A plane may be defined by
minimum of three noncolinear points. Since in the present study each
plane was characterized by more than thre points, eg, the rib ate
‘ion was characterized by upto 20 points, te least-squares technique
‘was used to yield the plane tha represented the best fi
“To provide an alternative measure ofthe end-pate angles mensured
herein, a parallel radiographic study was conducted. The end-plate
angles with espectto the transverse plane (EPI, EPIt) were measured
‘on ten lateral radiographs ofthe thoracic spine.
‘Area Parameters. The projected areas ofthe digitized part onto the
three planes ofthe local anatomic system were computed, The surface
oreross-sectional area was computed asthe square root of the sum ofthe
squares ofthe three projected areas. Mean and standard error of the
lean were then computed.
Statistical Analysis. Stasticl analysis using single-actor analysis
of variance (ANOVA) and Schefe'sF test were used to compare the
significant differences between parameter values as functions of verte-
bral level. A statistical significance level of 5% was used throughout
‘Accuracy of Measurements The accuracies in computing linear,
angular, and aea parameters were studied using an object with known
values A Plexiglas piece with sides cut at 30", 45", and 60" was
constrcted. A shet of milimeter graph pape, wit different shapes
(ciangles, rectangles, circles, ec.) dravn oni, was pasted on each of.
‘he cut surfaces. The pins (matkers) and three reference points similarto
those on a vericbra were identified on the Plexiglas piece, Thus
‘seudo-verteba with surfaces of known inclination and pars of known,
A) auRUAOIOP
01 powuoyod sem (ynONY] souEUE jo sisKeue Ue HojowieIEd Yous 105 sUBEL ot J 408 pIEDLE}S aul BOYS MOL NO} 4) SIR 'SUBBU a4 EMOYS mos Joddn OU YOIOUIRIEG YEO 104 °2/ON
ar ae ot ui Orh)SSCSSSCSSCSSC 1
es ve eZ so ru. ree 86 aL sor 6 zt gob moa
oor oro v0 ite eee esi ret igh rie erie | eee et
ey 6a 9 8 va eo oe 98 v9 ve su oy aga
oz zoe eee ie ogee tz aw nee wast
si ze vor zor coh Mt veLS LC eC ew ad
ore er re OOo 0 cet wr aiel eve cl ee ar
08 ez vo eo 6 szz eevee ee ween sig
(.) suojsuawp semnbuy
ovat es es (Oui Gre) Orel Ore ciel cue |= orel tonya = 00H)
1018 oli) £98 819 sis oz Sve gee week tht oa
cost SL 97 ws we use LL ewe 7)
(6-1) 608 20 v98 cy Cr a nT eT soa
(qui) eave yeuopoee-s8019
sr “a ot st sr ez 6 oz rs vt st 1 mas/Hos
6 et Br sr et at oz sr zz. wl et + aNd
¥60 sco veo sso WO zDD Dev
Meo eC OViN eae nO Ue ach ee I ev ies ures irr tien) Maa
evo ero e000 cg
west Giewsy (eer see sik otk zk es sHad
20 fonts ct YO) 90 0) crue ayo, isuee ecu)
oo (usu) £08 ze 29 zs 09 zs oz 59 6L ze waa
0 evo. PaO ee) O00) ee er cru) ogi rr)
88 8 aL “9 so 09 29 ss 02 ve ze aad
(uu) suojeuownp 2e9u
au ba oun aL aL aL on sn 1 aL a 4
7297 6198194 o708I0U
{eumjejouewon 20) @ ey4e], pue ¢ 6mnbiy 09g) suoysuOUNTHORACIC HUMAN VERTEBRAE + PANJABI ET AL 995
-{sesoyivored w unoys) aroun Aqueous ox0m sana} sgoUOA Yow (S0'0 >
Jeue ue “oowered yoes Joy sueNU By jo JOU PlepUEIS oY SMOUS MO: JOMD| aM 2
01 pounoyad sem (yAONY) BOUEIEA 05
wer eek ee
(ewes zea ces | ees ¥09 98
ot st Orbe (cries: | 6 eb oz
rip oSy els eza 08 oes His sus
(wu) suojsuoup 2290,
seo ez be cova gut - -
095 = Out ove 086 095 =e oe = = a
ve bee ost by 96 BL cL = =
093 oSh oor oe = (iL'enoes = o6e oan
- mia
jwojoee-4019,
zh oun a aL a 1 on a aL a
ono exgabeA BSeIOUL
Z alge] ue & oinbig 60g) eoeseDo!g esTonsaRI| BUR ‘enoUldg Sre|nOnaW-eNUT Sea “L
mL
quosed u unoys) wosoyp hyueoyutis 220M sono jexgauan YoU (S00 > d)
‘SmoYS mo! Jamo] au) oil 'SuLaW aul swoys mol Jedd Bip “eIoWEIEA Uses
01 pounoyied sem (yAONY) eauEyeR Jo SKeUe Ue sojowrered 4289 104 sUEBW ay Jo JOM POPU
sys woe el gue sealer yO ozer YO + ¥AO
= 198 ose. 199 ere. ue 108 ot,
er er zor cis 265 ess os9 mo
a uh ay sev ez, oe. ovr, zr
9 ‘gs ey eis z19 si9 cer swig
au 204 je
S99 oe eee 2 9 sos “is fs wo
= we oe ory. ws ez ay a
= ee 06 vee 19 00 Gn ‘gi wr
= 9 is? 389. str. we sez, ee,
= 08 Lor vee 292 re se iz aK
oze eo 98 399. wee. zs ee. we.
eur eis ves vis lr 690 ave ‘998 no
ue 6s v7 209) iss we cee, 189.
O19 (rae 98 Ly ves, ‘zp ele 292 ep anya
owe uoneinojue jeuqop2n0}s09,
zu ba oun aL a a ot aL aL aL a A
18197 FIgeHeA ORIOL
]DUBLION 10} O19@1 Bu 6 eInBLg 60g) SUO|SUOWIG SeEse00Id PUN GH “ODL896 SPINE + VOLUME 16 + NUMBER 8
Po] Seo
: ae a
3 EB ..
8 2. 2 epwueroy
& = ie epmero
: San ow
5 eae i
Be Bw.
c VERTEBRAL LEVEL . VERTEBRAL LeveL
Fig 4. Vertebral enc-plate dimensions as functions of vertebral
eng-pales.
indicated by whole nes, Dashedines depict the end-plate areas,
ofthe posterior wal ofthe vertebral body.
B, Ratio of enc-pale width lo depih fr the upper and lower end-pales,
levels Tt to T12 are shown, A, Linear dimensions ofthe upper and lower
, Surface areas of upper and lower end-plales are
‘assuming ne endplate shape tobe etter an lips ora langle, Heights
i