See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/42345855

An optimised method for quantifying glenoid orientation

Article  in  International Journal of Shoulder Surgery · April 2008

DOI: 10.4103/0973-6042.41407 · Source: PubMed

CITATIONS READS
9 457

5 authors, including:

Hippolite O Amadi Sughran Banerjee

Imperial College London Chiron Daycare Clinic
38 PUBLICATIONS   134 CITATIONS    14 PUBLICATIONS   116 CITATIONS   

SEE PROFILE SEE PROFILE

Ulrich Hansen Anthony M J Bull

Imperial College London Imperial College London
51 PUBLICATIONS   1,574 CITATIONS    328 PUBLICATIONS   7,029 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Neonatal hyperbilirubinaemia. View project

The fabella, a forgotten sesamoid View project

All content following this page was uploaded by Hippolite O Amadi on 29 May 2018.

The user has requested enhancement of the downloaded file.

 [Downloaded free from http://www.www.internationalshoulderjournal.org on Sunday, July, 13 2008]

Original Article
An optimised method for quantifying
glenoid orientation
Hippolite O. Amadi, Sughran Banerjee, Ulrich N. Hansen, Andrew L. Wallace1,
Anthony M. J. Bull

ABSTRACT Department of Bioengineering, Imperial

College London, South Kensington
A robust quantification method is essential for inter-subject glenoid comparison and planning of Campus, London SW7 2AZ, UK.
1
Shoulder Unit, Hospital of St John and
total shoulder arthroplasty. This study compared various scapular and glenoid axes with each
St Elizabeth, 60 Grove End Road, London
other in order to optimally define the most appropriate method of quantifying glenoid version and NW8 9NH, UK.
inclination.
Six glenoid and eight scapular axes were defined and quantified from identifiable landmarks of
Correspondence:
twenty-one scapular image scans. Pathology independency and insensitivity of each axis to inter- Dr Anthony MJ Bull
subject morphological variation within its region was tested. Glenoid version and inclination were Department of Bioengineering,
Imperial College London,
calculated using the best axes from the two regions.
South Kensington Campus,
The best glenoid axis was the normal to a least-square plane fit on the glenoid rim, directed London SW7 2AZ,
approximately medio-laterally. The best scapular axis was the normal to a plane formed by the United Kingdom.
E-mail: [email protected]
spine root and lateral border ridge. Glenoid inclination was 15.7° ± 5.1° superiorly and version
was 4.9° ± 6.1°, retroversion.
The choice of axes in the present technique makes it insensitive to pathology and scapular
morphological variabilities. Its application would effectively improve inter-subject glenoid version
comparison, surgical planning and design of prostheses for shoulder arthroplasty.

Key words: Version, inclination, morphology

INTRODUCTION approximately to the mid-glenoid level [Figure 1]. Although an

improvement on conventional X-ray methods[10] there remain
Effective surgical planning for total shoulder arthroplasty limitations to this technique in that the results are scanning-
requires a clear understanding of a patient’s glenoid version orientation dependent;[3,9,12] it is essential that the glenoid surface
and inclination,[1-6] Quantification of these parameters even is perpendicular to the plane of the CT slice. An improvement
in the presence of osseous pathology requires a robust and is to use ultrasound to define the perpendicular to the glenoid
reproducible technique.[7] Several methods have been proposed face.[12] It is known that the glenoid face is twisted in a superior-
for in vivo quantification of glenoid version; from the use of inferior direction[3,4] and therefore the use of two points from
conventional roentgenograms to axial-tomographic scans.[4,8,9] a subjective mid-glenoid slice will be susceptible to inherent
Computed tomographic (CT) methods are more reproducible errors. Others have used methods with either surface scanning[4]
and reliable compared to conventional X-ray methods.[10,11] or direct physical measurements[6] of ex vivo scapulae. These
methods suffer from scanning orientation dependency that is
Friedman et al.,[8] used a method that requires three landmarks set by eye,[4] or use of only two points to define an angle.[6]
to define glenoid version. They used CT scans in the axial
plane from the acromion to the inferior border of the glenoid. In another study, Couteau et al. [9] carried out a 3-D
Glenoid version was measured on the slice corresponding morphological and mechanical analysis of twelve shoulders

Please cite this article as: Amadi HO, Banerjee S, Hansen UN, Wallace AL, Bull AMJ. An optimised method for quantifying glenoid orientation. Int J Shoulder Surg 2008,2:2:0-0.

25 IJSS - April-June 2008 / Volume 2 / Issue 2 ♦

 [Downloaded free from http://www.www.internationalshoulderjournal.org on Sunday, July, 13 2008]

Amadi et al.: Optimised glenoid quantification

and four cryosectional image datasets. Sixteen of the specimens

were left shoulders, mean age was 60 years, range (57 years to 79
years). Nine of the image scans were of 1.00 mm slice thickness,
six (1.50 mm), four (1.40 mm) and two (1.25 mm).

Features or regions of interest within the field of view of any

standard shoulder or chest scan were defined. This includes
regions within the scapular distal half and the supraglenoid
tubercle. AMIRA image processing software (Mercury
Computer Systems Inc, Chelmsford, MA, USA) was applied to
segment and extract the three-dimensional locations describing
each feature of interest.

Least-square basic geometric shapes such as an ellipse, plane,

Figure 1: Mid glenoid section illustrating version angle due to Friedman line or triangle were numerically fitted on a given set of
et al. 1992 points to quantify axes on each specimen. These include
those normally applied by classical techniques for glenoid
using CT scans. In their method, the points defining the glenoid quantification and some novel ones. The specific axes that were
articular surface were extracted and their centroid calculated. defined are described below:
A least-square (LS) plane was mathematically fitted on the I. Glenoid rim normal (GNrim): This is the normal unit
extracted points and a normal unit vector to this quantified. vector to the best-fit plane over the rim of the glenoid. The
This represented the glenoid axis. A mid-transverse section outline of the glenoid rim was segmented, reconstructed
of the glenoid was defined as the axial slice corresponding and applied to mathematically quantify the least square
to the location of the centroid. The central axis of inertia of plane-fit over the points and the normal unit vector to it
this slice was quantified to represent the scapular axis. The [Figure 2].
version angle was finally calculated as the angle between the II. Glenoid fossa normal (GNfos): The normal unit vector
two representative axes. to the best-fit plane over the glenoid fossa.[9] The entire
glenoid fossa was segmented, reconstructed and applied to
Fundamentally, glenoid quantification can be seen as the quantify the plane and its normal [Figure 2].
measurement of the glenoid plane orientation relative to the III. Glenoid equatorial line (GEL): A line joining the anterior
scapular plane. All the earlier techniques achieved this by and posterior margins of the mid-glenoid slice. [4,8,10]
applying two axes, one each to represent the planes. Most of This is the axial slice midway along the glenoid height
these techniques rely on three or fewer landmark points that [Figure 3].
are susceptible to failures in the presence of pathologies. Again, IV. Coronal mid-glenoid superior axis (CMGS): A line joining
it is known that inter-subject variability in the morphology the inferior and superior margins of the mid-glenoid slice
of the scapula exists which none of these techniques from the coronal frames of an image scan [Figure 3]. This
addressed.[13,14] Therefore, comparison of glenoid quantification is the coronal slice midway along the glenoid width.
between individuals using these techniques might not be V. Bokor glenoid equatorial line (BGEL): This is a GEL based
reliably accurate. A more reliable technique could be developed on the proposals of Bokor et al.,[12] that scan orientation
based on axes that address the known limitations, having should be such that the glenoid surface is perpendicular to
minimal inter-subject variability as well as being pathology- the plane of the CT axial cut [Figure 3]. This was achieved
independent. using image processing software.
VI. Glenoid superior axis (GSA): A line directed superiorly
The aim of this work was to: from the most inferior aspect of the glenoid to the biceps
1. Compute the axes of the glenoid and scapula as defined tendon insertion.[5,6,15]
in the literature as well as other axes defined here from
clearly identifiable landmarks, For the scapula, the axes were:
2. To use weighting criteria to compute the best axes that I. Lateral border line (LBL): The best-fit inferior-superior
are least susceptible to morphometric variability to define line along the ridge of the scapular lateral border
glenoid version and inclination. [Figure 4].
II. Spine root line (SRL): The best-fit long-axis along the
MATERIALS AND METHODS root of the scapular spine [Figure 4].
III. Scapular normal (SN): The cross-product (unit vector)
Three-dimensional image datasets from standard shoulder scans between LBL and SRL [Figure 4]. This is directed
were assessed for obvious osseous pathology and twenty-one anteriorly.
of them selected. This comprised seventeen CT image scans IV. Scapular transverse axis (STA): A line drawn from the

♦ IJSS - April-June 2008 / Volume 2 / Issue 2 26

 [Downloaded free from http://www.www.internationalshoulderjournal.org on Sunday, July, 13 2008]

Amadi et al.: Optimised glenoid quantification

Figure 2: Normal unit vectors to the glenoid

Figure 4: Normal unit vector to the scapular body and its parent
vectors

axis was defined as the sum of its SDs from the 21 specimens.
All the insensitivity indices were normalized relative to the
smallest index which assumed a weighting value of 1. ‘Relative
Insensitivity’ of the rest of the axes was thus quantified.
These were also done for the scapular axes. In addition to
high insensitivity, the final criterion for the selection of the
best axes was based on pathology-independency. An axis of
Figure 3: (a) Re-slicing to conform to Bokor et al’s proposal (b) Glenoid which quantification was based on two or three points only
equatorial line and (c) Coronal mid-glenoid axis
had a risk of pathological failure if any of the quantification
landmarks was associated with any regular osseous pathology.
mid point of the glenoid fossa to the medial edge on the Such an axis was assigned a weighting of 1, otherwise this
mid glenoid transverse slice[8] [Figure 1]. was 0. Optimal glenoid version was defined as a measure
V. V. Bokor scapular transverse axis (BSTA) as proposed by of the angle between the best glenoid axis and that of the
Bokor et al.[12] scapula on the approximate transverse plane. Optimal glenoid
VI. Second Moment of Area transverse Axis (SMATA): The inclination was defined as a measure of the angle between the
medio-laterally directed principal axis of the second best glenoid and scapular axes on the approximate coronal
moment of area quantified on the closest axial slice to plane. Four different classical techniques were also applied to
the centroid of the glenoid fossa. quantify version of each specimen. These were: (I) Friedman
VII. Wong scapular transverse axis (WSTA): This is a line et al.,[8] angle between GEL and STA; (II) Bokor et al.,[12] angle
joining the spinoglenoid notch and the spine/medial between BGEL and BSTA; (III) Couteau et al.,[9] (modified)
border intersection.[5] angle between GNfos and SMATA; (IV) Churchill et al.,[6]
VIII. Churchill scapular transverse axis (CSTA): This is a line angle between BGEL and CSTA. Glenoid inclination was
joining the centre of the glenoid fossa and the spine/ quantified using two other methods: (I) Wong et al.,[5] method
medial border intersection.[6] as the included angle between WSTA and GSA; (II) Churchill
et al.,[6] method as the angle between CSTA and GSA. The
The corporate morphology of the glenoid or scapula was correlation coefficients between these and the optimal methods
characterized by these axes that were defined from landmarks. were also calculated.
It is therefore essential to identify a glenoid axis that integrates
the variations in the remaining axes in its make-up. Such an RESULTS
axis would therefore be relatively insensitive to changes in
glenoid morphology represented by inter-subject variations in The most insensitive axis of the glenoid is the normal to a LS
the remaining axes. For the scapular body also, the best axis plane fit on the glenoid rim (GNrim) while that of the scapula
capable of reflecting this quality was required. is the normal to the plane formed by LBL and SRL (SN). These
have Relative Insensitivity of 1.00 respectively [Tables 1 and 2].
The angles between all the glenoid axes were calculated in Quantification of these involved multitudes of points over their
all the specimens. The means and standard deviations (SD) landmarks. Optimal glenoid version is therefore a measure of
for these were quantified. A relatively insensitive axis would the angle between GNrim and SN. This produced a mean value
result in a smaller sum total of its SDs from the specimens of 4.9 ± 6.1°, retroversion; range: -16.4° to 10.7°. Mean glenoid
compared to the rest of the axes. The insensitivity index of an version using Friedman et al.[8] technique was 12.2° ± 8.4°,

27 IJSS - April-June 2008 / Volume 2 / Issue 2 ♦

 [Downloaded free from http://www.www.internationalshoulderjournal.org on Sunday, July, 13 2008]

Amadi et al.: Optimised glenoid quantification

retroversion; range: -30.6° to 0°; having correlation coefficient et al.[8] and Monk et al.[4] and hence avoids the subjective
of 0.08 with the optimal method. Bokor et al.,[12] technique on opinion of the sonographer. The rim of the glenoid has been
the same specimens produced mean glenoid version of 3.5° ± reported to be superoinferiorly twisted and might have the
4.8°, anteversion; range: -4.5° to 14.5°; and correlation coefficient presence of osseous pathology.[3,7,9] The fitting of LS plane
of 0.26 with the optimal method. By Cauteau et al.,[9] parallel over the glenoid face using over two thousand points across
technique, this was 15.8° ± 38.2°, anteversion and correlation the glenoid rim constitutes a better approximation of glenoid
coefficient of 0.12. Churchill et al.,[6] method produced 3.3° definition irrespective of the presence of the aforementioned
± 4.6°, anteversion; range: -4.6° to 13.1°; and correlation complications. The SN axis integrates most of scapular
coefficient of 0.23. The CSTA and WSTA with equal Relative morphology represented in over 5000 points from its parent
Insensitivity of 1.02 are the most insensitive scapular axes on axes (SRL and LBL). This is therefore a better representation
the approximate coronal plane. These were followed closely by of the scapular body compared to only two points applied by
SRL (relative insensitivity, 1.03). By pathology-independency the classical methods. None of the earlier techniques produced
criterion, the SRL was quantified with numerous points as a good correlation with the present technique because of its
against the two-point and pathology-dependent CSTA and unique approach. This used an ‘anterior-posterior’ axis for the
WSTA. This was therefore chosen as the best. This combines scapula instead of ‘medio-lateral’ axis applied by others.
with the glenoid’s GNrim to produce an ‘optimal’ mean glenoid
inclination of 15.7° ± 5.1°, superiorly; range: -7° to 27.4°. Wong Glenoid inclination has not been as extensively discussed in the
et al.,[5] method quantified a mean inclination of 0.9° ± 4.3°, literature as the version. This might suggest that the parameter
superiorly; range: -7.1° to 11.2°. Churchill et al.,[6] method is not seen to be so important during shoulder arthroplasty.
produced 5.2° ± 3°, superiorly; range: 0.8° to 11.5°. However, it is known that a more upward-facing glenoid
increases the risk of superior humeral head migration, possibly
DISCUSSION associated with the genesis of rotator cuff disease.[5] Similar
to the classical method of version calculation, inclination is
The classical methods of glenoid version quantification have based on defining a line joining two points only on the glenoid
been associated with various limitations such as scanning rim.[5,6] The second line has been differently defined in the
orientation dependency. [3,4,9,12] More recent studies have literature. Churchill et al.,[6] line joined the spine-medial
proposed other methods that addressed the orientation border intersection to the glenoid fossa centre while Wong
factor. However, these are also flawed for being sonographer- et al.,[5] joined this to the spinoglenoid notch. The present
dependent in ensuring preferred scanning orientation. study however, has demonstrated glenoid inclination based on
The technique proposed in the present study was based a multipoint approach, using GNrim and SRL.
on thousands of vectors to form the SRL, SN and GNrim
axes. GNrim integrates the corporate morphology of the Quantifications based on the present proposals are easily
glenoid rim rather than two points only compared to other realized using any standard shoulder or chest scans and do
methods.[6,8,12] This would therefore remain stable irrespective not require any special radiological scan of the patient. The
of the scanning orientation unlike the techniques of Friedman derivation and application of subject-invariant axes in this

Table 1: Relative insensitivity and pathology dependency in glenoid axes

Axes RI PD Direction No of points involved
GNrim (Glenoid rim normal) 1.00 0 medio-lateral Thousands
GNfos (Glenoid fossa normal) 1.67 0 medio-lateral Thousands
GSA (Novotny’s line) 1.23 1 infero-superior 2
BGEL (Bokor’s line) 1.33 1 antero-posterior 2
GEL (Friedman’s line) 1.50 1 antero-posterior 2
CMGS (Mid-glenoid i-s line) 2.44 1 infero-superior 2
RI - Relative insensitivity, PD - Pathology dependency

Table 2: Relative insensitivity and pathology dependency in scapular axes

Axes RI PD Direction No of points involved
SN (SRL-LBL plane normal) 1.00 0 antero-posterior Thousands
SRL (Spine Root line) 1.03 0 medio-lateral Thousands
WSTA (Wong’s line) 1.02 0 medio-lateral 2
LBL (Lateral Border Line) 2.00 0 infero-superior Thousands
CSTA (Churchill’s line) 1.02 1 medio-lateral 2
BSTA (Bokor’s line) 1.07 1 medio-lateral 3
SMATA (2nd Moment Area) 2.52 0 medio-lateral Thousands
STA (Friedman’s line) 1.89 1 medio-lateral 3
RI - Relative insensitivity, PD - Pathology dependency

♦ IJSS - April-June 2008 / Volume 2 / Issue 2 28

 [Downloaded free from http://www.www.internationalshoulderjournal.org on Sunday, July, 13 2008]

Amadi et al.: Optimised glenoid quantification

study would allow a more accurate inter-subject comparison 6. Churchill RS, Brems JJ, Kotschi H. Glenoid size, inclination, and
of glenoid quantification. This could allow better design of version: An anatomic study. J Shoulder Elbow Surg 2001;10:327-
32.
prostheses and ensure a more effective surfacing of the glenoid 7. Sugaya H, Moriishi J, Dohi M, Kon Y, Tsuchiya A. Glenoid rim
during total shoulder arthroplasty. The present technique’s morphology in recurrent anterior glenohumeral instability. J
sensitivity to numerically describing version and inclination Bone Joint Surg Am 2003;85:878-84.
and its insensitivity to scanning orientation suggest that this 8. Friedman RJ, Hawthorne KB, Genez BM. use of computerized
tomography in the measurement of glenoid version. J Bone Joint
has the potential to be a clinical tool in assessing glenohumeral Surg Am 1992;74:1032-7.
function. As a numerical technique, this can be automated 9. Couteau B, Mansat P, Darmana R, Mansat M, Egan J. Morpho-
and considerable time saved for the quantification of these logical and mechanical analysis of the glenoid by 3D geometric
parameters. Further studies will have to be conducted to reconstruction using computed tomography. Clin Biomech
2000;15:S8-12.
relate these parameters of version and inclination to clinical 10. Nyffeler RW, Jost B, Pfirrmann CW, Gerber C. Measurement of
outcome. glenoid version: Conventional radiographs versus computed
tomography scans. J Shoulder Elbow Surg 2003;12:493-6.
11. Kwon YW, Powell KA, Yum JK, Brems JJ, Iannotti JP. Use of three-
REFERENCES dimensional computed tomography for the analysis of glenoid
anatomy. J Shoulder Elbow Surg 2005;14:85-90.
1. De Wilde L, Berghs B, VandeVyver F, Schepens A, Verdonk RC. 12. Bokor DJ, O’Sullivan MD, Hazan GJ. Variability of measurement
Glenohumeral relationship in the transverse plane of the body. of glenoid version on computed tomography scan. J Shoulder
J Shoulder Elbow Surg 2003;12:260-7. Elbow Surg 1999;8:595-8.
2. Oosterom R, Rozing P, Bersee H. Effect of glenoid component 13. Ste hle J, Moore SM, Alaseirlis DA, Debski RE, McMahon PJ.
inclination on its fixation and humeral head subluxation in total Acromial morphology: Effects of suboptimal radiographs. J
shoulder arthroplasty. Clin Biomech 2004;19:1000-8. Shoulder Elbow Surg 2007;16:135-42.
3. Schlemmer B, Dosch J, Gicquel P, Boutemy P, Wolfram R, Kempf 14. Gallino M, Santamaria E, Doro T. Anthropometry of the scapula:
JF, et al. Computed tomographic analysis of humeral retrotorsion Clinical and surgical considerations. J Shoulder Elbow Surg
and glenoid retroversion. Rev Chir Orthop Reparatrice Appar 1998;7:284-91.
Mot 2002;88:553-60. 15. Novotny JE, Beynnon BD, Nichols CE. Modeling the stability
4. Monk A, Berry E, Limb D, Soames R. Laser morphometric of the glenohumeral joint during external rotation. J Biomech
analysis of the glenoid fossa of the scapula. Clin Anatomy 2000;33:345-54.
2001;14:320-3.
5. Wong AS, Gallo L, Kuhn JE, Carpenter JE, Hughes RE. The effect
of glenoid inclination on superior humeral head migration. J Source of Support: Nil, Conflict of Interest: None declared.
Shoulder Elbow Surg 2003;12:360-4.

29 IJSS - April-June 2008 / Volume 2 / Issue 2 ♦

View publication stats