Technology and Health Care 13 (2005) 497–510
IOS Press
497
Usefulness of quantitative ultrasound in
evaluating structural and mechanical
properties of bone: Comparison of
ultrasound, dual-energy X-ray
absorptiometry, micro-computed
tomography, and mechanical testing of
human phalanges in vitro
C. Wüstera , F. de Terlizzib , S. Beckera , M. Cadossib , R. Cadossib and R. Müllerc,∗
a Department
of Endocrinology, University of Heidelberg & Clinic for Endocrinology, Bahnhofplatz 2,
D-55116 Mainz, Germany
b IGEA Biophysics Laboratory, Carpi (MO), Italy
c Orthopedic Biomechanics Laboratory, Beth Israel Deaconess Medical Center and Harvard Medical
School, Boston, USA
Received in final form 23 March 2005
Abstract. Introduction: Ultrasound studies evaluating bone tissue generally concentrate on two parameters – velocity and
attenuation. This study aimed to determine whether ultrasound signal analysis techniques could provide additional information
on the structural and mechanical characteristics of bone.
Materials and methods: In vitro measurements were made on 26 left index fingers from human cadavers. Ultrasound
measurements at the distal metaphysis and epiphysis; dual-energy X-ray absorptiometry of the whole phalanx; micro-computed
tomography at the distal quarter of the phalanx (that is, the distal epiphysis and metaphysis), and mechanical three-point bending
tests were performed. Univariate and multivariate linear regression techniques were used to analyze the results.
Results: The ultrasound parameters, speed of sound and ultrasound peak amplitude correlated significantly with the three
micro-computed tomography measures used to describe the characteristics of mineralized material (r = 0.69–0.79, p < 0.05).
Low frequency ultrasound correlated significantly with micro-computed tomography parameters describing inter-trabecular or
marrow spaces (r = 0.68–0.78, p < 0.05). Comparison of ultrasound parameters with geometric characteristics showed that
while speed of sound and ultrasound peak amplitude were related to the cortical area, moment of inertia, and mechanical load
(r = 0.57–0.83, p < 0.05), the amplitude of the fastest part of the ultrasound signal and full width at 80% maximum of the low
frequency peak were related to the relative area of the medullary canal (r = 0.40–0.43, p < 0.05).
Discussion: Quantitative ultrasound may provide information on structural, material and mechanical characteristics of
bone to the same extent and even better than DXA Bone Mineral Density. These results have been obtained by a complete
and exhaustive use of QUS technology in situ but under clinical conditions. The ultrasound parameters, correctly used and
combined, seem to be effective tools for investigating bone tissue.
∗
Address for correspondence: Ralph Müller, PhD, Institute for Biomedical Engineering, Swiss Federal Institute of Technology
0928-7329/05/$17.00 © 2005 – IOS Press and the authors. All rights reserved
498
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
1. Introduction
Ultrasound is commonly used in clinical practice to determine the risk of fracture, especially in patients
with postmenopausal osteoporosis [1]. Numerous studies affirm its effectiveness in investigating the rapid
bone loss seen after the menopause [2,3], in identifying older patients who are at high risk of fracture [4–8]
and in following up the effects of treatment in patients with osteoporosis [9,10].
The human phalanx is one of the most metabolically active parts of the skeleton [11,12]. Bone
remodeling plays an important part in its biomechanical and structural equilibrium. As people age,
the phalanges undergo morphological changes such as resorption of the trabeculae at epiphyseal and
metaphyseal levels and enlargement of the medullary canal, connected with an endosteal resorption
which contributes to the thinning of the cortex. In addition, resorption occurs in the cortical tissue itself
and increases its porosity [13].
The phalanx is made up of trabecular bone (seen in greatest quantity at the epiphyseal and metaphyseal
levels) and cortical bone (which constitutes the diaphysis and, in part, the metaphysis). At the epiphysis,
trabecular organization together with bone mineral density is particularly important in transmitting the
ultrasound impulse, since the cortical component is limited to a thin shell. In the proximal metadiaphyseal regions, the geometric distribution of the cortical bone surrounding the medullary canal has
the greatest role in transmitting the ultrasound signal, especially in older people, since the trabecular
bone in this area is the first to be resorbed [14].
Several studies have aimed to determine exactly how the density and structure of bone tissue influence
the propagation of ultrasound energy [15]. Various in vitro and in vivo studies have shown how different
parts of the ultrasound signal can be used to describe the properties of bone density and architecture [16–
18]. The velocity with which ultrasound crosses bone tissue is well correlated with bone mineral
density [19–21]; on the other hand, parameters describing morphological characteristics can supply
information on elasticiy [22,23].
We aimed to identify quantitative ultrasound parameters which could provide information not only on
bone density but also on the architecture and mechanical resistance of bone tissue. These parameters
could then be used in clinical studies to collect the sort of data that are not usually provided by the more
commonly used densitometric techniques. We investigated the factors influencing the propagation of
ultrasound waves through bone tissue by in vitro measurements of phalangeal bone samples and then
compared these results with those obtained from dual-energy X-ray absorptiometry, micro-computed
tomography, and mechanical testing.
2. Materials and methods
We planned to analyze the proximal phalanx of the left index finger from 26 human cadavers and
to evaluate the relationship between quantitative ultrasound, dual-energy X-ray absorptiometry, microcomputed tomography, and mechanical tests at this area.
2.1. Dual X-ray absorptiometry
Densitometric evaluation of the phalanges was performed with a QDR 1000 Hologic densitometer
(Waltham, MA, USA). Bone mineral density measurements, expressed in g/cm 2 , were obtained for each
phalanx. The whole phalanx was radiographed in the anterior/posterior position.
(ETH) and University of Zürich, Moussonstrasse 18, CH-8044 Zürich, Switzerland. Tel.: +41 44 632 4592; Fax: +41 44 632
1214; E-mail: [email protected].
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
A
499
B
Fig. 1. Selection of the region of interest at the epiphyseal condyle site (A, black circle) and the section corresponding to the
meta-diaphyseal site (B, white square).
2.2. Quantitative ultrasound
Before measurement, phalanges were immersed overnight in water in a vacuum pump. Each phalanx
was measured twice (see Fig. 1). Firstly, ultrasound probes were placed at the distal epiphyseal condyles
(region A). Secondly, probes were placed at the meta-diaphyseal position where the margins of the
phalanx become parallel (region B). Measurements were carried out in the medio-lateral direction, as
they are in general made in clinical practice. Phalanges were kept in degassed water to which detergent
had been added to improve acoustic coupling.
The ultrasound measurements were performed using a DBM Sonic 1200 (IGEA, Carpi, Italy). This
instrument uses two ultrasound coaxial probes, each 14 mm diameter. Probes are mounted on a caliper.
An ultrasound impulse with a main frequency of 1.25 MHz is transmitted through the bone placed
between the emitting and receiving probes. The device is able to digitize and record the complete signal
received, thus enabling comprehensive analysis of the signal characteristics.
2.2.1. Ultrasound parameters
The fastest part of the signal received is the one propagated through bone. For this reason, we analyzed
only those parts of the signal corresponding to a propagation speed higher than the speed measured when
the bone sample was not present in the solution. Using this part of the ultrasound signal [22], we were
able to calculate a series of ultrasound parameters (Fig. 2). Measurements were made at the epiphysis
and the metaphysis as follows:
– Speed of sound, independent of ultrasound attenuation. Speed of sound is calculated as the ratio of
the distance traveled by the impulse (distance between the probes) and the time taken by the signal
to travel this distance.
– Ultrasound peak amplitude. The maximum amplitude of the signal, providing information on how
much the signal has been attenuated in passing through the sample.
500
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
US signal received
25
US signal received
25
20
20
15
US signal Am pliude
UPA
mV
15
mV
SoS
10
10
5
5
0
0
5
6
7
µs
8
9
10
5
6
US signal received
25
7
µs
8
9
10
Fourier spectrum
1000
20
Fourier coeff.
mV
15
FW80M low
800
Fast Wave Amplitude
FWA
10
5
600
Freq Low
400
200
0
0
5
6
7
µs
0,0
8
9
10
0,5
1,0
1,5
2,0
Freq (MHz)
2,5
3,0
Fig. 2. The ultrasound parameters investigated: speed of sound (SOS), ultrasound peak amplitude (UPA), fast wave amplitude
(FWA), low frequency (FREQ LOW), and full width at 80% maximum of the low frequency peak (FW80M).
– Fast wave amplitude. The amplitude of the first peak of the signal, the part that crossed the sample
and reached the receiving probe first.
– Frequency content. The analysis of the frequency domain of the signal. This showed the presence of
two frequency peaks – one at high frequencies (> 1 MHz) and one at low frequencies (< 1 MHz).
The two peaks were analyzed separately and two characteristic magnitudes were calculated: the
value in frequency of the centroid of the low frequency peak (called LOW FREQ) and the width of
that peak at 80% of its maximum amplitude (FW80M).
2.3. Micro-computed tomography
A Scanco µCT 20 (Scanco Medical, Bassersdorf, Switzerland) was used. The source used is an X-ray
tube with a 10 µm focal spot. To perform the measurement, the bone is placed in a sample holder
filled with saline on a support revolving along the axial direction. Measurements were carried out with
a nominal isotropic resolution of 34 µm. Computed tomography involved a region of approximately
13 mm, starting from the distal end of the phalanx. This resulted in the scanning of 386 sections for each
sample, comprising the whole of region A and part of region B. All data were collected on tape.
The micro-computed tomography investigation provides three-dimensional images of the site analyzed,
but it can also supply morphological and structural indices. In this case, indices were calculated in a
volume of interest, defined as the sphere of maximum possible volume within the distal end of the
phalanx, and containing only trabecular bone (see Fig. 1A).
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
501
2.3.1. Parameters describing mineralized bone (“full” volumes)
The parameters supplied can easily be grouped into those describing “full” volumes – that is, mineralized bone – and those describing the “empty” volumes – that is, the medullary canal and the medullary
bone cavities [24]. The parameters that describe the full volumes are the following:
– BV/TV, the ratio between the volume of the mineralized part and the total volume analyzed (volume
of interest). This represents the bone volume density of the region analyzed.
– BS/BV, the ratio between the surface of the mineralized part of the bone and the volume of the
mineralized part of the bone. This is an index of the geometric distribution of the mineralized part
of the bone.
– Tb.Th, the model-independent, mean thickness of the trabeculae [25].
2.3.2. Parameters describing medullary canal and empty cavities (“empty” volumes)
Those parameters that describe the “empty” volumes are as follows:
– BS/TV, the ratio between the surface of the mineralized part of the bone and the total volume analyzed.
This index expresses the amount of interfaces present in the volume of the region analyzed.
– Tb.Sp, the model-independent mean separation of the trabeculae [25], that is, the mean thickness of
the intra-trabecular cavities.
– H1, H2, H3, eigen values of the normalized mean intercept length tensor. These are directional
parameters measuring the mean distance of the interfaces between mineralized matrix and cavity [26].
2.3.3. Non-metric composite parameters
Two non-metric, composite parameters were also provided, and these were calculated on the whole
region within 13 mm of the distal end of the phalanx. These measures are as follows:
– The structure model index. This is a parameter characterizing the model type of the trabeculae –
that is, plate-like versus rod-like – where a perfect plate-like structure indicates a value of 0 and a
rod-like structure has a value of 3 [27].
– The trabecular bone pattern factor. This calculates the relative amount of convex to concave
structures in the bone [28]. This parameter is often used as an alternative measure of threedimensional trabecular connectivity.
2.3.4. Parameters related to cortical bone
We were able to analyze the digitized image of each section of the phalanx at any distance between
the distal end and 13 mm towards the diaphysis. Thus, we could identify exactly where the ultrasound
measurement had been performed in region B. This allowed morphometric characterization of the sample
by micro-computed tomography imaging by considering place at which the ultrasound measurement had
been performed (Fig. 1B) as the place corresponding to the transmission of the US energy produced in
the central region of the circular transducer. In the most proximal part of the scan, the micro-computed
tomography images already refer to the meta-diaphyseal area. In this area, parameters more related to
cortical bone can be obtained as follows:
– Cortical area. Measurement of the cortical area of the section analyzed is an index of the amount
of mineralized cortical bone in the diaphysis.
– Moment of inertia. This is an index of the amount and geometric distribution of mineralized cortical
bone in the diaphysis.
– Relative medullary canal area. This is obtained from the ratio between the medullary canal area
and the total phalanx area in the section analyzed. It is an index of the size of the medullary cavity
of the bone.
502
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
2.4. Biomechanical testing
All samples underwent three-point-bending mechanical testing. We used a universal testing system
(Frank GmbH, Weinheim, Germany) with a constant deformation velocity of 0.5 mm/min until failure
occurred. The maximum load (F MAX , at failure, and expressed in N) was obtained for each sample. In
order to adjust for the different thickness of the bones in the diaphysis, we defined a normalized maximum
load (FMAX /mm, expressed in N/mm) which was calculated by dividing F MAX by the anterio-posterior
diaphyseal thickness of the sample.
We also calculated the applied maximum stress (σ MAX ), defined as the applied maximum load divided
by the cross-sectional area of the bone at the site of failure. In calculating this we assumed that the
diaphyseal region of the phalanx was shaped like a hollow cylinder. The only phalangeal site involved
in this type of test was the meta-diaphyseal one.
2.5. Statistical analysis
The associations between the ultrasound, densitometric, mechanical, morphological, and structural
variables were studied with univariate and multivariate linear regression analysis. For each regression
analysis, we calculated the Pearson correlation coefficients with the relevant level of statistical significance of each independent variable. The level of significance in the univariate model was assumed to be
p < 0.05. The multivariate model was used to determine which of the ultrasound parameters best predicted certain geometric and microstructural characteristics of the bone. Stepwise forward multivariate
analysis was performed. Each independent variable, from all the ultrasound parameters involved in each
measurement, was entered in the model when its relative independent significance in the overall model
was p < 0.10. For each entered independent variable we reported the partial independent correlation
coefficient and the relative significance in the overall model. For each overall model we reported the
correlation coefficient and relative standard error of the estimate (RMSE). The statistical models were
calculated with SPSS software (SPSS Inc., Chicago, IL, USA).
3. Results
3.1. Bone samples
There were 26 phalanges – 12 from male cadavers and 14 from female cadavers; the age at death
ranged from 56–96 years. Soft tissue was removed from all samples and bones were kept in a 7%
formalin solution until further analysis.
3.2. Measurements at the epiphyseal site (position A)
We had to exclude 10 ultrasound measurements performed at the epiphysis; in those bones the
epiphyseal capsule was broken and most of the bone was missing. The selection of the volume of
interest for µCT analysis was therefore arbitrary and the results of the analysis not reliable. Furthermore
ultrasound propagation was extremely difficult and identification of ultrasound transmitted pulse from
background electronic noise was not possible. In the remaining 16 samples, the ultrasound speed at the
epiphyseal site was lower and was correlated with the speed measured in the meta-diaphyseal area, with
r = 0.83 (p < 0.01).
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
503
Table 1
Correlation coefficients (r) between quantitative ultrasound parameters plus bone mineral density determined by dual-energy
X-ray absorptiometry (DXA) and micro-computed tomography indices at the site of the epiphyseal condyles. In brackets the p
values are reported
Ultrasound and
DXA parameters
Indices of mineralized bone
Indices of intra-trabecular or marrow
Composite indices
(“full” spaces)
spaces (“empty” spaces)
BV/TV BS/BV Tb.Th
BS/TV Tb.Sp
H1
H2
H3
SMI
TBPf
Speed of sound
0.70
−0.72
0.71
0.19
−0.48 −0.17 −0.17 −0.32 −0.77
−0.77
(0.004) (0.002) (0.003)
(n.s.)
(n.s.) (n.s.) (n.s.) (n.s.) (0.001) (0.001)
Ultrasound peak amplitude
0.79
−0.72
0.71
0.41
−0.68 −0.35 −0.48 −0.53 −0.71
−0.71
(0.001) (0.002) (0.003)
(n.s.) (0.005) (n.s.) (n.s.) (0.04) (0.003) (0.003)
Fast wave amplitude
−0.18
0.11 −0.11
−0.17
0.22
0.19 0.17 0.23
0.25
0.22
(n.s.)
(n.s.)
(n.s.)
(n.s.)
(n.s.) (n.s.) (n.s.) (n.s.)
(n.s.)
(n.s.)
Centroid of low frequency
0.72
−0.53
0.50
0.70
−0.78 −0.71 −0.68 −0.71 −0.71
−0.76
peak (LOW FREQ)
(0.009) (n.s.)
(n.s.)
(0.01) (0.003) (0.01) (0.02) (0.01)
(0.01) (0.004)
Width of LOW FREQ at
0.46
−0.55
0.55
0.00
−0.25 −0.04 0.00 −0.10 −0.64
−0.55
80% of amplitude (FW80M) (n.s.)
(n.s.)
(n.s.)
(n.s.)
(n.s.) (n.s.) (n.s.) (n.s.)
(0.03)
(n.s.)
DXA bone mineral density
0.76
−0.69
0.70
0.37
−0.63 −0.36 −0.35 −0.48 −0.86
−0.79
(0.001) (0.005) (0.004)
(n.s.)
(0.01) (n.s.) (n.s.) (n.s.) (0.0001) (0.0001)
BV/TV = the ratio between the volume of the mineralized part and the total volume of bone analyzed; BS/BV = the ratio
between the surface of the mineralized part of the bone and the volume of the mineralized part of the bone; Tb.Th = the mean
thickness of the trabeculae; BS/TV = the ratio between the surface of the mineralized part of the bone and the total volume
analyzed; H1, H2, H3 = eigen values of the normalized mean intercept length tensor; SMI = structure model index; TBPf =
trabecular bone pattern factor.
The ultrasound parameters which correlated with the bone mineral density of the whole phalanx
were speed of sound (r = 0.77, p < 0.01) and FW80M (r = 0.68, p < 0.01). We then evaluated
the associations between the ultrasound parameters and the structural parameters obtained with microcomputed tomography. The correlation coefficients which proved to be statistically significant (p <
0.05) are reported in Table 1.
Speed of sound and ultrasound peak amplitude were significantly correlated with the micro-computed
tomography parameters describing the mineralized bone (full volumes: BV/TV, BS/BV, Tb.Th), while
LOW FREQ was significantly linked to the micro-computed tomography parameters describing the
medullary canal and cavities (empty spaces: BS/TV, Tb.Sp, H1, H2, H3). Fast wave amplitude showed
no significant correlations with any of the micro-computed tomography parameters.
The structure model index and the trabecular bone pattern factor are composite parameters indicating
an overall model of structure. These parameters were correlated with speed of sound, ultrasound peak
amplitude, and LOW FREQ. The structure model index was also correlated with FW80M.
We then evaluated the influence of the bone mineral density of the whole phalanx on all the parameters
determined with micro-computed tomography. Table 1 also reports the results of the linear correlations.
Bone mineral density correlated significantly only with those micro-computed tomography parameters
which had previously shown a significant correlation with speed of sound and ultrasound peak amplitude –
that is, BV/TV, BS/BV, and Tb.Th.
Use of multivariate analysis to determine which of the ultrasound parameters were the best predictors of
the variable BV/TV, showed that speed of sound (r p = 0.48, p = 0.02) and LOW FREQ (rp = 0.47, p =
0.02) contributed significantly and independently to the model, with a global correlation coefficient of
r = 0.88 (p = 0.002, RMSE = 3.10). Multivariate analysis also enabled us to establish that the best
ultrasound predictors of structure model index were speed of sound (r p = −0.53, p = 0.008) and LOW
FREQ (rp = −0.44, p = 0.019) with a correlation coefficient of r = 0.90 (p = 0.001, RMSE = 0.43).
Lastly, the best result achieved by multivariate analysis was the determination of the best predictors
of the trabecular bone pattern factor. In this case, speed of sound (r p = −0.51, p = 0.004) and LOW
504
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
Table 2
Correlation coefficients (r) between biomechanical, quantitative ultrasound and dual-energy X ray absorptiometry parameters
at the meta-diaphyseal site. P values are reported in brackets
Ultrasound and
DXA parameters
Speed of sound
Ultrasound peak amplitude
Fast wave amplitude
Centroid of low frequency peak (LOW FREQ)
Width of LOW FREQ at 80% of amplitude (FW80M)
DXA Bone Mineral Density
Maximum load at
failure (FMAX )
0.83 (0.0001)
0.54 (0.004)
0.20 (n.s.)
0.49 (0.01)
0.30 (n.s.)
0.89 (0.0001)
Biomechanical parameters
Normalised FMAX
Applied maximum
(FMAX /mm)
stress (σMAX )
0.75 (0.0001)
−0.12 (n.s.)
0.54 (0.004)
0.18 (n.s.)
0.29 (n.s.)
0.44 (0.02)
0.48 (0.01)
0.13 (n.s.)
0.34 (n.s.)
0.04 (n.s.)
0.78 (0.0001)
−0.07 (n.s.)
FREQ (rp = −0.51, p = 0.004) contributed independently to the model and the global correlation
coefficient obtained was r = 0.92 (p < 0.0001, RMSE = 0.40).
3.3. Measurements at the meta-diaphyseal site (position B)
Analyses of the correlation between ultrasound parameters and measurements of bone mineral density
on the whole phalanx showed that density was most influenced by speed of sound (r = 0.89, p < 0.01),
followed by ultrasound peak amplitude (r = 0.63, p < 0.01) and LOW FREQ (r = 0.55, p < 0.01).
The other ultrasound parameters at the meta-diaphyseal site were not significantly linked with bone
mineral density.
In evaluating correlations between the ultrasound parameters and maximum load obtained from mechanical testing, speed of sound, ultrasound peak amplitude, and LOW FREQ were significantly linked
to maximum load (see Table 2 and Fig. 3A), while fast wave amplitude and FW80M were not. Nevertheless, fast wave amplitude showed a modest, yet significant, correlation with maximum applied stress
(r = 0.44, p < 0.05).
To establish the extent to which bone density influenced the mechanical characteristics of the bone,
we also calculated the correlation coefficient between bone mineral density and mechanical parameters.
The bone mineral density value also correlated strongly with the maximum load (r = 0.89, p < 0.01)
(Fig. 3B), but it was not correlated with maximum stress. Multivariate analysis shows that best ultrasound
parameters predictors of normalized maximum load were: speed of sound (r p = 0.72, p < 0.0001) and
FW80M (rp = 0.28, p = 0.04) in an overall model with an r = 0.80 (p < 0.0001, RMSE = 13.0).
We studied the influence of the geometric characteristics of the bone on the ultrasound parameters
(Table 3). Speed of sound was closely correlated with cortical area and moment of inertia, but not with
the size of the medullary cavity, while fast wave amplitude correlated only with the size of the medullary
cavity and no significant correlation was found with cortical area. The other parameters ultrasound peak
amplitude, LOW FREQ and FW80M were equally linked to cortical area and relative medullary canal
area, though with lower values.
We then evaluated the extent to which the mechanical parameters depended on the geometric characteristics of the bone at the meta-diaphyseal site. The cortical area and the moment of inertia were correlated
with the maximum load (r = 0.79 and r = 0.73, respectively, p < 0.01), while relative medullary canal
area, which represents the size of the medullary cavity, was not correlated with any parameter obtained
by mechanical testing.
The dependence of the bone mineral density on the geometric characteristics of the bone was then
investigated. This gave analogous results: cortical area and moment of inertia were correlated with
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
505
Table 3
Correlation coefficients (r) between morphologic micro-computed tomography, ultrasound and dual-energy X-ray absorptiometry parameters related to cortical bone at the meta-diaphyseal site. P values are reported in brackets
Ultrasound and
DXA parameters
Speed of sound
Ultrasound peak amplitude
Fast wave amplitude
Centroid of low frequency peak (LOW FREQ)Q
Width of LOW FREQ at 80% of amplitude (FW80M)
DXA Bone Mineral Density
Micro-computed tomography parameters
Cortical area Moment of inertia Relative medullary canal area
0.79 (0.0001)
0.79 (0.0001)
−0.22 (n.s.)
0.57 (0.002)
0.37 (n.s.)
−0.60 (0.001)
0.18 (n.s.)
0.02 (n.s.)
−0.52 (0.006)
0.52 (0.007)
0.43 (0.03)
−0.40 (0.04)
0.40 (0.04)
0.43 (n.s.)
−0.43 (0.03)
0.88 (0.0001)
0.80 (0.0001)
0.45 (0.02)
2400
A
m/s
2200
2000
R = +0,83
1800
1600
0
200
400
N
600
800
0,550
1000
B
0,500
R = +0,89
g/cm2
0,450
0,400
0,350
0,300
0,250
0
200
400
N
600
800
1000
Fig. 3. (A) SOS at the meta-diaphyseal site FMAX . Regression line is shown together with correlation coefficient (R). (B) BMD
of the whole bone versus FMAX . Regression line is shown together with correlation coefficient (R).
bone mineral density (r = 0.88 and r = 0.80 respectively, p < 0.01), while the correlation between the
relative medullary canal area and bone mineral density was noticeably lower (r = 0.45, p < 0.05).
We used multivariate analyses to determine which ultrasound parameters were the best estimates of
certain geometric properties of bone at the meta-diaphyseal site. In particular, we observed that of all
the ultrasound parameters considered, the best predictors of the relative medullary canal area were fast
wave amplitude (rp = −0.31, p = 0.05) and ultrasound peak amplitude (r p = −0.43, p = 0.01), with
a global correlation coefficient of r = 0.71 (p = 0.001, RMSE = 5.81). With regard to cortical area, the
best predictors were speed of sound (r p = 0.76, p < 0.0001) and FW80M (rp = 0.34, p = 0.004) with
a global correlation coefficient of r = 0.86 (p < 0.0001, RMSE = 3.53).
506
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
4. Discussion
The study of ultrasound signal propagation through bone tissue has generally focused on evaluating
the speed of transmission and the attenuation of the signal itself. In recent years, many researchers have
been able to observe how the bone tissue changes the characteristics of the ultrasound signal detected
at the receiving probe. How osteoporosis alters the speed and attenuation of the signal itself and also
significantly modifies its frequency and morphological characteristics have often been described. This
reflects the complexity of the ultrasound/bone tissue interaction, which is influenced by the phenomena
of attenuation, reflection, absorption, and back-scattering [29,30].
Analysis of the ultrasound signal and, in particular, the frequency spectrum of the signal, have been
undertaken in some in vivo studies [1,31]. The results show that it is possible to distinguish different bone
properties, independently of bone mineral density, in conditions such as osteoporosis and osteomalacia,
and also to differentiate the bone tissue of a child from that of an elderly person with a fracture when
both have the same ultrasound speed of sound value.
The analytical method we used allowed us to investigate the modalities of the propagation of the
ultrasound signal through the sample measured in relation to determinant characteristics of the sample.
We evaluated a series of ultrasound parameters, assuming that each of these was influenced by particular
characteristics of the bone, and then studied the associations between the ultrasound parameters and the
characteristics of the samples measured by dual-energy X-ray absorptiometry, micro-ultrasound testing,
and mechanical testing.
We limited our analysis to the fastest part of the signal – that is, the part transmitted exclusively through
the bone. We were interested in investigating the different behavior of basic physical characteristics of
the ultrasound wave when traveling through the bone tissue, thus we consider in our analyses basic
parameters of characterization of an ultrasound wave, separately, such as: velocity, amplitude, frequency
content. Other studies show the effects of bone properties on broadband ultrasound attenuation, where
the whole ultrasound signal is processed together to extract relevant parameters [16,23].
We analyzed separately the data measured at the epiphyseal and diaphyseal sites. However, when
ultrasound measurements were compared with micro-computed tomography parameters, we observed
that at both the epiphyseal and metaphyseal levels, speed of sound and ultrasound peak amplitude were
correlated significantly with BV/TV, BS/BV, Tb.Th (epiphysis), and cortical area (metaphysis). All
these parameters describe areas of mineralized bone, as observed by other authors [17,19,20,23]. On
the other hand, fast wave amplitude and FW80M were mostly associated with BS/TV, Tb.Sp, H1, H2,
H3 (epiphysis), and relative medullary canal area (metaphysis), describing empty areas such as the
inter-trabecular spaces and the medullary canal.
It is interesting to note that the ultrasound parameters associated with bone mineral density, such as
speed of sound at the epiphysis, are those significantly correlated with the micro-computed tomography
parameters describing the mineralized bone volumes. These micro-computed tomography parameters
are, in turn, related to bone mineral density. Similarly, we observed that the bone mineral density of the
whole phalanx is highly significantly correlated with the speed of sound (r = 0.89, p < 0.01) and the
ultrasound peak amplitude (r = 0.63, p < 0.01) at the meta-diaphyseal site, and all these parameters
show a close association with the cortical area of the phalanx.
It is also interesting to observe that bone mineral density, speed of sound, and ultrasound peak
amplitude, linked with the mineralized volumes, are in no way associated with the parameters describing
the empty volumes inside the bone. The only exception is the significant correlation of bone mineral
density and ultrasound peak amplitude with Tb.Sp. Tb.Sp is, however, a parameter dependent on Tb.Th
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
507
(in the same volume of interest, as the one increases the other is observed to diminish) and the relative
medullary canal area. There is instead a high and significant negative correlation between LOW FREQ –
that is, the centroid of the peak representing the frequency content of the signal – and the parameters
describing the empty volumes. We had already observed that when a trabecular bone sample was
demineralized by EDTA, the frequency peak of the signal diminished as the duration of decalcification
increased [22]. The present results agree with these observations: the demineralization process performed
by EDTA gradually removes the trabeculae, resulting in an increase of the intra-trabecular spaces. As
these empty volumes increase, the frequency of the centroid of the peak (LOW FREQ) diminishes.
It must be noted that this close dependency of the frequency of the signal on the size characteristics
of the empty spaces is underlined by the fact that only the fast part of the signal was considered – that
is, the part transmitted preferentially inside the mineralized structure of bone. When the mineralized
structures are lacking and the porosity of the structure increases, the greatest effect is indeed observed
in the fastest part of the ultrasound impulse, which is seen to alter its morphological characteristics and
frequency content. This observation also agrees with Biot’s theory, which envisages the propagation of
two types of ultrasound waves: one (the faster) travelling through the mineralized material and the other
(the slower) travelling through the medullary inter-trabecular structure [32].
The multivariate analysis clearly showed that in describing micro-computed tomography parameters
such as the structure model index and trabecular bone pattern factor or relative medullary canal area,
ultrasound parameters associated with the mineralized spaces (speed of sound and ultrasound peak
amplitude) are clearly independent of those associated with the intra-trabecular and medullary spaces
(LOW FREQ, fast wave amplitude and FW80M). Chaffai et al. concluded that no significant independent
association could be found between microstructure and quantitative ultrasound parameters in human
cancellous bone [33]. These workers used bone specimens from the human calcaneus and measured
them with focused broadband transducers with a center frequency of 500 kHz. This was a much lower
than the 1.25 MHz frequency used by us. Furthermore, they investigated the relationships between
microstructure and speed of sound, broadband ultrasound attenuation, and broadband ultrasound backscatter. In their case, all the parameters were calculated on the whole ultrasound pulse received after
transmission or reflection. We, however, decided to consider only the fastest part of the ultrasound signal
and this is probably the main reason for such different results.
Furthermore, the analysis performed on the parameters obtained by the mechanical tests also led us
to conclude that there is a close interdependence between speed of sound, bone mineral density, cortical
area, and moment of inertia, and that all these parameters contribute equally to determining the maximum
breaking load. These parameters provide a similar type of information, which is linked to the quantity
of mineralized matrix in relation to the total bone volume and its geometric distribution.
Unlike the above parameters, fast wave amplitude is correlated with the relative medullary canal area.
This relationship was originally observed by Barkmann et al. [14]. They showed how the area below
the two fastest peaks of the ultrasound signal is linked to the relative medullary canal area, whether in
vivo or in vitro or in mathematical simulation. The area below the two fastest peaks is easily retraceable
to the amplitude of the fast peak of the ultrasound signal. Barkmann et al. also observed that speed
of sound was linked to the cortical area without being dependent on the relative medullary canal area,
and, similarly, how the area below the two fastest peaks was significantly associated with the relative
medullary canal area without being dependent on the cortical area.
Fast wave amplitude and applied maximum stress, the maximum tension applicable to a material of
cylindrical shape, are also correlated. Fast wave amplitude is the only ultrasound parameter that is
significantly correlated with applied maximum stress. This finding shows a dependence on the intrinsic
508
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
mechanical characteristics of the bone material, and is linked to its elastic properties. We previously
found a relationship between fast wave amplitude and the elastic properties of bone independent of
density [22]. In that case, the calculated density was the apparent density and the elastic properties
were assessed by mechanical compression tests on cylindrical trabecular specimens extracted from pigs.
Incidentally, similar observations were made in the analysis of this ultrasound parameter.
The main limitation of our present study concerns the choice of a peripheral site for the whole
instrumental evaluation, while the main current interest in osteoporosis is the evaluation of bone fragility
at the hip. We realize that the phalanx is not typically associated with osteoporotic fractures, but
osteoporosis assessment was not the aim of the study. Secondly, the low number of specimens available
and in good condition, especially at the epiphyseal site is undoubtedly a limitation to the power of the
obtained results. Furthermore when matching data of QUS and (CT images at metaphyseal level we
considered the site of QUS measurement the one corresponding to the ultrasound wave emitted in the
central part of the transducers and we do not consider the whole area interested by ultrasound beam
propagation. The type of mechanical testing was chosen because it was the most readily available
method, bearing in mind that torsion testing would be another possibility. Another limitation was the
lack of comprehensive information on the health status of the cadavers from which we extracted the
specimens. Studies on the characteristics of the ultrasound signal in large populations may elucidate the
clinical relevance of our observations.
We conclude that in this cadaver study some ultrasound parameters were selectively associated with
mineralized volumes and medullary inter-trabecular spaces, independently of the site of measurement
(being mainly trabecular or cortical). Nevertheless, speed of sound was always associated with bone
mineral density, and both of these parameters were also associated with mineralized bone volumes and
mechanical parameters of breakage load. However, applied maximum stress, related only to the intrinsic
elastic properties of the material, was associated with only one ultrasound parameter and showed no
relationship with bone mineral density assessed by dual-energy X-ray absorptiometry or structural microcomputed tomography parameters. This means that some ultrasound parameters provided information on
bone properties that were different from that obtained by dual-energy X-ray absorptiometry assessment.
Furthermore, quantitative ultrasound can provide information on structural properties of bone that is
richer and more exhaustive than those provided by BMD alone. This finding explains the potential of
quantitative ultrasound in a complete non-invasive investigation of bone tissue. Nevertheless, prospective
studies in vivo are warranted to find out how predictive these new ultrasound parameters will be in trials
with human patients.
References
[1]
[2]
[3]
[4]
[5]
C. Wüster, C. Albanese, D. de Aloysio, F. Duboeuf, M. Gambacciani, S. Gonnelli, C.C. Glüer, D. Hans, J. Joly,
J.Y. Reginster, F. de Terlizzi, R. Cadossi, et al., Phalangeal osteosonogrammetry study (PhOS): Age related changes,
diagnostic sensitivity and discrimination power, J Bone Miner Res 15 (2000), 1603–1614.
M. Mauloni, V. Ventura, D. de Aloysio, P.G. Crosignani, A.R. Genazzani, D. Marchesoni, C. Nappi, S. Schonauer, G.B.
Serra and T. Wierdis, Multicenter Italian study on the bone mass ultrasonographic evaluation at the phalanges in the
climacteric (VUMOF-CLIFE), It J Miner Electrol Metab 14 (2000), 23–26.
H. Rico, F. Aguado, I. Arribas, E.R. Hernandez, L.F. Villa, C. Seco and J.J. Gervas, Behaviour of phalangeal bone
ultrasound in normal women with relation to gonadal status and body mass index, Osteoporos Int 12 (2001), 450–455.
F.E. Alenfeld, C. Wüster, C. Funck, J.F.S. Pereira-Lima, T. Fritz, P.J. Meeder and R. Ziegler, Ultrasound measurements
at the proximal phalanxes in healthy women and patients with hip fractures, Osteoporosis Int 8 (1998), 393–398.
G. Guglielmi, C.F. Njeh, F. de Terlizzi, D.A. De Serio, A. Scillitani, M. Cammisa, B. Fan, Y. Lu and H.K. Genant,
Phalangeal quantitative ultrasound, phalangeal morphometric variables, and vertebral fracture discrimination, Calcif
Tissue Int 72 (2003), 469–477.
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
509
F. Hartl, A. Tyndall, M. Kraenzlin, C. Bachmeier, C. Guckel, U. Senn, D. Hans and R. Theiler, Discriminatory ability
of quantitative ultrasound parameters and bone mineral density in a population-based sample of postmenopausal women
with vertebral fractures: Result of the Basel Osteoporosis Study, J Bone Miner Res 17 (2002), 321–330.
C.C. Gluer, R. Eastell, D. Reid, D. Felsenberg, C. Roux, R. Barkmann, W. Timm, T. Blenk, G. Armbrecht, A. Stewart,
J. Clowes, F.E. Thomasius and S. Kolta, Association of five quantitative ultrasound devices and bone densitometry
with osteoporotic vertebral fractures in a population-based sample: The OPUS Study, J Bone Miner Res 19(5) (2004),
782–793.
B. Drozdzowska, W. Pluskiewicz and F. de Terlizzi, The usefulness of quantitative ultrasound at the hand phalanges in
the detection of the different types of nontraumatic fractures, Ultrasound in Med & Biol 29 (2003), 1545–1550.
U. Lange, U.Illgner, J. Teichmann and H. Schleenbecker, Skeletal benefit after one year of risedronate therapy in patients
with rheumatoid arthritis and glucocorticoid-induced osteoporosis: A prospective study, Int J Clin Pharm Res XXIV
(2004), 33–38.
M. Mauloni, L.C. Rovati, R. Cadossi, F. de Terlizzi, V. Ventura and D. de Aloysio, Monitoring bone effect of transdermal
hormone replacement therapy by ultrasound investigation at the phalanx. a four year follow up study, Menopause 7
(2000), 402–412.
M. Kleerekoper, D.A. Nelson, M.J. Flynn, A.S. Pawluszka, G. Jacobsen and E.L. Peterson, Comparison of radiologic
absorptiometry with dual energy X-ray absorptiometry and quantitative computed tomography in normal older white and
black women, J Bone Miner Res 9 (1994), 1745–1750.
M.E. Mussolino, A.C. Looker, J.H. Madans, D. Edelstein, R.E. Walker, E. Lydick, R.S. Epstein and J. Yates, Phalangeal
bone density and hip fracture risk, Arch Intern Med 157 (1997), 433–438.
J.A. Buckwalter, Bone Biology, J Bone Joint Surg A77(8) (1995), 1256–1289.
R. Barkmann, S. Lusse, B. Stampa, S. Sakata, M. Heller and C.C. Gluer, Assessment of the geometry of human finger
phalanges using quantitative ultrasound in vivo, Osteoporos Int 11 (2000), 745–755.
R. Cadossi, F. de Terlizzi, V. Canè, M. Fini and C. Wuster, Assessment of bone architecture with ultrasonometry:
Experimental and clinical experience, Horm Res 54(1) (2000), 9–18.
S. Sakata, R. Barkmann, E.M. LochMüller, M. Heller and C.C. Gluer, Assessing bone status beyond bone mineral
density: Evaluation of bone geometry and porosity by quantitative ultrasound of human finger phalanges, J Bone Miner
Res 19 (2004), 924–930.
P.H.F. Nicholson, R. Müller, G. Lowet, X.G. Cheng, T. Hildebrand, P. Rüegsegger, G. Van Der Perre, J. Dequeker and
S. Boonen, Do quantitative ultrasound measurements reflect bone structure independently of density in human vertebral
cancellous bone? Bone 3 (1998), 425–431.
R. Cadossi and V. Canè, Pathways of transmission of ultrasound energy through the distal metaphysis of the second
phalanx of pigs: An in vitro study, Osteoporosis Int 6 (1996), 196–206.
D. Hans, C. Wu, C.F. Njeh, S. Zhao, P. Augat, D. Newitt, T. Link, Y. Lu, S. Majumdar and H.K. Genant, Ultrasound
velocity of trabecular cubes reflects mainly bone density and elasticity, Calcif Tissue Int 64 (1999), 18–23.
C. Wu, D. Hans, Y. He, B. Fan, C.F. Njeh, P. Augat, J. Richards and H.K. Genant, Prediction of bone strength of distal
forearm using radius bone mineral density and phalangeal speed of sound, Bone 26 (2000), 529–533.
R. Hodgskinson, C.F. Njeh, J.D. Currey and C.M. Langton, The ability of ultrasound velocity to predict the stiffness of
cancellous bone in vitro, Bone 21 (1997), 183–190.
F. de Terlizzi, S. Battista, F. Cavani, V. Canè and R. Cadossi, Influence of bone tissue density and elasticity on ultrasound
propagation: An in vitro study, J Bone Min Res 15 (2000), 2458–2466.
F. Cavani, M. Fini, F. de Terlizzi, M. Cadossi, L. Ciminelli, S. Ortolani, R. Cherubini, D. De Aloysio, G. Giavaresi, R.
Cadossi and V. Canè, Effect of trabecular orientation on mechanical resistance and ultrasound propagation in specimens
of equine vertebrae, Ultrasound in Med & Biol 29 (2003), 1777–1785.
P. Rüegsegger, B. Koller and R. Müller, A microtomographic system for the nondestructive evaluation of bone architecture,
Calcif Tissue Int 58 (1996), 24–29.
T. Hildebrand, A. Laib, R. Müller, J. Dequeker and P. Rüegsegger, Direct 3-D morphometric analysis of human cancellous
bone: Microstructural data from spine, femur, iliac crest, and calcaneus, J. Bone Miner Res 14 (1999), 1167–1174.
T.P. Harrigan and R.W. Mann, Characterisation of microstructural anisotropy in orthotropic materials using a second rank
tensor, J Mater Sci 19 (1984), 761–767.
T. Hildebrand and P. Rüegsegger, Quantification of bone microarchitecture with the structure model index, Compar
Method Biomech Biomed Eng 1 (1997), 15–23.
M. Hahn, M. Vogel, M. Pompesius-Kempa and G. Delling, Trabecular bone pattern factor – a new parameter for simple
quantification of bone microarchitecture, Bone 13 (1992), 327–330.
P.H.F. Nicholson, R. Strelitzki, R.O. Cleveland and M.L. Bouxsein, Scattering of ultrasound in cancellous bone: Predictions from a theoretical model, J Biomechanics 33 (2000), 503–506.
C.F. Njeh, D. Hans, T. Fuerst, C.C. Gluer and H.K. Genant, Quantitative Ultrasound: Assessment of Osteoporosis and
Bone Status, London: Martin Duniz, 1999, 391–399.
510
[31]
C. Wüster et al. / Phalangeal BMD, structure and ultrasound properties
G. Luisetto, V. Camozzi and F. de Terlizzi, Use of quantitative ultrasonography in differentiating osteomalacia from
osteoporosis: Preliminary study, J Ultrasound Med 19 (2000), 251–256.
[32] M.L. McKelvie and S.B. Palmer, The interaction of ultrasound with cancellous bone, Phys Med Biol 36(10) (1991),
1331–1340.
[33] S. Chaffai, F. Peyrin, S. Nuzzo, R. Porcher, G. Berger and P. Laugier, Ultrasonic characterization of human cancellous
bone using transmission and backscatter measurements: Relationships to density and microstructure, Bone 30 (2000),
229–237.