Bone quality assessment based on conebeam computed tomography imaging

Yan HuaOlivia NackaertsJoke DuyckFrederik MaesReinhilde Jacobs

Authors’ affiliations:Yan Hua, Olivia Nackaerts, Reinhilde Jacobs,Oral Imaging Center, School of Dentistry, OralPathology and Maxillofacial Surgery, Faculty ofMedicine, Katholieke Universiteit Leuven, Leuven,BelgiumJoke Duyck, Biomat Laboratory, School ofDentistry, Oral Pathology and MaxillofacialSurgery, Faculty of Medicine, KatholiekeUniversiteit Leuven, Leuven, BelgiumFrederik Maes, ESAT-PSI, Medical Imaging Center,Faculty of Medicine and Faculty of Engineering,Katholieke Universiteit Leuven, Leuven, Belgium

Correspondence to:Reinhilde JacobsOral Imaging Center,School of DentistryOral Pathology and Maxillofacial SurgeryFaculty of MedicineKatholiekeUniversiteit LeuvenKapucijnenvoer7, 3000 LeuvenBelgiumTel/Fax: þ32 16 332410e-mail: reinhilde.jacobs@med.kuleuven.be

Key words: cone beam computed tomography, fractal analysis, fractal dimension,

morphometry

Abstract

Objectives: The aim of this in vitro study was to investigate the accuracy of fractal analysis

and morphometry for bone quality assessment as measured with dual energy X-ray

absorptiometry (DXA).

Material and methods: Nineteen mandibular bone samples were used for the creation of

artificial bone lesions (n¼5) or decalcification (n¼12) to simulate osteoporosis; two

samples were used as controls. Cone beam computed tomography (CBCT) and DXA scans

were made before and after processing the samples. The image data obtained from the

CBCT scans were used to calculate the mean fractal dimension (FD), bone area and density

(morphometric analysis) of the samples. Bone mineral density (BMD) was obtained from the

DXA scans and set as a reference value for bone quality. The correlation between BMD and

FD and between BMD and morphometric results were calculated.

Results: A significant correlation between FD and BMD (r¼ þ0.71 to þ0.75; Po0.05) was

observed. Bone area and BMD of the specimens (r¼ þ0.69 to þ0.85; Po0.05) were also

significantly related, in contrast to the density analysis, for which no significant correlation

to BMD was found.

Conclusions: The results of this study suggest that fractal analysis and bone area

measurement have potential to evaluate bone quality on CBCT images, while density

measurement does not seem to be valid.

The definition of osteoporosis refers to

both low bone mass and micro-architec-

tural deterioration of the bone scaffold. The

presence of one or both factors results in

increased bone fragility and therefore in-

creased susceptibility to fractures (White

2002). Esposito et al. (1998) identified low

bone quality as one of the factors associated

with oral implant failures due to biological

causes. Such low bone quality or a quality

reduction can be caused by osteoporosis.

Therefore, preoperative bone quality as-

sessment is important for oral implant

surgery planning.

Bone mineral density (BMD) assessment

as well as micro-architectural analysis of

the bone, both have a nonnegligible role in

the diagnosis of osteoporosis (Apostol et al.

2006).

BMD, as a gold standard for the diagnosis

of osteoporosis, can be measured in differ-

ent ways. Dual-energy X-ray absorptiome-

try (DXA) can be used for osteoporosis

diagnosis at the spine and/or hip, and

also, although until now it has been used

solely in experimental settings, for jaw

bone evaluation (von Wowern 2001; Nack-

aerts et al. 2007).

Date:Accepted 22 October 2008

To cite this article:Hua Y, Nackaerts O, Duyck J, Maes F, Jacobs R. Bonequality assessment based on cone beam computedtomography imaging.Clin. Oral Impl. Res. 20, 2009; 767–771.doi: 10.1111/j.1600-0501.2008.01677.x

c� 2009 The Authors. Journal compilation c� 2009 John Wiley & Sons A/S 767

For studies on bone micro-architecture,

texture analysis has been applied earlier

(Apostol et al. 2006). Fractal analysis is

one of the approaches for texture analysis,

which received particular attention for ap-

plication in 2D radiographs (Apostol et al.

2006). In fractal analysis, there are different

methods to calculate fractal dimension

(FD), which can be used to compare normal

bone with osteoporotic bone (Geraets &

van der Stelt 2000). For example, Southard

et al. (2000) found that the average FD

decreased from 1.26 to 1.1 in radiographs

of decalcified human alveolar bone. In an

in vivo study (Southard et al. 2001) they

confirmed the clinical relevance of this

finding. FD and density of the alveolar

process bone were highly correlated, mean-

ing the FD could represent an objective

way to quantify bone quality, irrespective

of variations in exposure settings.

In the 3D domain, bone quality assess-

ment methods have also been studied.

Texture analysis has been applied in mi-

cro-CT (Apostol et al. 2006), while Houns-

field units (HU) have been used in spiral

CT as a measure, related to jaw BMD

(Stoppie et al. 2006).

Cone beam computed tomography

(CBCT) is a more recent development

than spiral CT. Its clinical application in

the field of dentomaxillofacial radiology is

gaining importance and spreading widely

(Guerrero et al. 2006; Loubele et al. 2006;

Scarfe et al. 2006). While the clinical re-

levance for presurgical assessment of jaw

bone density has been clearly demonstrated

(for a review see Esposito et al. 1998), a

reliable and easily applicable clinical tool

for this assessment does not yet exist. The

available research on CBCT-based bone

quality assessment is scarce and hampered

by the inherent technical constraints of

CBCT image data sets.

Indeed CBCT data have a larger amount

of scattered X-rays than conventional spiral

CT. This may enhance the noise in recon-

structed images, and thus affect the low-

contrast detectability (Endo et al. 2001).

Because of scatter and artifacts, HU values

in CBCT are not valid, and therefore the

method of correlating BMD to HU values

from CBCT is not ideal. Moreover, the

scatter and artefacts in CBCT get worse

around inhomogenous tissues with reduced

HU values up to 200 HU (Yoo & Yin

2006), which confirms that the HU in

CBCT is not a valid method for bone

quality assessment. Beam hardening is a

phenomenon resulting from the increase of

mean energy of the X-ray beam when it

passes through an object. Because of beam

hardening, the HU of certain structures

such as soft tissue and bone alters. The

single detector CBCT has a larger beam

width than the conventional multi-detec-

tor row CT. This causes a non-uniform

angular distribution of the X-ray beam

intensity known as the heel effect, which

also leads to HU that have no uniformity

either. This confirms that up till now

CBCT-based bone quality assessment is

neither accurate nor reliable, and thus,

there is an urgent need to find methods to

circumvent the shortcomings of this parti-

cular development, so as to have a reliable

way to assess bone quality.

Given the fact that HU values are not

valid in CBCT, there is a need for methods,

other than density measurements, for bone

quality assessment. Texture analysis may

thus come into play, which is strengthened

by the fact that bone quality may be

expressed by its micro-architectural com-

position.

The aim of this in vitro study was to

investigate the accuracy of fractal analysis

and morphometry for bone quality assess-

ment as measured with dual energy X-ray

absorptiometry (DXA). Fractal and mor-

phometric analyzes were based on CBCT

images.

Material and methods

Material

Human mandibular dry bone specimens

were obtained from the Department of

Anatomy (K. U. Leuven), after approval by

the ethical committee of the Catholic Uni-

versity of Leuven. A total of 19 specimens

were randomly allocated to two separate

groups: artificial bone lesion (group 1;

n¼ 5) and decalcification (group 2;

n¼ 12). Two samples were used as controls.

Methods

Preparation of bone specimens to simulatebone loss

Samples in group 1 were used to simulate

destruction of the trabecular bone struc-

ture. This was done by creating small bone

defects (diameter between 1 and 1.5 mm)

using a spoon excavator. Samples in group

2 were used to simulate bone loss by

decalcification. The samples were decalci-

fied with an HCl solution (Decals

, Serva,

Heidelberg, Germany) twice during

10 min. At each interval, the samples

were first rinsed with distilled water for a

few seconds in order to suppress decalcifi-

cation and then dried.

Methods for bone quality assessment

Imaging procedure. All samples were

weighed on a scientific scale, AB304-Ss

(N. V. Mettler-Toledo S. A., Zaventem,

Belgium; precision¼ 0.1 mg), before and

after the procedures for bone removal.

All samples were scanned with CBCT

and DXA before and after the procedures

for bone removal. For the CBCT images,

the samples were put into a polystyrene

container with water for soft tissue simula-

tion and scanned with an i-CATs

(Imaging

Sciences International Inc., Hatfield, PA,

USA) using the predefined protocol ‘Max-

illa 6 cm, 20 s’.

DXA scans were carried out with a

Hologic QDR-4500as

(Hologic Inc., Bed-

ford, MA, USA) to establish a reference

standard on bone density. The DXA scan-

ner was calibrated daily in accordance with

the manufacturer’s recommendations. The

regional high-resolution mode of the small

animal scan protocol [scan field 5

(width) � 7.4 (height) cm2, line spacing

and point resolution 0.0311 cm] was used.

The specimens were positioned on a plexi

support (thickness¼ 2 cm).

The CBCT image data were analyzed

using several bone quality assessment

methods, potentially suitable for CBCT

image analysis: 2D and 3D fractal analysis

and morphometric analysis. The methods

are described below.

2D fractal analysis. FD, which is the result

of fractal analysis, is a quantification of

surface roughness: the rougher the surface,

the larger the computed magnitude of FD

(Southard et al. 2001). FD was based on the

box counting method, which is only one of

several existing methods for fractal analysis.

MevisLabs

(MeVis Research GmbH,

Bremen, Germany) was used for this ana-

lysis. CBCT slices were consecutively pro-

cessed using this programming environ-

ment. The bone region on the slices was

segmented (‘live wire’ and ‘region grow-

Hua et al . Bone quality on CBCT images

768 | Clin. Oral Impl. Res. 20, 2009 / 767–771 c� 2009 The Authors. Journal compilation c� 2009 John Wiley & Sons A/S

ing’) and thresholding (window/leveling)

was applied to obtain a binary image (Fig.

1). Either the whole bone (cortical and tra-

becular bone) or only the trabecular bone

were set as the region of interest (ROI) and

fractal analysis was applied in Matlabs

(MathWorks Inc, Natick, MA, USA), using

the open source project FracLab. The mean

value of FD for each sample was calculated.

3D fractal analysis. Within the program-

ming environment of MevisLabs

, a 3D

cube counting method was developed for

fractal analysis as an extension to the 2D

box counting method. For this purpose, all

slices were processed in the volume mode.

Either the whole bone (cortical and trabe-

cular bone) or only the trabecular bone was

set as the volume of interest and fractal

analysis was applied. FD was determined

by the relationship between the number of

voxels covering bone and the voxel size

(cube size). The slope of the logarithm-

fitted curve of inverse voxel size vs. the

number of cubes needed to cover the bony

structure represented the FD (Zhang et al.

2006). An illustration of such curve is

shown in Fig. 2.

Morphometry. Morphometry was done

with the commercially available software

AxioVisions

(Carl Zeiss MicroImaging

GmbH, Koln, Germany). Each slice was

saved as a jpg image, on which morpho-

metry was performed. In this study, the

mean density value based on gray values

and the area of bony structure in each slice

were measured.

Data analysis

As data were not normally distributed,

Spearman’s r was calculated between FD

and the corresponding BMD and between

morphometric results and the correspond-

ing BMD. To assess the change in FD and

morphometric results, only descriptive sta-

tistics were used because of the small

sample size.

Results

Weight

The mean weight of the samples is shown

in Table 1. The weight of all samples

except for the control samples, decreased

after bone removal procedures.

2D fractal analysis

The mean FD of all CBCT slices was

calculated for each sample. Table 2 shows

the mean FD of the processed samples for

both methods. Although we were not able

to perform statistical analysis due to the

size of the experimental groups in this

exploratory study, it became evident that

the FD decreased after the modifications.

This was true for the analysis including the

cortical bone and the one including only

trabecular bone. Analysis of the control

samples showed a stable FD.

3D fractal analysis

The number of voxels containing bone for

the same cube size decreased after the

modifications to the bone samples (struc-

ture and Ca content). The FD decreased

when analyzing the entire bone sample.

However, FD increased when analyzing

only the trabecular bone.

Morphometry

The mean density, based on gray values,

and bone area of each sample are listed in

Table 3.

Fig. 1. (a) Cone beam computed tomography

(CBCT) slice in binary format, including cortical

bone. (b) CBCT slice in binary format, excluding

cortical bone.

1/Voxel size

N v

oxe

ls c

ove

rin

g b

on

e vo

lum

e o

f in

tere

st

Fig. 2. Illustration of the fractal analysis of one sample in 3D. The slope, a measure for fractal dimension (FD),

decreases after decalcification.

Table 1. Changes in bone specimen weight based on various procedures

Weight beforemodification (g)

Weight aftermodification (g)

Change % change

Artificial bone lesion 2.88 2.72 � 0.16 � 5.69Decalcification 2.97 2.31 � 0.66 � 22.11Control 1.89 1.88 � 0.01 � 0.27

Table 2. Results of 2D fractal analysis: (a) method 1 (both cortical and trabecular bone),(b) method 2 (trabecular bone only)

Mean FD (SD) beforemodification

Mean FD (SD) aftermodification

FD change % change

(a)Control samples 1.00 (0.07) 0.99 (0.08) � 0.01 � 1.0Modified samples 1.05 (0.04) 0.98 (0.04) � 0.07 � 6.7

(b)Control samples 0.71 (0.05) 0.71 (0.05) 0.00 0.0Modified samples 0.88 (0.14) 0.82 (0.13) � 0.06 � 6.8

Hua et al . Bone quality on CBCT images

c� 2009 The Authors. Journal compilation c� 2009 John Wiley & Sons A/S 769 | Clin. Oral Impl. Res. 20, 2009 / 767–771

The area decreased after the modifica-

tions to the samples. The density measured

in gray values increased.

BMD

Each sample was scanned twice with the

DXA device. The correlation between

BMD and FD, and between BMD and

area were calculated for both scans. A

significant correlation of þ 0.71 to þ 0.75

(Po0.05) was found between FD and

BMD. Likewise, area and BMD correlated

significantly (r¼ þ 0.69 to þ 0.85;

Po0.05). In contrast, density measured in

gray values did not significantly correlate

with BMD (r¼ þ0.13 to þ 0.21; P40.5).

Discussion

Considering the importance of bone quality

in jaw bone implant surgery or related

therapy (Bryant 1998), the need for an

accurate and reliable clinical tool for quan-

tifying this preoperatively is evident. Un-

fortunately such tool is not yet available.

At the same time, in many cases, low dose

CBCT can be advised for implant place-

ment, considering the possibility to gather

clinically relevant 3D data at a low dose

(Guerrero et al. 2006). However, CBCT

does not allow reliable and accurate bone

quality assessment when focusing on the

inherent radiographic density information

that is otherwise expressed by HU (Yama-

shina et al. 2008). This has led to the

present study from which the outcome

illustrates that there may be a potential

for structural analysis to become the

method for standard clinical jaw bone qual-

ity assessment. FD and bone area represent

objective measurements that avoid the dif-

ficulties associated with bone density eva-

luation in CBCT.

For the present study, we opted for an in

vitro approach to allow a reliable validation

and accuracy assessment of the new tech-

niques. Trabecular bone destruction and

decalcification of the bone specimens,

representing micro-architectural bone

deterioration and decreased bone mass,

were an attempt to characterize osteoporo-

tic bone.

The main finding of this study was that

2D FD of CBCT images was significantly

related to the actual jaw BMD. This was in

accordance with a study of Southard et al.

(2001), who found that the average FD

decreased from 1.26 to 1.1 in radiographs

of decalcified human alveolar bone.

Besides, the present study showed that

3D fractal analysis of the CBCT image data

facilitated quantification of the induced

decrease in bone mass by either artificial

bone lesion creation or decalcification.

When performing 3D fractal analysis based

on the trabecular bone alone, the calculated

FD increased after osteoporosis simulation.

Because this was a preliminary study on

development of an innovative method, this

study could not be related nor compared

with previous studies regarding this parti-

cular method in CBCT image data assess-

ment. In the 2D domain, there has been

previous evidence that the FD increased

after osteoporosis simulation. For example,

Berry et al. (1996) decalcified human ver-

tebral bone. Digitized radiographs were

made every 30 s and the FD increased

from 2 to 3. These findings suggest that

FD does not consistently decrease after

osteoporosis simulation. Therefore, it

could be more meaningful to observe both

bone mass and FD, or more in general

structural properties, rather than only con-

sidering FD (Jiang et al. 1999). We will

continue in vitro studies with larger sample

size and validate this approach in clinical

prospective studies.

The morphometric measurements

showed a decreased bone area, correspond-

ing well with the bone mineral reduction

for all samples after the modifications.

BMD does obviously not only depend on

the calcium content of bone alone, but also

on the structural characteristics. The ab-

sence of a clear correlation between density

based on the gray values and BMD based on

DXA measurements illustrated that den-

sity measurements based on CBCT images

were not useful as such, because of inten-

sity inhomogeneity.

As far as the researchers know, the pre-

sent study was the first one proposing an

objective bone quality assessment study

based on CBCT technology for dentomax-

illofacial applications. It might also be

stressed that this was the primary study

reporting the use of radiographic CBCT

slices in jaw bone morphometry.

Conclusions

From the present preliminary results, frac-

tal analysis and bone area measurement

seem to have some potential for bone

quality assessment on CBCT images,

while density measures do not seem valid.

More elaborate studies are necessary to

verify these results and test their clinical

applicability.

References

Apostol, L., Boudousq, V., Basset, O., Odet, C., Yot,

S., Tabary, J., Dienten, J.M., Boller, E., Kotzki,

P.O. & Peyrin, F. (2006) Relevance of 2D radio-

graphic texture analysis for the assessment of 3D

bone micro-architecture. Medical Physics 33:

3546–3556.

Berry, J.L., Towers, J.D., Webber, R.L., Pope, T.L.,

Davidai, G. & Zimmerman, M. (1996) Change

in trabecular architecture as measured by fractal

Table 3. Results for morphometry: (a) in goup 1 (artificial bone lesion), (b) in group 2(decalcification), and (c) in control group

Sample Before lesion After lesion Change % change

(a)Density (gray value) 240.47 261.25 20.78 8.64Area (pixel) 65,197 60,815 � 4382 � 6.70

Sample Beforedecalcification

Afterdecalcification

Change % change

(b)Density (gray value) 226.26 238.02 11.76 5.20Area (pixel) 40,814 32,846 � 8328 � 20.40

Sample Before rinsing After rinsing Change % change

(c)Density (gray value) 212.28 218.37 6.09 2.87Area (pixel) 34,216 33,466 750 2.20

Hua et al . Bone quality on CBCT images

770 | Clin. Oral Impl. Res. 20, 2009 / 767–771 c� 2009 The Authors. Journal compilation c� 2009 John Wiley & Sons A/S

dimension. Journal of Biomechanics 29: 819–

822.

Bryant, S.R. (1998) The effects of age, jaw site, and

bone condition on oral implant outcomes. Inter-

national Journal of Prosthodontics 11: 470–490.

Endo, M., Tsunoo, T., Nakamori, N. & Yoshida, K.

(2001) Effect of scattered radiation on image noise

in cone beam CT. Medical Physics 28: 469–474.

Esposito, M., Hirsch, J.M., Lekholm, U. & Thom-

sen, P. (1998) Biological factors contributing to

failures of osseointegrated oral implants. Eur-

opean Journal of Oral Sciences 106: 721–764.

Geraets, W.G.M. & van der Stelt, P.F. (2000) Fractal

properties of bone. Dentomaxillofacial Radiology

29: 144–153.

Guerrero, M.E., Jacobs, R., Loubele, M., Schutyser,

F., Suetens, P. & van Steenberghe, D. (2006)

State-of-the-art on cone beam CT imaging for

preoperative planning of implant placement. Clin-

ical Oral Investigations 10: 1–7.

Jiang, C., Giger, M.L., Chinander, M.R., Martell,

J.M., Kwak, S. & Favus, M.J. (1999) Character-

ization of bone quality using computer-extracted

radiographic features. Medical Physics 26: 872–

879.

Loubele, M., Maes, F., Schutyser, F., Marchal, G.,

Jacobs, R. & Suetens, P. (2006) Assessment

of bone segmentation quality of cone-beam CT

versus multislice spiral CT: a pilot study. Oral

Surgery, Oral Medicine, Oral Pathology, Oral

Radiology and Endodontics 102: 225–234.

Nackaerts, O., Jacobs, R., Horner, K., Zhao, F.,

Lindh, C., Karayianni, K., van der Stelt, P., Pavitt,

S. & Devlin, H. (2007) Bone density measure-

ments in intra-oral radiographs. Clinical Oral

Investigations 11: 225–229.

Scarfe, W.C., Farman, A.G. & Sukovic, P. (2006)

Clinical applications of cone-beam computed to-

mography in dental practice. Journal of Canadian

Dental Association 72: 75–80.

Shapurian, T., Damoulis, P.D., Reiser, G.M., Grif-

fin, T.J. & Rand, W.M. (2006) Quantitative

evaluation of bone density using the Hounsfield

index. The International Journal of Oral & Max-

illofacial Implants 21: 290–297.

Southard, T.E., Southard, K.A., Jakobsen, J.R., Hil-

lis, S.L. & Najim C, A. (2000) Fractal dimension

in radiographic analysis of alveolar process

bone. Oral Surgery, Oral Medicine, Oral Pathol-

ogy, Oral Radiology and Endodontics 89: 569–

576.

Southard, T.E., Southard, K.A. & Lee, A. (2001)

Alveolar process fractal dimension and postcranial

bone density. Oral Surgery, Oral Medicine, Oral

Pathology, Oral Radiology and Endodontics 91:

486–491.

Stoppie, N., Pattijn, V., Van Cleynenbreugel, T.,

Wevers, M., Vander Sloten, J. & Naert, I. (2006)

Structural and radiological parameters for the

characterization of jaw bone. Clinical Oral

Implants Research 17: 124–133.

von Wowern, N. (2001) General and oral aspects of

osteoporosis: a review. Clinical Oral Investiga-

tions 5: 71–82.

White, S.C. (2002) Oral radiographic predictors of

osteoporosis. Dentomaxillofacial Radiology 31:

84–92.

Yamashina, A., Tanimoto, K., Sutthiprapaporn, P. &

Hayakawa, Y. (2008) The reliability of computed

tomography (CT) values and dimensional

measurements of the oropharyngeal region using

cone beam CT: comparison with multidetector

CT. Dentomaxillofacial Radiology 37: 245–251.

Yoo, S. & Yin, F.F. (2006) Dosimetric feasibility of

cone-beam CT-based treatment planning com-

pared to CT-based treatment planning. Inter-

national Journal of Radiation Oncology, Biology

and Physics 66: 1553–1561.

Zhang, L., Liu, J.Z., Dean, D., Sahgal, V. & Yue,

G.H. (2006) A three-dimensional fractal analysis

method for qualifying white matter structure in

human brain. Journal of Neuroscience Methods

150: 242–253.

Hua et al . Bone quality on CBCT images

c� 2009 The Authors. Journal compilation c� 2009 John Wiley & Sons A/S 771 | Clin. Oral Impl. Res. 20, 2009 / 767–771