Bone Abstracts

Peripheral quantitative computed tomography (pQCT) is receiving considerable attention in the diagnosis and monitoring of human bone diseases. It is well accepted that lower image resolution compared to micro-computed tomography (micro-CT) affects bone morphometry. With advances in micro-CT evaluation techniques such as sample-specific remodeling simulations or dynamic bone morphometry, there is the potential to also allow the application of such techniques to clinical pQCT scans. Therefore, virtual high-resolution image reconstruction was considered to improve image resolution and with that to allow advanced quantification schemes. We hypothesized that upscaling pQCT images either preserves or enhances bone morphometry.

Accuracy was investigated by downscaling 16 ex vivo human vertebral grayscale scans from 17.4 to 87.0 μm and subsequent upscaling to higher image resolutions (17.4, 34.8, 52.2, and 69.6 μm). The morphometric indices, bone volume fraction, specific bone surface, trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), trabecular number (Tb.N), structural model index, degree of anisotropy, and connectivity density were compared to the lowest and the highest image resolution. Reproducibility was assessed by precision errors of 14 times three repeated cadaveric forearms scanned at 82 μm and virtually reconstructed at 41 μm resolution. Sensitivity was investigated by a clinical study of 100 fractured and 105 non-fractured human forearm pQCT scans.

Regarding accuracy, the scans upscaled from the 87-μm-resolution-images deviated maximum 11.1% (Tb.N, 17.4 μm) and 42.3% (Tb.Sp, 17.4 μm) from the original 17.4-μm-resolution images, indicating that bone morphometry could be preserved but not enhanced. The technique was reproducible (1.96–7.88%) and sensitive to changes as in the clinical study, all indices (except Tb.Th) were significantly worse in the fractured group at 41-μm-resolution (P<0.05). These results agreed with the differences at 82-μm-resolution where all indices showed significant differences (P<0.05). We conclude that virtual high image-resolution reconstruction can be applied to pQCT scans, however, it does not provide more information than the original lower image resolution.