A technical and clinical evaluation of digital X-ray radiogrammetry

Abstract.

Skeletal assessment by morphometry at peripheral sites (e.g. metacarpal index), although simple to perform and widely available, was limited by poor precision and technical aspects of radiogrammetry. Digital X-ray radiogrammetry (DXR) uses the principles of this long-established method but applies more sophisticated methodology to overcome these problems. The aims of this study were to (a) investigate the effects of radiographic technique on bone mineral density (BMD) measurement by DXR, (b) compare DXR to dual-energy X-ray absorptiometry (DXA) and single-energy X-ray absorptiometry (SXA) and (c) determine the applicability of DXR in identifying individuals who most appropriately might be referred for axial DXA. Different radiographers performing the radiograph do not adversely affect precision. Precision, unstandardised (CV%) and standardised (sCV%), is good with both double (DF)- and single (SF)-sided emulsion radiographic film, but better with SF (CV% 0.92 vs 1.12 DF; SCV% 1.76 vs 2.93 DF). Repeat analysis precision was determined on SF (CV% 0.24, sCV% 0.55). A significant (p<0.001), systematic difference was found between BMD measured from DF and SF (mean difference 0.017 g/cm²). The overall percentage difference between the methods was 2.98% (range 0.18–5.78%). Correlations between DXR BMD and DXA were moderately good (r=0.56–0.77, p<0.001); with SXA of the forearm they were excellent (r=0.91, p<0.001). The sensitivity and specificity of DXR for detecting women with osteopaenia or osteoporosis (DXA T-score less than −1; World Health Organisation) was determined at the spine [area under curve (AUC)=0.82, standard error (SE)=0.04], femoral neck (AUC=0.84, SE=0.04) and total hip (AUC=0.84, SE=0.04). Based on femoral neck BMD for detection of osteopaenia, a DXR T-score threshold of −1.05 would be appropriate for detection of patients who might benefit most from axial DXA measurements. The DXR is quick and simple to use, having potential for application in a variety of settings as analysis can be performed in a central unit, with radiographs taken in sites over a wide geographical area. Retrospective analysis may also be performed, e.g. on radiographs taken to monitor rheumatoid arthritis. The technique may also provide a simple, widely available and relatively inexpensive method to assess patients at risk of osteopaenia or osteoporosis, and who most appropriately could be referred for axial DXA. This may be particularly relevant in those who suffer low-trauma fractures and attend accident and emergency or fracture clinics, where investigation for osteoporosis is often overlooked.

Appendix: Standardisation of BMD

The BMD measurements were measured on both the Hologic QDR 4500 (n=154) and the Lunar DPX-L (n=50). In order to combine the BMD measurements and T-scores from these machines, standardisation was necessary. The European Spine Phantom [1] is scanned daily in our department. Twenty measurements were taken over the time period of this study and used to produce standardisation equations for conversion of lumbar spine (L2–L4) and femoral neck BMD:

$$ y = \alpha (1 - e^{ - \beta x} ) $$

(1)

where y is the measured density, x is the standardised density, and α and β are parameters unique to each machine [2]. Solving the equation for x enables measured values to be converted to standardised values. For our scanners, the standardisation equations are:

$$ {\rm Hologic}\;{\rm QDR - 4500: Standardised}\;{\rm density}\;{\rm (g/cm2)} = 6.693{\rm }\log _e \left( {{{6.686} \over {6.686 - y_h }}} \right) $$

(2)

and

$$ {\rm Lunar}\;{\rm DPX - L:}\;{\rm Standardised}\;{\rm density}\;{\rm (g/cm}^{\rm 2} {\rm )} = 7.8869018\;\log _e \left( {{{9.181} \over {9.181 - y_1 }}} \right) $$

(3)

where y _h and y _l are the densities measured on the Hologic and Lunar machines, respectively.

Recalculation of T-scores

To combine T-scores from both machines, the standardised BMD is converted back into one of the manufacturer's units using the inverse of the above equations. In this study the standardised Lunar data was recalculated into Hologic data using the following equation:

$$ y_h = 6.686{\rm }(1 - e^{ - 0.149x} ) $$

(4)

The BMD was then converted to T-scores (NHANES III) using an equation provided by T. Kelly (Hologic Inc., Pers. commun.):

$$ {\rm Lumbar}\;{\rm spine}\;{\rm (L2 - L4)}\;{\rm T - score = }{{\left( {LSBMD{\rm - 1}{\rm .079}} \right)} \over {0.110}} $$

(5)

and

$$ {\rm Femoral}\;{\rm neck}\;{\rm (FN)}\;{\rm T - score = }\left( {{{FNBMD{\rm - 0}{\rm .849}} \over {0.111}}} \right) $$

(6)

Appendix references

The following are references for the Appendix:

• 1. Kalender W, Felsenberg D, Genant H et al. (1995) European Spine Phantom: a tool for standardization and quality control in spinal bone mineral measurements by DXA and QCT. Eur J Radiol 20:83–92

• 2. Pearson J, Dequeker J, Henley M et al. (1995) European semi-anthropomorphic spine phantom for the calibration of bone densitometers: assessment of precision, stability and accuracy. Osteoporos Int 2:174–184