http://pih.sagepub.com/Medicine

Engineers, Part H: Journal of Engineering in Proceedings of the Institution of Mechanical

http://pih.sagepub.com/content/225/1/48The online version of this article can be found at:

 DOI: 10.1243/09544119JEIM810

2011 225: 48Proceedings of the Institution of Mechanical Engineers, Part H: Journal of Engineering in MedicineP J Watson, P O'Higgins, M J Fagan and C A Dobson

Validation of a Morphometric Reconstruction Technique Applied to a Juvenile Pelvis  

Published by:

http://www.sagepublications.com

On behalf of: 

  Institution of Mechanical Engineers

can be found at:Proceedings of the Institution of Mechanical Engineers, Part H: Journal of Engineering in MedicineAdditional services and information for

    

  http://pih.sagepub.com/cgi/alertsEmail Alerts:

 

http://pih.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

http://pih.sagepub.com/content/225/1/48.refs.htmlCitations:  

by guest on September 15, 2011pih.sagepub.comDownloaded from

Validation of a morphometric reconstruction techniqueapplied to a juvenile pelvisP J Watson1*, P O’Higgins2, M J Fagan1, and C A Dobson1

1Department of Engineering, University of Hull, UK2Department of Anatomy, Hull York Medical School, University of York, UK

The manuscript was received on 1 February 2010 and was accepted after revision for publication on 17 June 2010.

DOI: 10.1243/09544119JEIM810

Abstract: Three-dimensional reconstructions of bone geometry from microCT (computedtomography) data are frequently used in biomechanical and finite element analyses. Digiti-zation of bone models is usually a simple process for specimens with a complete geometry, butin instances of damage or disarticulation it can be very challenging. Subsequent to digitization,further imaging techniques are often required to estimate the geometry of missing bone orconnecting cartilage. This paper presents an innovative approach to the reconstruction of in-complete scan data, to reproduce proper anatomical arrangements of bones, including absentconnecting cartilaginous elements. Utilizing geometric morphometric tools, the reconstructiontechnique is validated through comparison of a reconstructed 9 year old pelvis, to the originalCT data. A principal component analysis and an overlay of the two pelves provide a measure ofthe accuracy of the reconstructed model. Future work aims to investigate the biomechanicaleffects of any minor positional error on the bone’s predicted structural properties through theuse of finite element analysis.

Keywords: Geometric morphometric methods, morphometric reconstruction, juvenile pelvis

1 INTRODUCTION

The influence of mechanical strain upon bone growth

and remodelling has been widely studied [1–5], and it

has been established that bone optimizes its struc-

ture to withstand the most commonly experienced

loading regimes in accordance with the Wolff’s law

[6, 7]. The lower extremities are frequently analysed

to examine the relationship between mechanical

loading and remodelling of weight-bearing bones

[8, 9], particularly in areas such as the femoral head

[10, 11]. However, strain-related bone development

is also evident within the human pelvis, where

distinct regions of highly dense trabecular and thick

cortical bone are found to match the stress induced

by load transfer during bipedal locomotion [12, 13].

The current work aims to investigate the influence

of loading regimens brought about by age-related

movements (i.e. walking and crawling) upon the

growth and remodelling of the internal trabecular

network of the pelvis. This was prompted in part by

the discovery of trabecular organization in prenatal

pelvic specimens [14, 15], which is similar to that

found in adulthood, contradicting the previously

held view that trabecular organization is dependent

on external forces arising from locomotion during

growth [1, 16]. To investigate this finding further,

biomechanical and finite element (FE) analyses are

to be conducted on a range of juvenile ages, to ascer-

tain the strain distribution in the pelvis throughout

locomotor development.

Until now, biomechanical analyses examining the

influence of mechanical loading on human pelvic bone

growth have been predominantly performed on adult

specimens [17–19], where trabecular densities are

clearly evident and loading regimes can be determined

relatively easily. In contrast, juvenile bone growth re-

search is often hampered by the lack of available speci-

mens and difficulty in acquiring ethical approval for

biomechanical testing. FE analysis offers a method of

examining rare juvenile material in an ethical and non-

destructive way.

*Corresponding author: Department of Engineering, University

of Hull, Cottingham Road, Hull HU6 7RX, UK.

email: p.watson@eng.hull.ac.uk

48

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

by guest on September 15, 2011pih.sagepub.comDownloaded from

Ossification of the juvenile skeleton is incomplete

until late adolescence, and therefore fusion of the

three pelvic bones (ilium, ischium, and pubis) (see

Fig. 1) into the single innominate bone does not

occur in humans until this time [20]. Prior to fusion,

cartilage provides connectivity between bones, to

such a degree that the acetabulum is not fully ossified

until 16–18 years of age [21]. Therefore, when con-

sidering juvenile bones that have incomplete fusion,

it is essential for any FE analysis that the relative

positions of the bones and the geometry of the

cartilaginous material are modelled accurately to

represent the situation in vivo.

Computerized biomechanical analyses of bone

structures require the digitization of magnetic reson-

ance imaging (MRI) or computed tomography (CT)

data to produce accurate three-dimensional volume

models [22, 23]. There are many competent software

packages that aid in the segmentation and construc-

tion of these models, enabling the internal and

external architecture to be replicated faithfully from

the scan data.

Geometric morphometric methods (GMMs) [24–26]

facilitate comparison of configurations of landmarks

between forms and, since they preserve geometry at all

stages of analysis, visualization of results in terms of

geometry is straightforward. Variations among land-

mark configurations can be represented through the

use of thin-plate splines [27] to warp associated

volumes and surfaces [28]. Here these methods were

used digitally to warp a developed hemi-pelvis to a

known juvenile geometry, hence providing a template

to enable the repositioning of disarticulated bones.

This paper presents a new approach to the recon-

struction of incomplete scan data to produce FE

models that replicate proper anatomical arrangements

of bones, including absent connecting cartilaginous

elements. It is presented using statistical and warping

tools from geometric morphometrics to reconstruct

a model of a juvenile hemi-pelvis from scan data

containing each of its three constituent bones.

The technique was validated through the use of a

complete clinical CT scan of a 9 year old (i.e. soft

tissue included). The three bones of the left hemi-

pelvis were individually segmented from the scan,

and then repositioned within a warped hemi-pelvic

template of a similar form. The relative orientation

of the positioned bones was then compared to their

positions in the original complete CT scan.

2 METHOD

The general geometric morphometric reconstruction

will first be described, followed by the details of the

specific data used for the validation of the technique.

Fig. 1 Anatomy of the hemi-pelvis and positioning of the morphometric landmarks used for thewarping and reconstruction in this paper

Validation of a morphometric reconstruction technique 49

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

by guest on September 15, 2011pih.sagepub.comDownloaded from

2.1 Geometric morphometric reconstructionmethod

The procedure for undertaking the morphometric

reconstruction can be described in four stages, as

shown in Fig. 2:

Stage 1 . Morphometric datasets comprising 27 land-

marks (identified initially on complete museum

specimens from the Natural History Museum) were

gathered for juvenile pelves (see Fig. 1 and Table 1)

spanning infancy to adulthood (Stage 1.1 of Fig. 2).

A CT scan of a fully fused adult pelvis was then

segmented and the same anatomical landmarks

identified manually (Stage 1.2 of Fig. 2). Finally, the

disarticulated CT scan data of a juvenile hemi-pelvis

requiring reconstruction were segmented (Stage 1.3

of Fig. 2).Stage 2 . Based on the iliac depth of the disarticulated

juvenile hemi-pelvis (measured as the distance

between the ASIS and PSIS [29]; see points 2 and 9

on Fig. 1), the mean landmark configuration for a

comparably sized individual was estimated using a

multivariate regression of Procrustes registered land-

mark configurations. This was performed on the

original morphometric dataset within the GMM

software Morphologika [30]. These landmarks were

then used to create a template of appropriate form

(size and shape), using warping tools (a triplet of thin-

plate splines – the ‘Bookstein’ function) within

AMIRAH Image Segmentation Software (TGS Inc,

USA) to warp the adult pelvis to fit the estimated size

and shape of the landmark configuration. The result

was an anatomical estimate of a continuous hemi-

pelvic structure of the required size to be used as the

template for subsequent stages of the reconstruction.Stage 3 . The individual disarticulated pelvic bones

were then overlaid upon the warped template to

gauge their position and relative orientation.Stage 4 . Finally, the volume and geometry of the

connective cartilage were estimated by examining

the spacing between the repositioned bones once

placed within the template, to produce a complete

pelvic structure.

2.2 Validation of the method using a complete 9year old pelvic CT

For Stage 1.2, CT data from a fully fused 9 year old

pelvis were segmented and digitized into a volu-

metric hemi-pelvic model, with a slice thickness of

0.48 mm60.48 mm60.25 mm (see Fig. 3(a)). The co-

ordinates of the same 27 pelvic landmarks were also

taken from this specimen and used to control

subsequent warping.

Clinical CT scan data from a 9 year old subject

including the pelvis was obtained from the research

collection of Medizinische Universitat Innsbruck,

courtesy of Dr Wolfgang Recheis. The pelvis was

segmented and a volumetric model of the left hemi-

pelvis reconstructed using AMIRAH with a slice

thickness of 4.98 mm64.98 mm62.5 mm. This re-

construction identified and defined the separate

constituents of the hemi-pelvis (individual bones

and cartilaginous material), as illustrated in Fig. 4(a).

For the purpose of validating the morphometric

Fig. 2 Flowchart showing the four stages of the morphometric reconstruction modellingprocedure

50 P J Watson, P O’Higgins, M J Fagan, and C A Dobson

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

by guest on September 15, 2011pih.sagepub.comDownloaded from

Table 1 Description of morphometric landmark locations used for warping method and reconstruction

Landmark number Definition

1 Midpoint of iliac crest between anterior and posterior superior iliac spines2 Posterior superior iliac spine (PSIS)3 Posterior limit of auricular surface (PIIS)4 Upper (posterior) point of inflexion of auricular surface5 Junction of superior limit of sacro-iliac joint and iliac crest6 Lower (anterior) point of inflexion of auricular surface7 Point of maximal curvature of greater sciatic notch8 Ischial spine9 Anterior superior iliac spine (ASIS)10 Anterior inferior iliac spine (AIIS)11 Midpoint of scar of ilio–pubic epiphysis as it crosses the pelvic brim12 Point at which the superior pubic ramus forming the border of the obturator foramen is thinnest13 Inferior point on the obturator foramen border closest to the ischial tuberosity14 Inferior point on the obturator foramen border closest to the pubic symphysis15 Inferior-most point on pubic symphysis16 Superior-most point on pubic symphysis17 Posterior point at maximum width of pubic symphysis18 Anterior point at maximum width of pubic symphysis19 Midpoint of scar of ilio–pubic epiphysis at the acetabular rim20 Point directly opposite midpoint of acetabular notch on acetabular rim21 Midpoint of scar of ilio–pubic epiphysis at the acetabular rim22 Deepest central point in the acetabulum23 Midpoint of acetabular notch over ischiopubic epiphysis at limit of acetabulum24 Anterior limit of ischial tuberosity towards pubis25 Medial point of widest diameter of ischial tuberosity26 Superior limit of ischial tuberosity27 Lateral point of widest diameter of ischial tuberosity

Fig. 3 Digitized volumetric models of: (a) a fully fused 19 year old hemi-pelvis and (b) the samepelvis, warped to an estimated mean landmark configuration with an iliac depth of104 mm (the circles show the location of the 27 landmarks used in the geometricmorphometric analysis)

Validation of a morphometric reconstruction technique 51

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

by guest on September 15, 2011pih.sagepub.comDownloaded from

technique, the individual bones were then separated

from the connecting cartilaginous material (see

Fig. 4(b)) to obtain disarticulated bones (Stage 1.3

of Fig. 2).

The iliac depth of the 9 year hemi-pelvis was meas-

ured to be 104 mm. For Stage 2 the mean landmark

configuration for a hemi-pelvis with a comparative

iliac width was created within Morphologika, and

used to warp the adult pelvis to fit the estimated form

of the landmark configuration (see Fig. 3(b)).

The disarticulated 9 year pelvic bones were then

positioned manually within the warped template to

produce an anatomically realistic reconstructed

hemi-pelvis, as shown in Fig. 5 (Stage 3 of Fig. 2).

Due to the age of the disarticulated bones, the ischi-

um and pubis were expected to be close to fusion,

which begins between 5 and 10 years [20]. Conse-

quently, these two elements were first positioned

within the template with the spacing between the

two representing the vertical acetabular flange. The

ilium was subsequently positioned relative to this

structure, paying particular attention to the ilio-pu-

bic ridge and ilio-ischial articular surfaces, where the

anterior and posterior acetabular flanges are located.

The full hemi-pelvic structure was then completed

by digitally reconstructing the cartilaginous compo-

nents, by smoothly filling the gaps between the bones

once sited within the template (see Fig. 6). Using the

contours of the template between these gaps, the

curvature of the posterior and inferior acetabular

flanges was used to reconstruct a spherical joint.

Additional cartilage was defined to ensure a complete

obturator foramen. Due to the differences in form

between the 9 year bones and the warped template

model, some manual smoothing of cartilage was also

required to ensure continuity of material between the

pelvic bones. This occurred in regions where the

warped template slightly protruded past the exterior

of the disarticulated bones. In such instances the

surplus material of the warped template was cropped

until it correctly adjoined the disarticulated bones.

The accuracy of the reconstructed hemi-pelvis was

assessed through comparing its form to that of the

original articulated CT data. This was initially per-

formed through a principal component analysis (PCA),

to examine the extent of the difference in form be-

tween the reconstructed and originally articulated

hemi-pelves. The two hemi-pelves were then overlaid

to identify visually where any geometric deviations

occurred.

3 RESULTS

The PCA results (see Fig. 7) were calculated by

defining the 27 landmarks on both the reconstructed

and originally articulated hemi-pelves, and adding

these forms to the original morphometric dataset

(from Stage 1.1 of Fig. 2). A multivariate regression

analysis of this dataset within Morphologika enabled

comparison of the principal components between

the reconstructed and articulated models. For an

accurate reconstruction, the points representing the

originally articulated and reconstructed hemi-pelves

would be expected to be close to each other along all

principal components.

Fig. 4 Digitized volumetric model of the 9 year hemi-pelvis showing: (a) the complete structureincluding articulated bones and connecting cartilaginous material, and (b) the con-stituent disarticulated bones

52 P J Watson, P O’Higgins, M J Fagan, and C A Dobson

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

by guest on September 15, 2011pih.sagepub.comDownloaded from

Analysis of the PCA does indeed confirm an excellent

agreement between the two hemi-pelves along all

principal components (see Fig. 7). Through comparing

the positioning between the two in Fig. 7(a) and (b) to

that of Fig. 7(c), it is observed that the accuracy of the

reconstruction decreases along the first principal

component (PC1). Accuracy along the second principal

component (PC2) and third principal component

(PC3) is highlighted in Fig. 7(c) where the originally

articulated and reconstructed forms are overlying.

Therefore, although the accuracy of the reconstructed

model appears close to that of original articulation,

there is a slight difference between the forms.

To identify the locations where the original and

reconstructed hemi-pelves differ, the ilia of both

hemi-pelves were aligned in two planes, where one

axis was defined along the line connecting the

ASIS and PSIS and the other between the PSIS and

PIIS (see Fig. 1 and Table 1). Three anatomical

features of the ischium (spine, superior symphysis,

and inferior foramen) and pubis (superior symphy-

sis, inferior symphysis, and inferior foramen) were

selected, based on their ease of manual identifica-

tion, and the vectoral distances between these

points in the two models were then calculated to

quantify the error in the reconstructed pelvic bone

placement.

Thus, the actual extent to which the reconstructed

ischium and pubis were misplaced is presented in

Fig. 8. The largest error occurs at the superior pubic

symphysis, where a misplacement of 2.79 mm was

found. However, this is only 2.69 per cent of the

Fig. 5 Overlay of 9 year disarticulated pelvic bones into the geometry of the warped template

Fig. 6 Complete reconstruction of the 9 year hemi-pelvis with rearticulated bones andconnecting cartilage included

Validation of a morphometric reconstruction technique 53

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

by guest on September 15, 2011pih.sagepub.comDownloaded from

Fig. 7 Results of the multivariate regression analysis performed on the original morphometricdataset and both the articulated and the reconstructed 9 year hemi-pelvic forms, showing:(a) PC1 versus PC2, (b) PC1 versus PC3, and (c) PC2 versus PC3 (X 5 original mor-phometric dataset, &5 warped template, + 5 original articulated hemi-pelvis,65 recon-structed hemi-pelvis)

54 P J Watson, P O’Higgins, M J Fagan, and C A Dobson

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

by guest on September 15, 2011pih.sagepub.comDownloaded from

overall hemi-pelvic width (PW) (calculated as the

distance between the ASIS and PSIS). In contrast, the

smallest error is found at the ischical spine with a

displacement of 2.22 mm (2.14 per cent of PW).

Overall the average error in the ischial and pubic

landmarks of the reconstructed model was only

2.5 mm (2.41 per cent of PW), with a standard

deviation of 0.2 mm.

As the errors in the ischial and pubic landmarks are

closely matched, this would suggest that the relative

position between the ischium and pubis is close to

that of the original. This is confirmed through

calculating the vectoral distance between the superior

ischial tuberosity and pubic symphysis, and the

inferior ischial and pubic foramen on both models.

The intermodel comparison of these distances shows

that the error in the reconstructed positioning of the

ischial–pubic structure is minimal (0.83 mm and

0.04 mm respectively). Therefore, this suggests that it

is the positioning of this structure to the ilium that

may be slightly inaccurate. Finally, Fig. 9 shows an

overlay of the reconstructed and original articulated 9

year hemi-pelvic models, confirming that the ischium

and pubis may have been inaccurately positioned, to

produce a more posteriorly rotated ischial–pubic

structure compared to the original articulation; al-

though, as demonstrated above, the differences are in

fact very small.

Fig. 8 Positional error (in mm) between articulated and rearticulated models, as defined by thedistance between corresponding landmarks

Fig. 9 Overlay of reconstructed and original articulated 9 year hemi-pelvic models (light areasrepresent the original model, dark areas the reconstruction)

Validation of a morphometric reconstruction technique 55

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

by guest on September 15, 2011pih.sagepub.comDownloaded from

4 DISCUSSION

The aim of this paper was to illustrate and validate a

modelling technique to reconstruct damaged or dis-

articulated anatomical structures. Validation of the

method was performed by reconstructing a hemi-

pelvis from the disarticulated bones of a 9 year old,

and comparing this to the original articulated CT

data.

A PCA and intermodel overlay shows an accep-

table level of accuracy for the reconstructed hemi-

pelvis in comparison to the original articulation (see

Figs 7 and 8). A small difference between forms is

possibly due to the posteriorly placed ischial–pubic

structure in relation to the ilium when sited within

the template (see Fig. 9).

The PCA also highlights the differences between

the reconstructed model and the warped template.

Figure 7 displays noticeable differences between the

forms along all principal components, which are

partially evident in Fig. 5. In this instance, identifi-

cation of anatomical features such as the ilio–ischial,

ilio–pubic, and ischial–pubic articulation sites have

aided in the creation of an accurate reconstruction.

However, when reconstructing juvenile pelves of

younger ages, such articulating features are not

easily identifiable. Hence, the utilization of warped

templates is increasingly important in such cases.

The anatomical accuracy of the reconstructed

hemi-pelves is dependent on the geometry of the

warped template, which is in turn influenced by

landmark positioning. Accordingly, meticulous at-

tention is required in positioning of the landmarks to

ensure congruence between the morphometric data

and manually defined points. Accuracy could be

increased further through the application of more

landmarks, although their number will ultimately be

limited by the fact that the bony features of adult

pelves are not always identifiable on juvenile geome-

tries. However, as evident in Fig. 3(b), the warped

estimate of the obturator foramen and acetabulum

appears to provide an accurate anatomical repre-

sentation. It is possible that the reconstruction of the

pelvic bone alignment and cartilaginous geometry

could be susceptible to operator variance. Unfortu-

nately, as the reconstruction is a laborious activity,

comparing the results from a range of operators

would prove prohibitively time-consuming. How-

ever, operator variance is almost certainly reduced

by ensuring that the pelvic bones are aligned not only

by the warped template but also through anatomi-

cal features (i.e. acetabular articulation sites). This

should minimize interoperator variation in the

resulting morphologies.

The comparisons in the bone orientations are

made with respect to the manually defined anato-

mical landmarks, but of course accuracy in deter-

mining the exact location of an anatomical feature

can be subject to the quality of the CT scan resolu-

tion and to personal interpretation. Therefore, dur-

ing intermodel validation, landmarks selected for the

analysis are based on ease of identification, both on

an intra- and intermodel level, again with the aim of

reducing differences in morphology resulting from

different individual operators.

Additionally, the resulting digitized model can be

used in further GMM-based warping, to replicate the

geometry for a range of ages. A regression analysis of

the juvenile morphometric data enables the estima-

tion of shape variations throughout the age range.

Consequently, landmark positioning for a desired

age can be obtained and used to warp the created 9

year old model into different geometries. This ap-

proach gives the capacity to estimate complete hemi-

pelvic anatomies at any desired age, although the

differing degrees of ossification at different ages

would also have to be taken into consideration.

Future work plans to determine the effect of the

small positional differences witnessed in the recon-

struction of the 9 year old hemi-pelvis, through the

use of FE modelling. This will aim to determine the

sensitivity of the positioned errors on the strain

trajectories required for analysis of biomechanical

modelling. The effect of using different adult speci-

mens on which the warping is performed will also

be examined, to investigate the effect of different

morphologies on the subsequent warping.

ACKNOWLEDGMENTS

The authors gratefully acknowledge Dr WolfgangRecheis for providing access to clinical CT data of ajuvenile pelvis. Additional thanks are made to Prof.S. Black of the Department of Anatomy and ForensicAnthropology, University of Dundee.

F Authors 2011

REFERENCES

1 Turner, C. H. Three rules for bone adaptation tomechanical stimuli. Bone, 1998, 23(5), 399–407.

2 Frost, H. M. Bone ‘mass’ and the ‘mechanostat’: aproposal. Anat. Rec., 1987, 219, 1–9.

3 Huiskes, R., Ruimerman, R., van Lenthe, G., andJanssen, J. D. Effects of mechanical forces onmaintenance and adaptation of form in trabecularbone. Nature, 2000, 405, 704–706.

56 P J Watson, P O’Higgins, M J Fagan, and C A Dobson

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

by guest on September 15, 2011pih.sagepub.comDownloaded from

4 Rauch, F. and Schoenau, E. The developing bone:slave or master of its cells and molecules. Pediat.Res., 2001, 50(3), 309–314.

5 Ehrlich, P. J. and Lanyon, L. E. Mechanical strainand bone cell function: a review. Osteoporosis Int.,2002, 13, 688–700.

6 Wolff, J. Das Gesetz der Transformation derKnochen, 1892 (Hirschwald, Berlin).

7 Wolff, J. The law of bone remodelling (Translatedby P. Maquet and R. Furlong), 1986 (SpringerVerlag, Berlin).

8 Drapeau, M. S. M. and Streeter, M. A. Modellingand remodelling responses to normal loading inthe human lower limb. Am. J. Phys. Anthropol.,2006, 129(3), 403–409.

9 Ulrich, D., van Rietbergen, B., Laib, A., andRuegsegger, P. The ability of three-dimensionalstructural indices to reflect mechanical aspects oftrabecular bone. Bone, 1999, 25, 55–60.

10 Ryan, T. M. and Krovitz, G. E. Trabecular boneontogeny in the proximal femur. J. Human Evolu-tion, 2006, 51(6), 591–602.

11 Tsubota, K., Suzuki, Y., Yamada, Y., Hojo, M.,Makinouchi, A., and Adachi, T. Computer simula-tion of trabecular remodelling in human proximalfemur using large-scale voxel FE models; approachto understanding Wolff’s law. J. Biomechanics,2009, 42, 1088–1094.

12 Macchiarelli, R., Bondioli, L., Galichon, V., andTobias, P. V. Hip bone trabecular architectureshows uniquely distinctive locomotor behaviour inSouth African australopithecines. J. Human Evolu-tion, 1999, 26(2), 211–232.

13 Rook, L., Bondiolo, L., Kohler, M., Moya-Sola, S.,and Macchiarelli, R. Oreopithecus was a bipedalape after all. Evidence from the iliac cancellousarchitecture. Proc. Natl Acad. Sci. USA, 1999,96(15), 8795–8799.

14 Cunningham, C. and Black, S. Development of thefetal ilium – challenging concepts of bipedality.J. Anatomy, 2008, 214(1), 91–99.

15 Cunningham, C. and Black, S. Anticipating biped-alism: trabecular organization in the newbornilium. J. Anatomy, 2009, 214, 817–829.

16 Gosman, J. H. and Ketcham, R. A. Patterns inontogeny of human trabecular bone from Sun-Watch in the prehistoric Ohio Valley; general fea-tures of microstructural change. Am. J. Phys.Anthropol., 2008, 138(3), 318–332.

17 Dalstra, M., Huiskes, R., Odgaads, A., and vanErning, L. Mechanical and textural properties of

pelvic trabecular bone. J. Biomechanics, 1993, 26(5),523–535.

18 Dalstra, M. and Huiskes, R. Load transfer acrossthe pelvic bone. J. Biomechanics, 1995, 28(6), 715–724.

19 Majumder, S. and Roychowdhury, A. Biomecha-nical analysis of human pelvis under musculo-skeletal load using 3D finite element method. Int. J.Appl. Mechanics and Engng, 2005, 10(4), 647–665.

20 Scheuer, L. and Black, S. Developmental juvenileosteology, edition 1, 2000 (Academic Press, Lon-don).

21 Palastanga, N., Field, D., and Soames, R. Anatomyand human movement: structure and function,edition 4, 2002 (Butterworth-Heinemann, Edin-burgh).

22 Anderson, A. E., Peters, C. L., Tuttle, B. D., andWeiss, J. A. Subject-specific finite element model ofthe pelvis: development, validation and sensitivitystudies. J. Biomechanics, 2005, 117, 364–373.

23 Phillips, A. T. M., Pankaj, P., Howie, C. R., Usani,A. S., and Simpson, A. H. R. W. Finite elementmodelling of the pelvis: inclusion of muscular andligamentous boundary conditions. Med. EngngPhysics, 2006, 29, 739–748.

24 Dryden, I. L. and Mardia, K. V. Statistical shapeanalysis, 1998 (John Wiley & Sons Ltd, Chichester).

25 O’Higgins, P. Advances in approaches to the studyof morphological variation in the hominid fossilrecord: biology, landmarks and geometry. J. Anat-omy, 2000, 197, 103–120.

26 Zelditch, M. L., Swiderski, D. L., Sheets, H. D., andFink, W. L. Geometric morphometrics for biologists:a primer, 2004 (Elsevier Academic Press, NewYork).

27 Bookstein, F. L. Principal warps: thin plate splinesand the decomposition of deformations. IEEETrans. on Pattern Anal. Mach. Intell., 1987, 11,567–585.

28 Lapeer, R. J. A. and Prager, R. W. 3D shaperecovery of a newborn skull using thin-platesplines. Computerised Med. Imaging and Graphics,2000, 24, 194–304.

29 Kirkwood, R. N., Culham, E. G., and Costigan, P.Radiographic and non-invasive determination ofthe hip joint center location: effect on hip jointmoments. Clin. Biomechaincs, 1999, 14, 227–235.

30 O’Higgins, P. and Jones, N. Facial growth inCercocebus torquatus: an application of threedimensional geometric morphometric techniquesto the study of morphological variation. J. Anat-omy, 1998, 193, 251–272.

Validation of a morphometric reconstruction technique 57

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

by guest on September 15, 2011pih.sagepub.comDownloaded from