Assessing Bone Status Beyond BMD: Evaluation of Bone Geometry and Porosity by Quantitative Ultrasound of Human Finger Phalanges


  • Satoru Sakata,

    1. Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig-Holstein, Kiel, Germany
    Search for more papers by this author
  • Reinhard Barkmann,

    Corresponding author
    1. Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig-Holstein, Kiel, Germany
    • Address reprint requests to: Reinhard Barkmann, PhD, Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig-Holstein, Campus Kiel, Schittenhelmstrausse 12, Kiel D-24105, Germany
    Search for more papers by this author
    • Dr Barkman receives unrestricted grants from Igea srl. Dr Glüer served as a consultant for Igea srl. All other authors have no conflict of interest

  • Eva-Maria Lochmüller,

    1. Universitätsfrauenklinik der Ludwig Maximilians, Universität München, München, Germany
    Search for more papers by this author
  • Martin Heller,

    1. Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig-Holstein, Kiel, Germany
    Search for more papers by this author
  • Claus-Christian Glüer

    1. Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig-Holstein, Kiel, Germany
    Search for more papers by this author


In an in vitro study, we found significant associations between QUS variables and properties and geometrical parameters of the compact bone of human finger phalanges. QUS variables were not only related to BMD but also to other skeletal properties, which explained 70% of the variability of speed of sound.

Introduction: Transverse transmission quantitative ultrasound (QUS) measurements at the finger phalanges are known to be correlated with BMD and to predict osteoporotic fractures. To determine which other skeletal properties are affected by ultrasound, we investigated the impact of density, geometry, and porosity on QUS variables in vitro.

Materials and Methods: Ultrasound variables were correlated with density, porosity, and geometrical characteristics of cortical bone. Additionally, we tested which combinations of geometry and bone properties best predicted the ultrasound results observed. Forty-four proximal phalanges from the middle finger were investigated at their distal metaphysis, similar to the typical clinical measurement procedure. Donor age ranged from 52 to 98 years (15 males and 29 females; mean age, 80.9 ± 9.4 years). QUS variables were measured on the metaphysis of the phalanges using the DBMSonic 1200. Quantitative CT was used for the measurement of BMD, and high-resolution MRI was used for the measurement of porosity and geometrical variables.

Results: Speed of sound (SOS) and the clinically used variable AD-SOS correlated significantly with area, relative area, density, and porosity of the compact bone (R2 = 0.28-0.58, p < 0.0001). Signal amplitude correlated significantly only with relative area of the compact bone and area of the medullary canal (R2 = 0.18-0.20, p < 0.01). The combination of cortical area, density, and porosity improved the determination of SOS to R2 = 0.70, with a residual unexplained variability of 54 m/s (3.2%).

Conclusions: These results clarify the impact of skeletal properties on QUS variables. SOS is affected by cortical area, cortical bone density, and cortical porosity, whereas attenuation only depends on geometry of the medulla. AD-SOS, the variable routinely measured in clinical practice, is primarily affected by cortical area. QUS of the finger phalanges is not only related to BMD but also to other skeletal properties.


THE MEASUREMENT OF bone mass is of central importance for the diagnosis of osteoporosis and in the evaluation of efficacy of various therapeutic regimens designed to reduce fractures associated with this disease. Bone density is the most relevant quantitative parameter for the prediction of osteoporotic fracture risk. However, recent clinical trials have suggested that bone quality is an additional important determinant of skeletal fractures, because changes in vertebral fracture rate correlate only moderately with the magnitude of change in spinal bone mass.(1,2) Considering that fracture risk increases with age and that prevalent osteoporotic fractures represent a risk factor for future fracture independent of BMD, it can be assumed that the measurement of additional properties of bone quality might improve the prediction of fracture risk. The observations have prompted a new interest in the development of methodologies directed at measuring bone quality.

In the past 10 years, quantitative ultrasound (QUS) techniques have been introduced to evaluate skeletal status.(3–5) Several studies have shown that variables of ultrasound measurements of calcaneal trabecular bone measured in transverse transmission mode correlate moderately with BMD measured using DXA on different sites,(6) but strongly when measured on the same site.(7) Ultrasound variables are influenced by bone structure,(5,8,9) although it is difficult to show an independent correlation with histomorphometric measures, in part because structure and BMD are correlated.(7) Although speed of sound (SOS) has been found to be associated with bone structure,(9) the impact of structure remains small when adjusted for BMD.(10)

QUS assessment of bone status of the finger phalanges is an alternative approach and has been evaluated by several researchers.(11–19) The phalanx is easily accessible for measurements using QUS. The bone of the phalanx contains both cortical and trabecular bone and is sensitive to early changes in bone mass.(20,21) At the measurement location chosen in clinical practice, the phalanx of adult subjects can be expected to mostly consist of cortical bone,(17) although no extensive study has been reported regarding this issue. QUS at the phalanges correlates with hand bone density measured by densitometric techniques such as dual photon absorptiometry (DPA), DXA, or computed radiogrammetry(11,22) discriminates between osteoporotic and healthy women(23) and predicts incident fractures.(24) From simulation studies(17) we can expect that QUS variables would not only be affected by BMD as measured by the above-mentioned techniques but also by geometric properties. This was confirmed in vitro in a former study that showed strong correlations between QUS variables and geometrical properties of compact bone.(17) Moreover, DXA-based BMD measures themselves depend on volumetric mineral density and geometry of the bone. In phalanges, BMD depends on thickness and shape of the cortex. This makes it difficult to identify the determinants of QUS variables in vivo.

Ultrasound transmission is complex and depends on a variety of material properties like density, elasticity, and structure. After penetrating a structured medium such as bone, the shape of the ultrasound signal changes substantially, which complicates the interpretation of the results. At first sight, this may be considered to be a disadvantage. However, a variety of variables can be extracted from the ultrasound signal received that might reflect different aspects of skeletal status. To gain further insight into the clinical validity of this “multiparametric” approach, we initiated a study to investigate relationships between QUS variables and BMD and other skeletal properties in vitro.



Forty-four proximal middle phalanges from cadavers, 52-98 years of age (15 males and 29 females; mean age, 80.4 ± 9.9 years), were used. Specimens were obtained from autopsy courses in Munich and stored in formalin.(25,26)



QUS measurements of the phalanges were performed using the DBM Sonic 1200 (Igea, Carpi, Italy). The device was modified for research purposes to allow measurements with the probes immersed in a water bath. To exclude air effects on ultrasound transmission, the specimens were degassed under vacuum (∼13 mbar) for 30 minutes to extract air bubbles and transferred to the water bath without exposure to air. The specimens were positioned between the probes, which also were immersed in this water bath. In clinical use, ultrasound probes are mounted on a caliper, which is manually positioned on the phalanx in patient measurements. A button on the caliper has to be pressed by the operator to store the optimal signal. In the research device, this button was removed and placed distally to prevent ingress of moisture into the button. Apart from storage of the signal over a wider time range, which does not affect the measurements (but allows measurement of additional variables), no further differences existed between the clinical and the research device. For quality control, a phantom was measured on a regular basis just as in clinical routine.

The middle fingers were measured four times in the medio-lateral direction with repositioning. The ultrasound signal passed laterally (radial to ulnar) through the distal metaphysis of the phalanges. In all measurements, the distance between the probes was 20 mm. To avoid detection of ultrasound signals passing alongside the face of the phalanges, the ultrasound beam was collimated at both the sender and receiver side by ultrasound-blocking foam barriers with 8-mm-diameter holes. Specimens were positioned between the probes according to the procedure used in vivo. By slightly sliding and rotating them between the probes, the position with the best trace on the monitor was selected. In this device, ultrasound with a frequency of 1.25 MHz was used. For calculating velocity, the time when the electrical signal, generated by the ultrasound mechanical wave at the receiving probe, reached an amplitude of 2 mV was chosen, and thus the measured velocity depended on the signal amplitude. Therefore, it is reported as “amplitude-dependent speed of sound” (AD-SOS).(15,27) To disentangle the velocity and attenuation, we separately measured speed of sound (SOS) as a velocity parameter independent of the amplitude and the average amplitude of the first oscillation (AMP). SOS was calculated from the time of flight of the very first part of the ultrasound wave, specifically when first exceeding a low trigger level. AMP was calculated from the area under the first two peaks (reflecting the first positive and negative component of the first oscillation) of the signal received divided by their combined duration. Signal processing has been previously described in detail.(17)

High-resolution MRI:

To estimate cortical area and porosity independent of BMD, we selected high-resolution MRI (HRMRI) because, unlike X-ray-based methods, the signal is derived from the marrow and thus is unaffected by bone material properties. HRMRI was performed using a Siemens Vision Magnetom (Siemens AG Medical Solutions, Erlangen, Germany) using a specially developed spin echo sequence, with TE = 13 ms, TR = 65 ms, B = 1.5 T, and an examination time of 15 minutes.(28) Spatial resolution was 152 × 152 × 280 μm. The phalanges were mounted in a customized holder placed in a plastic box filled with 1.0 mM gadolinium-diethyltriamine pentaacetic acid (Gd-DTPA). To avoid artifacts caused by air within the phalanges, the phalanges were degassed for at least 30 minutes before measurement.

The HRMRI slices were positioned in the same location that is measured by QUS, with a central slice at 40% phalanges length below the proximal inter-phalangeal (PIP) joint. Five 1-mm-thick slices were obtained in the region. From HRMRI, we estimated geometrical parameters (cross-sectional area of the cortex [CA], cross-sectional area of the medullary canal [MCA], and relative cross-sectional cortical area [RCA]). RCA was calculated as the ratio of the cross-sectional area of cortex and the whole bone. We also evaluated cortical porosity (CP), which was calculated as the ratio of porosity area and the whole cortical area. Pores are defined as closed areas of bone marrow inside the compact bone. Porosity is the ratio of the integrated areas of all pores divided by the area of the compact bone (including pores). All parameters were analyzed using NIH Image 1.61 (National Institutes of Health, Bethesda, MD, USA). A global threshold was manually chosen offering the best differentiation between bone and marrow on the inner cortical surface. Image resolution was limited, and therefore only larger pores (>300 μm) could be visualized. However, for the group of specimens evaluated, substantial variability in porosity of this magnitude could be observed (Fig. 1).

Figure FIG. 1..

HRMRI cross-sectional images of eight phalanges with variable cortical area, shape, and porosity.

High-resolution quantitative CT:

To assess cortical BMD, we performed high-resolution quantitative CT (HRQCT) measurements. HRQCT was carried out using a Siemens Somatom plus CT (Siemens AG Medical Solutions) scanner. A high-resolution mode was implemented with a slice thickness of 1 mm and a high exposure setting of 330 mA at 120 kVp and matrix size of 256 × 256 pixels. To calibrate BMD, we used two phantoms: one water equivalent and the other incorporating 200 mg/cm3 hydroxyapatite. The average density in the cortical area was calculated from five adjacent slices of 1 mm thickness obtained in the same area as HRMRI selected to match the area measured by finger QUS measurements. Despite the high-resolution mode, the spatial in-plane resolution was insufficient for an accurate depiction of the cortical thickness (Fig. 2). Because of partial volume effects, the measured density is affected by the thickness of the cortex. Using the cortical area accurately determined as CA on HRMRI images, we corrected the measured average density by the ratio of (CA − porous area) divided by the cortical area measured by HRCT. This should largely correct for the partial volume induced decrease of measured density. To reflect the limitation of the approach, the resulting density was denoted as “apparent cortical density” (ACD). All images were analyzed with custom software (NIH Image 1.61). A fixed threshold was used for the segmentation of the cortex.

Figure FIG. 2..

HRQCT cross-sectional images of eight phalanges; these are the same bones as depicted in Fig. 1. Variability in gray level reflects local pores (even if below resolution of the image) and variability in mineralization, modulated by partial volume effects.

Statistical analysis

The degree of correlation between variables was determined by means of linear regression analysis. Best combinations were evaluated using multifactorial stepwise regression. All statistical computations were processed using JMP (SAS Institute, Cary, NC, USA).


Descriptive statistics of the variables measured are summarized in Table 1.

Table Table 1.. Descriptive Statistics of the QUS Variables and Skeletal Properties
original image

Correlation between geometry, density, and porosity

ACD did not correlate with CA or RCA and only slightly with CP (R2 = 0.1, p = 0.04). CP correlated negatively with RCA (R2 = 0.36, p < 0.0001) and CA (R2 = 0.15, p < 0.01).

Impact of geometry on ultrasound variables

SOS correlated significantly with CA (R2 = 0.58, p < 0.0001) and RCA (R2 = 0.41, p < 0.0001) but not MCA (R2 = 0.03). AMP correlated significantly with MCA (R2 = 0.18, p < 0.01) and RCA (R2 = 0.20, p < 0.01) but not with CA (R2 = 0.03). AD-SOS correlated significantly with CA (R2 = 0.50, p < 0.0001) and RCA (R2 = 0.47, p < 0.0001) but not MCA (R2 = 0.07; Table 2). SOS and AD-SOS increased with increasing CA, whereas AMP decreased with increasing MCA. Combining CA and RCA improved the determination of SOS and AD-SOS. The geometrical properties explained 64% of the variability of SOS, 18% of the variability of AMP, and 60% of the variability of AD-SOS.

Table Table 2.. Associations Between Ultrasound Variables SOS, AMP, and AD-SOS, and Geometrical Variables CA, MCA, and RCA
original image

Impact of density and porosity on ultrasound variables

SOS and AD-SOS correlated significantly with CP (R2 = 0.30, p < 0.001; R2 = 0.30, p < 0.001) and ACD (R2 = 0.28, p < 0.001; R2 = 0.24, p < 0.001). There was no correlation between AMP and ACD (R2 = 0.03) or CP (R2 = 0.03; Table 3). Combination of ACD and porosity could not improve the determination of SOS or AD-SOS. Porosity explained 30% of the variability of SOS and AD-SOS.

Table Table 3.. Associations Between Ultrasound Variables SOS, AMP, and AD-SOS and CP and ACD
original image

Determinants of ultrasound variables in combined models

Using step-wise regression, we evaluated which set of bone properties contributes to the determination of the QUS variables. SOS was determined best by a combination of CA, porosity, and ACD (R2 = 0.70, SE of the estimate [SEE] = 54 m/s, p < 0.0001). AD-SOS was determined best by a combination of CA, RCA, and ACD (R2 = 0.65, SEE = 57 m/s, p < 0.0001). No combination could improve the determination of AMP (R2 = 0.18). In summary, 70% of the variability of SOS, 65% of the variability of AD-SOS, and 18% of the variability of AMP could be explained.


Aging is associated with accelerated bone loss and increased bone fragility. For the assessment of osteoporosis, QUS became attractive because it is free of ionizing radiation, is less expensive, and has a potential to assess bone material, strength, and structure, providing information relevant to bone strength beyond BMD.(5,9,29,30) Recently, QUS transverse transmission techniques on the phalanx had been evaluated by several researchers(11–17) because human phalanges are sensitive to early changes in bone mass.(20,21) QUS is generally based on a measurement of the velocity of the ultrasonic wave propagating through the bone segments and on the attenuation of the ultrasound signal. A strong correlation was observed between the velocity of the ultrasonic wave and cortical thickness in phantoms made of Perspex(16) in pig phalanges(15) and human phalanges.(31,32) A more detailed analysis of the ultrasound signal has been performed by our group,(17) showing stronger associations between SOS and cross-sectional area of the compact bone rather than cortical thickness. On the other hand, AMP correlated with the cross-sectional area of the medullary canal. Similar associations could be found in vivo, in vitro, and on phantoms. Our study confirms these results and provides more detailed information about the determinants of QUS variables (e.g., the impact of cortical porosity and density).

At the standard measurement location selected for QUS studies in vivo, the phalanx mostly consists of cortical bone, with some degree of endosteal trabecularization, and rarely, a few trabeculae are present (Fig. 1). Characteristics of the cortex thus should be sufficient to assess relationships with ultrasound variables. The structure of cortical bone can be described by two features: cortical thickness and cortical porosity. Additionally, for ultrasound interaction, the shape of the cortex and the medullary canal needs to be considered. For example, different sizes of the medullary canal at identical levels of cortical thickness and porosity can be expected to result in differences in ultrasound propagation (data not shown). The endosteal bone surface is metabolically active and very involved in bone remodeling. Cortical thinning or the expansion of the inner diameter during aging actually takes place at the endosteal surface, and the trabecularization that is observed on the inner surface of the cortex is an outcome of endosteal resorption.(33,34) On the periosteal surface, bone apposition occurs, predominantly, but not exclusively, in men.(35) Within the cortex, cortical remodeling results in increased porosity caused by expansion of the Haversian canal area.(36) Taken together, the cortical shell of long bones gets thinner and becomes more porous because of aging and disease. Our results show that both of theses processes affect QUS variables; for some of them the effect is independent. Cortical porosity was significantly associated with SOS as well as AD-SOS in our study but not with signal amplitude. The association between CA and SOS could be improved from R2 = 0.56 to R2 = 0.64 by adding porosity as a covariate.

We also investigated the impact of cortical density on QUS variables. Here the combined effect of the level of mineralization and the presence of smaller pores (that could not be resolved by HRMRI because of the limited resolution) within the cortex was estimated by the variable ACD. ACD correlated with SOS and AD-SOS but not with AMP. Adding ACD to the association between cortical area and porosity on one hand and SOS on the other hand led to a significant increase in R2 (0.64-0.70). When measuring cortical density, it has to be considered that compact bone is not homogenous. Apparent bone density is an average over variably mineralized parts of the compact bone and “nonmineralized” parts like the Haversian canals, and therefore, decreases with increased porosity. A decrease in mineralization causes a decrease in ultrasound velocity because of a lower material stiffness.(37) We excluded pores from the calculation of density, thereby separating porosity and density. However, spatial resolution in MRI was 0.156 mm, which only allows the detection of greater pores. Smaller pores, the majority of the Haversian canals, remain out of consideration. Thus, we cannot separate between lower mineralization and larger areas (or increased number) of the canals. A typical confounder in density measurements is the partial volume effect. Although we corrected our density values by using the cortical area assessed from the higher resolved MRI images to correct density values obtained from CT, a potentially remaining error caused by this effect cannot be ruled out. Our study does not have the power to show if any variations in mineralization alone have an impact on the QUS data. However, for ACD, the reported independent associations document that, despite the limited spatial resolution, this cortical variable seems to capture additional cortical properties that independently affect QUS variables. Both effects, increases in porosity and decreases in mineralization, result in lower SOS values, that is, in a diagnostically correct direction, because both impair the strength of the bone.

It is interesting to note the lack of any association between porosity and signal amplitude. A porous structure could be assumed to enhance scattering and hence increase attenuation. This assumption is not supported by our results. Porosity seems to have little or no impact on signal amplitude. However, porosity has a strong effect on SOS and AD-SOS. The strongest association was observed between SOS and a combination of cortical area and apparent cortical density, which might at least in part be caused by the impact of microporosity on ACD. ACD and CA together explain 70% of the variation in SOS. For AD-SOS, 65% of the variability was explained, with SEE = 54 m/s. This is equivalent to a residual error of 77% of 1 population SD (young reference population).

During aging, bone is lost, and it can be expected that all bone-related measures change their values with age. Therefore, it has to be examined if the correlations of different measures of bone properties are causally determined and not only an effect of ageing. However, in our study group, no significant correlation of any of the variables with age could be found, which might be because of the high ages of the subjects. Additional adjustment for age did not affect the findings, that is, significant associations remained significant and nonsignificant ones remained nonsignificant. Therefore, it can be assumed that aging did not cause any bias on the correlations between bone properties and ultrasound results.

Other limitations of our study include the small sample size, limited age range, and the use of formalin fixation, which might affect ultrasound propagation.(38) Nevertheless, the associations between QUS variables and variables of bone geometry were similar to those observed in vivo, on phantoms, and in simulations of ultrasound propagation through cortical long bones.(17) Therefore, it is very likely that our findings regarding the impact of cortical porosity on SOS are realistic, although this should be confirmed in additional studies.

We investigated associations between QUS variables measured with a clinical device and geometry and density variables measured with clinical devices of CT and MRI. Because of the limited resolutions of CT and MRI, not all of the Haversian canals could be detected. Therefore, it would be appropriate to continue these kinds of examinations with higher resolution methods such as μCT. A deeper insight into bone microstructure and material properties, however, requires the application of destructive methods like acoustical microscopy or histology, which were not available for our study. Also, the clinical relevance of our results, specifically the impact of bone geometry, porosity, and density on mechanical properties, still has to be examined using tests of bone strength and comparison with osteoporotic fracture status.

The standard variable evaluated in regular clinical practice is AD-SOS, whereas SOS and AMP measurements are not implemented in the clinical devices. However, because raw data are stored on the computer, these variables can be calculated retrospectively. Another variable available on clinical devices is “bone transmission time” (BTT). BTT depends positively on the amount of bone inside the phalanx, largely independent of soft tissue thickness. BTT would be equal to zero if no bone were present. The variable SOS, as calculated by us, is strongly correlated to BTT,(39) and hence, results for SOS should also be valid for BTT. From our results it can be assumed that SOS and BTT are not only sensitive to endosteal resorption (i.e., cortical thinning) but also to increased porosity. A decrease in porosity caused by therapy might also be detectable. In this context, it is interesting to note that the IGEA device performed well for monitoring response to alendronate treatment. Machado et al.(40) reported a significant increase in BTT in the course of 2 years of treatment. Whether this was because of decreases in porosity or increases in mineralization because of bisphosphonate treatment is unclear, but both aspects are captured by the QUS approach. Further studies are warranted to investigate the potential for monitoring treatment effects and perhaps separating different skeletal changes such as changes in porosity and mineralization.

A very common criticism of QUS methods revolves around the question of what skeletal properties are measured by QUS. Our data shows that 65% of the variability of AD-SOS can be explained by the skeletal and geometric properties investigated. This error includes inaccuracies of the imaging methods. If better imaging methods had been used, the percentage explained would in all likelihood have been even larger. In absolute units, the residual error is 57 m/s or 81% of the population variance reported for a young reference population. How does this compare with accuracy errors of DXA? For comparison, for spinal and femoral DXA, accuracy errors of 5-16% were reported,(41,42) corresponding to 30-90% of the population variance observed for those established techniques, comparable with the values reported here for AD-SOS. For SOS (or BTT), the residual errors can be assumed to be even smaller than for AD-SOS because the coefficient of determination is higher. However, no reference data are available today, and therefore, no direct comparison with DXA can be made. Our findings document that residual errors in the associations between QUS and bone properties are comparable with errors in the association between central DXA values and BMD. In other words, for phalangeal transverse transmission ultrasound, the question of which skeletal properties are measured can be answered with residual errors comparable with central DXA.

In summary, we identified different bone properties that explain a large portion of the variability of QUS variables for transverse transmission measurements of the finger phalanges. This showed that an expanded evaluation of the ultrasound signal might lead to a refined analysis of different bone properties. The calculation of different skeletal properties from QUS parameters might enable a differential diagnosis of different kinds of bone disorders and monitoring of disease-specific bone patterns.


This work was supported by the Japanese Foundation for Age and Health, the Alexander von Humboldt-Stiftung, and through an unrestricted grant by Igea srl.