Clinical validation of coronal and sagittal spinal curve measurements
based on three-dimensional vertebra vector parameters
Szabolcs Somoske€ oy, MD*, Mikl? os Tunyogi-Csap? o, MD, PhD, Csaba Bogy? o, MD,
Tam? as Ill? es, MD, PhD, DSc
Department of Orthopedic Surgery, Institute of Musculoskeletal Surgery, University of P? ecs Clinical Center, 1 Ak? ac utca, P? ecs, Hungary H-7632
Received 31 July 2011; revised 13 February 2012; accepted 25 August 2012
AbstractBACKGROUND CONTEXT: For many decades, visualization and evaluation of three-
dimensional (3D) spinal deformities have only been possible by two-dimensional (2D) radiodiag-
nostic methods, and as a result, characterization and classification were based on 2D terminologies.
Recent developments in medical digital imaging and 3D visualization techniques including surface
3D reconstructions opened a chance for a long-sought change in this field. Supported by a 3D Ter-
minology on Spinal Deformities of the Scoliosis Research Society, an approach for 3D measure-
ments and a new 3D classification of scoliosis yielded several compelling concepts on 3D
visualization and new proposals for 3D classification in recent years. More recently, a new proposal
for visualization and complete 3D evaluation of the spine by 3D vertebra vectors has been intro-
duced by our workgroup, a concept, based on EOS 2D/3D, a groundbreaking new ultralow radiation
dose integrated orthopedic imaging device with sterEOS 3D spine reconstruction software.
PURPOSE: Comparison of accuracy, correlation of measurement values, intraobserver and inter-
rater reliability of methods by conventional manual 2D and vertebra vector–based 3D measure-
ments in a routine clinical setting.
STUDY DESIGN: Retrospective, nonrandomized study of diagnostic X-ray images created as part
of a routine clinical protocol of eligible patients examined at our clinic during a 30-month period
between July 2007 and December 2009.
PATIENT SAMPLE: In total, 201 individuals (170 females, 31 males; mean age, 19.88 years) in-
cluding 10 healthy athletes with normal spine and patients with adolescent idiopathic scoliosis (175
cases), adult degenerative scoliosis (11 cases), and Scheuermann hyperkyphosis (5 cases). Overall
range of coronal curves was between 2.4 and 117.5?. Analysis of accuracy and reliability of mea-
surements was carried out on a group of all patients and in subgroups based on coronal plane de-
viation: 0 to 10?(Group 1; n536), 10 to 25?(Group 2; n525), 25 to 50?(Group 3; n569), 50 to 75?
(Group 4; n549), and above 75?(Group 5; n522).
METHODS: All study subjects were examined by EOS 2D imaging, resulting in anteroposterior
(AP) and lateral (LAT) full spine, orthogonal digital X-ray images, in standing position. Conven-
tional coronal and sagittal curvature measurements including sagittal L5 vertebra wedges were
determined by 3 experienced examiners, using traditional Cobb methods on EOS 2D AP and
LAT images. Vertebra vector–based measurements were performed as published earlier, based on
computer-assisted calculations of corresponding spinal curvature. Vertebra vectors were generated
by dedicated software from sterEOS 3D spine models reconstructed from EOS 2D images by
the same three examiners. Manual measurements were performed by each examiner, thrice for ster-
EOS 3D reconstructions and twice for vertebra vector–based measurements. Means comparison
t test, Pearson bivariate correlation analysis, reliability analysis by intraclass correlation coefficients
FDA device/drug status: Approved (EOS 2D/3D orthopedic imaging
system; and sterEOS 3D workstation software for surface 3D reconstruc-
tion of the spine in adults, adolescents, and children).
Author disclosures: SS: Nothing to disclose. MT-C: Nothing to dis-
close. CB: Nothing to disclose. TI: Speaking/Teaching Arrangements: Al-
phaTechSpine SA (C), SpineGuard SA (B); Endowments: SpineGuard SA
(B), AlphaTech SA (C).
The disclosure key can be found on the Table of Contents and at www.
* Corresponding author. Department of Orthopedic Surgery, Institute
of Musculoskeletal Surgery, University of P? ecs Clinical Center, 1 Ak? ac
utca, P? ecs, Hungary H-7632. Tel.: (36) 72-536841; fax: (36) 72-536840.
E-mail address: firstname.lastname@example.org (S. Somoske€ oy)
1529-9430/$ - see front matter ? 2012 Elsevier Inc. All rights reserved.
The Spine Journal - (2012) -
for intraobserver reproducibility and interrater reliability were performed using SPSS v16.0
RESULTS: In comparison with manual 2D methods, only small and nonsignificant differences
were detectable in vertebra vector–based curvature data for coronal curves and thoracic kyphosis,
whereas the found difference in L1–L5 lordosis values was shown to be strongly related to the mag-
nitude of corresponding L5 wedge. Intraobserver reliability was excellent for both methods, and
interrater reproducibility was consistently higher for vertebra vector–based methods that was also
found to be unaffected by the magnitude of coronal curves or sagittal plane deviations.
CONCLUSIONS: Vertebra vector–based angulation measurements could fully substitute conven-
tional manual 2D measurements, with similar accuracy and higher intraobserver reliability and in-
terrater reproducibility. Vertebra vectors represent a truly 3D solution for clear and comprehensible
3D visualization of spinal deformities while preserving crucial parametric information for vertebral
size, 3D position, orientation, and rotation. The concept of vertebra vectors may serve as a starting
point to a valid and clinically useful alternative for a new 3D classification of scoliosis. ? 2012
Elsevier Inc. All rights reserved.
Keywords:Vertebra vectors; EOS 2D/3D; Interrater reliability; Coronal and sagittal curve measurements; Scoliosis curve
For many decades, the term of scoliosis has been consid-
ered analogous to the lateral (LAT) deviation of the spinal
column as depicted in anteroposterior (AP) radiographs. Vi-
sualization and evaluation of spinal deviations have only
been possible in a two-dimensional (2D) fashion by con-
ventional X-ray radiography, depicting the spinal column
in AP and, less frequently, LAT projections. As a result,
characterization and classification of spinal deformities
have been based on 2D terminologies, as well [1–3].
This situation started to change in recent years, thanks to
developments in medical X-ray imaging and visualization
by three-dimensional (3D) reconstruction techniques [4–6].
More recently, a new ultralow radiation dose–integrated
2D/3D orthopedic radioimaging system based on a Nobel-
prize winning X-ray detection technology represented
a breakthrough in this field . A significant benefit of us-
ing this new system is related to its ability of achieving su-
periorimage quality while
considerably lower radiation doses compared with conven-
tional computed radiography (CR) or digital radiography
systems or for obtaining 3D reconstructions, computed to-
mography examinations . A recent publication reported
on dosimetry measurements of entrance skin dose ratios
of 2.9 to 9.2 depending on the anatomic location—all in fa-
vor of the EOS system compared with a standard CR sys-
tem. Notably, average dose values by the EOS system
were between 0.11 and 0.30 mGy, whereas doses by
a CR system for identical images ranged between 0.59
and 2.47 mGy . Other unique advantages of the system
include its ability of simultaneously capturing full body
standing orthogonal AP and LAT radiographs, allowing
an accurate and realistic 3D reconstruction of the skeletal
system, including the spine and pelvis.
According to these developments, the definition of sco-
liosis is being transformed into a true 3D entity, exhibiting
spinal deformities involving deviations in all 3D—an LAT
translation of vertebrae in coronal, mainly a lordotic devi-
ation of the normal curvature in sagittal, and very character-
istically for scoliosis, a vertebral axial rotation and torsion
in transverse (or horizontal) plane.
In clinical practice, spinal geometry is traditionally char-
acterized by 2D angulation values as determined by various
2D curve measurement methods. Among various conven-
tional measurement methods, the original Cobb method for
coronal curves and a modified Cobb method for sagittal cur-
vature became most widely used and accepted for clinical
routine, thanks to their simplicity and relatively easy repro-
duction. These methods have been shown to exhibit good in-
traobserver reliability and good-to-excellent interrater
cause a relatively greater interrater variability has been
shown related to variable definition of the end vertebrae of
spinal curves and because of drawbacks from the essential
problems of interpreting measurements for 3D objects like
vertebral end plates based on projections in 2D planes .
Nevertheless, routine evaluation and classification of spinal
morphology remained predominantly 2D [12,13].
Increasing concern of diagnostic and surgical conse-
quences due to limitations of 2D terminology used for
a 3D entity has compelled Scoliosis Research Society
(SRS) to propose a 3D terminology of spinal deformity
and emphasize the need for a new 3D classification based
on new 3D diagnostic techniques .
As a result of several years of work along these princi-
ples, new proposals for a 3D visualization and 3D classifi-
cation of adolescent idiopathic scoliosis (AIS) have been
published recently [15–17], including those by members
of the SRS 3D Classification Committee about evaluation
and schematic visualization of the spine by auxiliary planes
of maximum curvature (PMCs) using the so-called ‘‘da
Vinci representation’’ [18–20].
2 S. Somoske€ oy et al. / The Spine Journal - (2012) -
A new concept of vertebra vectors has been introduced
recently by our workgroup to visualize and evaluate spinal
column deformities in a truly 3D manner . A vertebra
vector is, by its definition, a replacement for a real vertebra
while preserving crucial information for its size, position,
orientation, and rotation in 3D. Vertebra vectors represent-
ing the full spine are placed within a coordinate system that
is based on an individually calibrated scale, provide simple,
well-established mathematical ways for quantitative evalu-
ation of 3D spinal geometry. Methods for measuring con-
ventional coronal and sagittal spinal curvatures, as well as
calculation of vertebral axial rotation, have also been de-
scribed in this article, allowing a complete 3D characteriza-
tion of the spine.
The present study was aimed at the validation of the con-
cept of vertebra vectors in its ability for accurate and reli-
able characterization of spinal curvature in coronal and
sagittal planes. The accuracy, reliability, and reproducibility
of measurements based on vertebra vectors are reported,
compared with measurements based on conventional 2D
Materials and methods
In total, 201 patients (170 females, 31 males; mean age,
19.88 years) were included in this nonrandomized retro-
spective study who were referred to our Department Outpa-
tient Clinic within a 30-month period between July 2007
and December 2009. Ten healthy athletes with normal
spine (with coronal deviation smaller than 5?) and patients
with AIS (175 cases), adult degenerative scoliosis (11
cases), and Scheuermann hyperkyphosis (5 cases) were en-
tered as study subjects. Range of coronal plane deviations
varied between 2.4 and 117.5?. Analysis of accuracy and
reliability of measurements was carried out either on all pa-
tients as a whole or in patient groups based on their coronal
plane deviation as follows: 0 to 10?(Group 1; n536), 10 to
25?(Group 2; n525), 25 to 50?(Group 3; n569), 50 to 75?
(Group 4; n549), and above 75?(Group 5; n522).
Each subject was put through a routine clinical protocol.
Full body standing orthogonal AP and LAT X-ray images
were created by an EOS 2D/3D system (EOS Imaging,
Paris, France), using a standardized arm positioning that
was reported for optimized sagittal visualization of cervical
and thoracic spine regions . Digital images were stored
in the institutional Picture Archiving and Communication
System network (Aspyra AccessNet v6.2, Aspyra LLC,
Jacksonville, FL, USA). Conventional curve measurements
were performed manually on digital AP and LAT EOS 2D
images, using standard Picture Archiving and Communica-
tion System workstation software tools (Aspyra AccessNet
MedView v6.2; Aspyra LLC). Coronal curves were mea-
sured according to the Cobb method. To avoid the most im-
variability, fixed, predefined end vertebrae were used for
each case agreed on by examiners. For sagittal curve
manual 2D measurements, T4–T12 kyphosis and L1–L5
lordosis were determined, according to the modified Cobb
method, using LAT EOS images and the same software
tools as mentioned previously. L5 vertebral wedge in sagit-
tal plane was also determined manually in 2D on LAT EOS
images by determining the angle between lines correspond-
ing to the upper and lower end plates of the L5 vertebra.
All manual 2D measurements were performed thrice by
three independent experienced examiners. Operator bias
from learning was controlled by random selection of cases
for each measurement set, and patient identities were
masked from observers.
Surface 3D reconstructions of vertebrae in thoracic and
lumbar regions were created using sterEOS 3D workstation
software version 22.214.171.12440 (EOS Imaging, Paris, France),
performing the ‘‘full 3D’’ procedure using AP and LAT
EOS 2D images . Three-dimensional reconstructions
were carried out by the same three experienced examiners
as for manual 2D measurements and repeated once for each
case. Completed sterEOS 3D reconstructions were used for
vertebra vector generation.
Vertebra vectors were generated by a method based on
EOS 3D reconstructions as recently described in detail
. In summary, vertebra vectors for T1–T12 and L1–L5
vertebrae were produced and visualized by a dedicated soft-
ware utility after sterEOS 3D reconstruction was finished. A
3D coordinate system for vertebra vectors was created with
a scale based on the individual interacetabular distance of
each patient. The vertebra vector generation software utility
was developed in MATLAB (MathWorks, Inc., Natick, MA,
USA) with an ability to directly query 3D coordinate data of
vertebra vector–specific elements stored inside sterEOS 3D
models, as well as to visualize generated vertebra vectors in
coronal, sagittal, and horizontal plane projections. Vertebra
vector–associated parameters—coordinates X, Y, Z for left/
right pedicular centroids, initial point A and terminal point
B of vertebra vectors—were recorded and angulation values
calculated as follows. For coronal curves, frontal plane pro-
jections of vertebra vectors were used and the angle formed
by lines connecting the left and right pedicular centroids of
upper and lower end vertebrae was calculated (Fig. 1, Top).
For sagittal curve values, sagittal plane projections of verte-
bra vectors were used and angles between lines correspond-
ing to vertebra vector T4, T12 for thoracic kyphosis (Fig. 1,
Middle) and L1, L5 for lumbar lordosis (Fig. 1, Bottom),
respectively, were computed.
Means comparison t test, Pearson bivariate correlation
analysis, reliability analysis by calculating intraclass corre-
lation coefficients (ICCs) for intraobserver reproducibility
and interrater reliability were all performed using a statisti-
cal software package (SPSS v16.0; IBM Corp., Armonk,
NY, USA). Reliability analysis was based on the alpha
two-way random model with consistency (for intraobserver
ICC) or absolute agreement (for interrater ICC). Statisti-
cally significant results were accepted as valid with signif-
icance of p!.05 or below.
3 S. Somoske€ oy et al. / The Spine Journal - (2012) -
A representative clinical case of a 17-year-old girl with
AIS comprising a right convex major thoracic curve (Lenke
Type 1A) is demonstrated in Fig. 2.
EOS 2D AP and LAT X-ray images (panels A and B),
sterEOS 3D model superimposed on respective AP and
LAT X-rays in coronal and sagittal plane views (panels C
and D), and corresponding vertebra vectors in coronal
and sagittal plane views (E and F) are shown. sterEOS
3D top view image of the reconstructed 3D spine model
with the pelvis and a corresponding horizontal plane view
of vertebra vectors with acetabula are visualized in panels
G and H, respectively.
Results of intraobserver reproducibility of curve mea-
surements performed by the three examiners for all patients
as calculated by intraobserver ICC values are demonstrated
in Table 1. Manual 2D and vertebra vector–based 3D mea-
surements were found to be performed by each examiner at
Accuracy and correlation analysis results of curve mea-
surements based on vertebra vectors compared with manual
2D methods are shown in Table 2, presenting only nonsig-
nificant, minute differences for coronal curve and kyphosis
values (?0.02?and 0.57?) but a significantly smaller value
for lordosis by 9.03?. Relationship between measurement
values by the two methods was found to be strongly
Interrater reliability of curve measurements for all pa-
tients as computed by interrater ICCs is demonstrated in
Table 3. Reliability of manual 2D measurements were
good-to-excellent (ICC values of 0.971, 0.844, and 0.845
for coronal curves, thoracic kyphosis, and lumbar lordosis,
respectively) but were surpassed by significantly higher
(0.991, 0.982, and 0.971, respectively).
Results of accuracy and correlation of curve measure-
ments within patient groups are presented in Table 4. Differ-
ences compared with corresponding manual 2D values were
found to be small and nonsignificant, except for lordosis
parameters in all patient groups. Pearson correlations be-
tween corresponding measurements were medium to strong
and statistically significant, except for coronal curve param-
eters in Group 1 (?0.055; p5nonsignificant) and lordosis
parameters in Group 5 (0.305; p5nonsignificant).
Fig. 3 shows results of interrater reliability analysis of
curve measurement within patient groups. Interrater ICC
values for vertebra vector–based measurements were found
to be excellent, ranging from 0.994 to 0.915 for coronal
curves, from 0.987 to 0.978 for kyphosis, and from 0.985
to 0.961 for lordosis, in patient Groups 1 to 5. These values
were consistently higher than those for manual 2D in every
patient group, and unlike their 2D manual counterparts,
their values did not seem to significantly change with the
increase of scoliotic curve magnitude.
Fig. 1. Angulation measurement methods based on vertebra vectors for
(Top) a thoracolumbar curve between vertebrae T6–L1, (Middle) T4–
T12 thoracic kyphosis, and (Bottom) L1–L5 lumbar lordosis.
4 S. Somoske€ oy et al. / The Spine Journal - (2012) -
Fig. 2. A representative adolescent idiopathic scoliosis case visualized by EOS 2D, sterEOS 3D, and vertebra vectors. (A, B) EOS 2D anteroposterior, lateral
X-rays. (C, D, G) sterEOS 3D model in coronal, sagittal, and horizontal plane views. (E, F, H) Vertebra vectors in coronal, sagittal, and horizontal plane
5 S. Somoske€ oy et al. / The Spine Journal - (2012) -
Relationship between L5 wedge angles and differences
in lumbar lordosis measurement values are presented in
Fig. 4, documenting only small and statistically nonsignif-
icant disparities, either for all patients as a whole or within
patient Groups 1 to 5.
Characterization of spinal column geometry is tradition-
ally based on measurements of spinal curvature in coronal
and sagittal planes. In addition to simplicity and compre-
hensiveness, accuracy, reliability, and reproducibility are
very important features of any measurement method to be-
come a standard in clinical practice. The Cobb method for
coronal curves and the modified Cobb method for sagittal
curvature undoubtedly meet these requirements, although
they are fundamentally 2D measurements carried out on
plain X-ray images and, as a consequence, are laden with
problems and errors in faithfully translating true character-
istics of a 3D object into values based on 2D projections
[1,11]. In addition, when end vertebrae of scoliotic curves
are not pre-agreed on observers, only poor-to-moderate in-
terrater reproducibilities are reported for coronal measure-
ments . Replacing manual measurement methods with
computer-assisted digital measurement tools did not im-
prove interrater reproducibility .
and sagittal curves by vertebra vectors is presented, com-
pared with standard manual 2D measurements based on the
Cobb method. Accuracy of conventional spinal angulation
values was found to beverygood forcoronal curvesand tho-
racic kyphosis. Although L1–L5 lumbar lordosis values
showed differences compared with values obtained with
the modified Cobb method, the overall discrepancy of
9.03?was within reported values of interobserver variations
for lumbar lordosis measurements . Moreover, any dis-
crepancy was shown to be strongly related to the magnitude
of corresponding L5 vertebral wedge. Intraobserver reliabil-
ity and interrater reproducibility were shown to be both
ing conventional manual methods.
Accuracy of any vertebra vector–related parameters di-
rectly depend on a highly precise registration of 3D land-
marks that determine a vertebra vector, ie, pedicular
centroids, vector initial and terminal points. In this case,
this problem was overcome by performing the ‘‘full 3D’’
procedure of sterEOS 3D reconstruction, which has been
shown to be very precise compared with 3D reconstructions
based on computed tomography [23,26] and readily fur-
nishes 3D information for vertebra vector data. As a result
of that, our measurements based on vertebra vectors were
shown to be just as reliable and reproducible when per-
formed on cases with severe scoliotic curves.
manual 2D coronal and sagittal plane curvature measure-
ments, with similar accuracy and higher intraobserver reli-
ability and interrater reproducibility. Positioning of vertebra
vectors inside vertebral bodies at pedicular centroid level
and along with the sagittal symmetry axis of the vertebral
body may create a measurement base with inherently less
Intraobserver reproducibility of curve measurements for all patients
Manual 2DVertebra vectors 3D
Mean6SD ICC* Mean6SD ICC*
Examiner no. 1
Examiner no. 2
Examiner no. 3
2D, two-dimensional; 3D, three-dimensional; SD, standard deviation;
ICC, intraclass correlation coefficient.
* Two-way random model with consistency used; for all results,
Accuracy and correlation of curve measurements for all patients
Vertebra vectors 3D
Vertebra vectors 3D
Vertebra vectors 3D
SD, standard deviation; 2D, two-dimensional; 3D, three-dimensional.
* Means shown are average values of examiners averages for measure-
yp Value not significant.
Interrater reliability of curve measurements for all patients
Manual 2DVertebra vectors 3D
Lower UpperLower Upper
T4–T12 kyphosis 0.844
2D, two-dimensional; 3D, three-dimensional; ICC, intraclass correla-
tion coefficient; CI, confidence interval.
95% CI: 95% upper and lower boundaries for confidence interval of
* Single measures, two-way random model with absolute agreement
used; for all results, p!.001.
6S. Somoske€ oy et al. / The Spine Journal - (2012) -
variability and better reproducibility than conventional
methods based on vertebral upper and lower end plates.
Horizontal plane axial rotation of vertebrae could also be
directly and accurately evaluated by calculation of vertebra
vector angle aHas shown earlier . Visualization of the
spinal column by vertebra vectors provides a truly 3D
method to evaluate scoliotic deformities in 3D, conveying
clear views to orthopedic surgeons that are readily accessi-
ble, backward-compatible with traditional images, and are
easy to understand, especially in horizontal plane. Vertebra
vector parameters and values calculated from 3D vector co-
ordinates preserve all important information of vertebrae,
including size, 3D position and orientation, and rotation.
The 3D coordinate system using an individual scale based
on patient-specific geometry creates a unique possibility
for interindividual comparisons, assessment of surgical
results by preoperative/postoperative comparison, and
follow-up of individual progression. Preliminary data sug-
gest that numerical characterization of different scoliotic
curve types by vertebra vector parameters could easily dif-
ferentiate between clinically different configurations, with-
out unnecessary and difficult abstractions (Somoske€ oy S
et al., 2012, unpublished data).
A report presented by The Working Group on 3D Termi-
nology of Spinal Deformity of SRS in 1993 summarized the
approach to be adopted to create a new terminology that en-
deformities in3D .Thisapproach included development
of ‘‘quasi-3D’’ measurements that overcome the difficult
challenge of 3D visualization of the spine. This is to be
achieved by making extensive use of auxiliary planes in
which spinal curves are to be projected and analyzed. It has
resented a reasonable compromise between mathematical
purity and conceptual/practical limitations, ones that origi-
nal geometry into curves projected in those auxiliary planes.
The ability to visualize and analyze spinal deformities
by truly 3D methods might be the only valid approach
for a fundamental understanding of underlying 3D mecha-
nisms in scoliosis. Routine 2D diagnostical methods lack-
ing insights into horizontal plane events and current 3D
surgical correction strategies emphasizing deformity cor-
rection by vertebral axial derotation are in stark contrast.
This issue is convoluted by a current scoliosis classification
system that is based on coronal and sagittal assessment
only, completely ignoring specific and important deviations
in the horizontal plane. A new, clinically valid 3D classifi-
cation system needs to be based on new clinical imaging
modalities with optimally automated 3D visualization and
3D analysis that may reveal a significantly novel, mathe-
matically described relationship among elementary steps
leading to different types of scoliotic deformities. A case
representing a certain class in this new 3D classification
should not only be clearly differentiated from a case of
Accuracy and correlation of curve measurements within patient groups
Group 1 (n536)
Group 2 (n525)
Group 3 (n569)
Group 4 (n549)
Group 5 (n522)
2D, two-dimensional; 3D, three-dimensional.
Group 1: Cobb 0 to 10?; Group 2: Cobb 10 to 25?; Group 3: Cobb 25
to 50?; Group 4: Cobb 50 to 75?; Group 5: Cobb O75?.
2D: manual 2D measurements; 3D: 3D measurements based on verte-
* Means shown are average values of examiners averages for measure-
yp Value not significant.
7S. Somoske€ oy et al. / The Spine Journal - (2012) -
another class by its differing 3D phenotype, the distinction
should have a direct impact on the subsequent clinical treat-
ment, as well. This includes surgical strategies being used
(ie, anterior vs. posterior approach, vertebral translation
in 3D with or without expressive axial derotational
maneuvers, length of fused segments, and so forth), assess-
ment for surgical correction, and methods of follow-up.
Treatment outcomes must be evaluated by the same 3D
principles using identical 3D methods.
In the last decade, new concepts were developed along
the approach introduced by SRS, and attempts have been
made at creation of a new scoliosis classification based
on 3D visualizations [15–19]. A recent article by members
of the SRS Committee for 3D Classification of Scoliosis
summarized the work accomplished. This article states that
based on the preliminary work done in the past 4 years,
a valid and clinically useful new 3D classification is within
reach . The presented concept proposes 3D analysis of
the scoliotic spine by a schematic visualization of simpli-
fied horizontal plane projections of spinal curves in auxil-
iary planes determined according to the original concept
of PMC described by Stagnara and Queneau .
It is a readily acceptable endeavor that a new 3D classifi-
cation of scoliosis might involve certain compromises and
simplifications, even by the introduction of auxiliary planes
or abstraction of spinal curves into such planes. To be valid
and clinically useful, however, the proposed solution must
be clear and easily translated into clinical practicewhile pre-
mation as possible. Unfortunately, the concept proposed by
the SRS 3D Committee is—although based on fundamental
and pioneering work on 3D visualization of the spine—not
entirely convincing in this respect. Comprehending the ab-
straction of the spine into three PMCs is not an easy task in
routine clinical practice. This is additionally complicated
top view, using a so-called da Vinci representation.
The result is a highly abstract visual presentation that is
hard to reconnect with the 3D reality or the 3D reconstruc-
tion—not speaking of 2D diagnostic images—of the spinal
column involved. Furthermore, data are demonstrated in
a complicated multi-scaled diagram that is difficult to read
and interpret interms of conventional clinical routine.Based
on the few samples of the couple of curve types presented, it
3D mechanism in scoliosis or to see its clinical applicability
Fig. 3. Interrater reliability of curve measurements within patient groups.
Barplots for (Top) coronal curve, (Middle) thoracic kyphosis, and (Bottom)
lumbar lordosis measurements. Interrater ICC values (color bars) with up-
per/lower limits of 95% confidence interval (error bars) for single mea-
sures are shown. Two-way random model with absolute agreement was
used; for all values, p!.001. Group 1: Cobb 0 to 10?; Group 2: Cobb 10
to 25?; Group 3: Cobb 25 to 50?; Group 4: Cobb 50 to 75?; Group 5: Cobb
O75?. ICC, intraclass correlation coefficient; 2D, two-dimensional; 3D,
three-dimensional; VV, vertebra vector.
Fig. 4. L5 wedge and difference between manual 2D and VV 3D values
for L1–L5 lordosis means (color bars) and standard deviations (error bars).
Means shown are averages of examiners average values for measurements.
Group 1: Cobb 0 to 10?; Group 2: Cobb 10 to 25?; Group 3: Cobb 25 to
50?; Group 4: Cobb 50 to 75?; Group 5: CobbO75?. p Values for each dif-
ferences are not significant.
8 S. Somoske€ oy et al. / The Spine Journal - (2012) -
for the evaluation of scoliotic curves. The authors also note Download full-text
that further validation of the proposed method is needed;
and reproducibility of numerical results of their method or
the cross-validated classification based on it. Basically, no
exact numerical results or quasi-3D measurement values
have been presented that would quantify the few visual sam-
ples shown. Overall, the proposal of this new method—and
a new 3D classification based on it—is not completely con-
by the SRS Working Group in 1993.
On the other hand, the concept of vertebravectors and 3D
visualization and characterization of spinal deformities
based on vertebra vectors could plausibly satisfy that ap-
proach. Although the concept of vertebra vectors is also
a simplified visual presentation of the 3D reconstructed
spine, it is placed inside a standard cartesian coordinate sys-
tem, not requiring introduction of any auxiliary planes. As
a result, it represents a 3D system for true 3D measurements,
based on a patient-specific individual scale. The substitution
of vertebrae by vertebra vectors preserves important infor-
mation of vertebral size, 3D position and orientation, as well
as rotation. While providing clear and comprehensible vi-
sual views in any of the standard plane views—especially
in horizontal plane top view—it does not compromise
purely mathematical calculations for3D evaluation of spinal
The concept of vertebra vectors was introduced recently
as a new, alternative method for 3D visualization and char-
acterization of the spinal column. This study reported on the
successful validation of the concept using normal patients
and patients with a wide range of scoliotic deformities, doc-
umenting highly accurate measurement values for coronal
and sagittal curves based on vertebra vectors, with excellent
intraobserver reliability and a consistently higher interrater
reproducibility compared with traditional 2D measurement
methods. As a result, the concept of vertebra vectors may
serve as a starting point to a valid alternative for a new
3D classification of scoliosis.
 Vrtovec T, Pernus F, Likar B. A review of methods for quantitative
evaluation of spinal curvature. Eur Spine J 2009;18:593–607.
 King HA, Moe JH, Bradford DS. The selection of fusion levels in
thoracic idiopathic scoliosis. J Bone Joint Surg Am 1983;65:
 Lenke LG, Betz RR, Harms J, et al. Adolescent idiopathic scoliosis:
a new classification to determine extent of spinal arthrodesis. J Bone
Joint Surg Am 2001;83:1169–81.
 Aubin CE, Dansereau J, Parent F, et al. Morphometric evaluations of
personalised 3D reconstructions and geometric models of the human
spine. Med Biol Eng Comput 1997;35:611–8.
 Mitulescu A, Skalli W, Mitton D, et al. Three-dimensional surface
rendering reconstruction of scoliotic vertebrae using a non stereo-
corresponding points technique. Eur Spine J 2002;11:344–52.
 Le Bras A, Laporte S, Mitton D, et al. Three-dimensional (3D) de-
tailed reconstruction of human vertebrae from low-dose digital ster-
eoradiography. Eur J Orthop Surg Traumatol 2003;13:57–62.
 Dubousset J, Charpak G, Dorion I, et al. A new 2D and 3D imaging
approach to musculoskeletal physiology and pathology with low-dose
radiation and the standing position: the EOS system. Bull Acad Natl
 Kalifa G, Charpak G, Maccia C, et al. Evaluation of a new low-dose
digital X-ray device: first dosimertric and clinical results in children.
Pediatr Radiol 1998;28:557–61.
 Desch^ enes S, Charron G, Beaudoin G, et al. Diagnostic imaging of
spinal deformities: reducing patients radiation dose with a new
slot-scanning x-ray imager. Spine 2010;35:989–94.
adolescent idiopathic scoliosis measurements. Spine 2005;30:444–54.
 Gstoettner M, Sekyra K, Walochnik N, et al. Inter- and intraobserver
reliability assessment of the Cobb angle: manual versus digital mea-
surement tools. Eur Spine J 2007;16:1587–92.
 Lenke G. Lenke classification system of adolescent idiopathic scoli-
osis: treatment recommendations. Instr Course Lect 2005;54:537–42.
 Clements DH, Marks M, Newton PO, et al. Did the Lenke classifica-
tion change scoliosis treatment? Spine 2011;36:1142–5.
 Stokes I. Three-dimensional terminology of spinal deformity: a report
presented to the Scoliosis Research Society by the Scoliosis Research
Society Working Group on 3D terminology of spinal deformity.
 Poncet P, Dansereau J, Labelle H. Geometric torsion in idiopathic
scoliosis: 3-D analysis and proposal for a new classification. Spine
 Duong L, Cheriet F, Labelle H. Three-dimensional classification of
spinal deformities using fuzzy clustering. Spine 2006;31:923–30.
 Duong L, Mac-Thiong JM, Cheriet F, et al. Three-dimensional sub-
classification of Lenke type 1 scoliotic curves. J Spinal Disord Tech
 Sangole AP, Aubin C-E, Labelle H, et al. Three-dimensional classifi-
cation of thoracic scoliotic curves. Spine 2009;34:91–9.
 Stokes IA, Sangole AP, Aubin C-E. Classification of scoliosis defor-
mity three-dimensional spinal shape by cluster analysis. Spine 2009;
 Labelle H, Aubin C-E, Jackson R, et al. Seeing the spine in 3D: how
will it change what we do? J Pediatr Orthop 2011;31:S37–45.
 Ill? es T, Tunyogi-Csap? o M, Somoske€ oy S. Breakthrough in three-
dimensional scoliosis diagnosis: significance of horizontal plane view
and vertebra vectors. Eur Spine J 2011;20:135–43.
 Steffen JS, Obeid I, Aurouer N, et al. 3D postural balance with regard
to gravity line: an evaluation in the transversal plane on 93 patients
and 23 asymptomatic volunteers. Eur Spine J 2010;19:760–7.
 Humbert L, De Guise JA, Aubert B, et al. 3D reconstruction of the
spine from biplanar X-rays using parametric models based on trans-
versal and longitudinal inferences. Med Eng Phys 2009;31:681–7.
 Dimar JR, Carreon LY, Labelle H, et al. Intra- and inter-observer re-
liability of determining radiographic sagittal parameters of the spine
and pelvis using a manual and a computer-assisted methods. Eur
Spine J 2008;17:1587–92.
 Polly DW, Kilkelly FX, McHale KA, et al. Measurement of lumbar
lordosis: evaluation of intraobserver, interobserver, and technique
variability. Spine 1996;21:1530–5.
 Rousseau M-A, Laporte S, Chavary-Bernier E, et al. Reproducibility
of measuring the shape and three-dimensional position of cervical
vertebrae in upright position using the EOS stereoradiography sys-
tem. Spine 2007;32:2569–72.
 Stagnara P, Queneau P. Scolioses ? evolutives en p? eriode de croissance:
aspects cliniques, radiologiques, propositions th? erapeutiques. Rev
Chir Orthop Reparatrice Appar Mot 1953;39:378–452.
9S. Somoske€ oy et al. / The Spine Journal - (2012) -