Three-dimensional biplanar radiography as a new means of accessing femoral version: a comparitive study of EOS three-dimensional radiography versus computed tomography.

Abstract

To validate femoral version measurements made from biplanar radiography (BR), three-dimensional (3D) reconstructions (EOS imaging, France) were made in differing rotational positions against the gold standard of computed tomography (CT). Two cadaveric femurs were scanned with CT and BR in five different femoral versions creating ten total phantoms. The native version was modified by rotating through a mid-diaphyseal hinge twice into increasing anteversion and twice into increased retroversion. For each biplanar scan, the phantom itself was rotated -10, -5, 0, +5 and +10°. Three-dimensional CT reconstructions were designated the true value for femoral version. Two independent observers measured the femoral version on CT axial slices and BR 3D reconstructions twice. The mean error (upper bound of the 95% confidence interval), inter- and intraobserver reliability, and the error compared to the true version were determined for both imaging techniques. Interobserver intraclass correlation for CT axial images ranged from 0.981 to 0.991, and the intraobserver intraclass correlation ranged from 0.994 to 0.996. For the BR 3D reconstructions these values ranged from 0.983 to 0.998 and 0.982 to 0.998, respectively. For the CT measurements the upper bound of error from the true value was 5.4-7.5°, whereas for BR 3D reconstructions it was 4.0-10.1°. There was no statistical difference in the mean error from the true values for any of the measurements done with axial CT or BR 3D reconstructions. BR 3D reconstructions accurately and reliably provide clinical data on femoral version compared to CT even with rotation of the patient of up to 10° from neutral.