Determination of vertebral and femoral trabecular morphology and stiffness using a flat-panel C-arm-based CT approach.

The importance of assessing trabecular architecture together with bone mineral density to determine bone stiffness and fracture risk in osteoporosis has been well established. However, no imaging modalities are available to assess trabecular architecture at clinically relevant sites in the axial skeleton. Recently developed flat-panel CT devices, however, offer resolutions that are potentially good enough to resolve bone architecture at these sites. The goal of the present study was to investigate how accurate trabecular architecture and stiffness can be determined based on images from such a device (XperCT, Philips Healthcare). Ten cadaver human C3 vertebrae, twelve T12 vertebrae and 12 proximal femora were scanned with XperCT while mimicking in-vivo scanning conditions and compared to scans of the same bones with microCT. Standard segmentation and morphology quantification algorithms were applied as well as finite element (FE) simulation based on segmented and gray value images. Results showed that mean trabecular separation (Tb.Sp) and number (Tb.N) can be accurately determined at all sites. The accuracy of other parameters, however, depended on the site. For T12 no other structural parameters could be accurately quantified and no FE-results could be obtained from segmented images. When using gray-level images, however, accurate determination of cancellous bone stiffness was possible. For the C3 vertebrae and proximal femora, mean bone volume fraction (BV/TV), Tb.Sp, Tb.N, and anisotropy (C3 only) could be determined accurately. For Tb.Th, structure model index (SMI, femur only), and anisotropy good correlations were obtained but the values were not determined accurately. FE simulations based on segmented images were accurate for the C3 vertebrae, but severely underestimated bone stiffness for the femur. Here also, this was improved by using the gray value models. In conclusion, XperCT does provide a resolution that is good enough to determine trabecular architecture, but the signal to noise ratio is key to the accuracy of the morphology measurement. When the trabeculae are thick e.g. in the femur or the noise is low, e.g. cervical spine, architecture and stiffness could be determined accurately, but when the trabeculae are thin and the noise is high, e.g. thoracic spine, architecture could not be determined accurately and the connectivity was lost and hence no mechanical properties could be calculated directly.