Resolution dependence of the non-metric trabecular structure indices

Non-metric indices of topological features of trabecular bone structure, such as structure model index (SMI), connectivity density (Conn.D), and degree of anisotropy (DA), provide unique information relevant to bone quality. With recent technological advancement, in vivo assessment of these indices may be possible from images acquired using high-resolution imaging techniques such as high-resolution peripheral quantitative computed tomography (HR-pQCT). However, more detailed investigation of the dependence of non-metric indices on spatial resolution is needed to determine their applicability. The purpose of this study was to determine whether these three non-metric indices are affected by the spatial resolution of CT images. First, the SMI, Conn.D, and DA were calculated for trabecular bone specimens with varying plate-like and rod-like structures from resampled muCT images across a range of spatial resolutions and compared to the reference values. To account for differences in size across different species and anatomical sites, the results are reported in normalized resolution units. Next, the impact of resolution on the non-metric indices for cores of human distal tibia trabecular bone from clinical HR-pQCT images was evaluated to determine the applicability of the non-metric indices to in vivo imaging. We found that the non-metric indices of trabecular bone structure were affected by spatial resolution of CT images. Particularly, the SMI deviated from the high-resolution muCT reference value depending on the structure type, whether plate-like or rod-like. Both Conn.D and DA were underestimated in the images obtained at an in vivo resolution. It is not trivial to determine absolute threshold for validity of these non-metric indices without considering a specific study design (e.g. relative resolution, the size of the treatment effect to detect, and specimen type). The results of this study provide an upper bound for the accuracy of the non-metric indices under limited resolution scenarios.