Impact of spatial resolution on the prediction of trabecular architecture parameters

Although the efficacy of various measures for the assessment of trabecular bone architecture has been widely studied, the impact of spatial resolution on the estimation of these measures has remained relatively unexplored. In this study, ten cubes each of human trabecular bone from the femur and vertebral bodies were obtained from nine cadavers (four males and five females), aged 23-67 years (mean 42.3 years), These specimens were serially milled and imaged at a resolution of 40 mu m to produce three-dimensional digitizations from which traditional morphometric and structural anisotropy measures could be computed based on a three-dimensional approach, The cubes were then artificially degraded to an in-plane resolution of 100 mu m and an out-of-plane (slice) resolution of 100-1000 mu m, These resolutions mimicked in vivo resolutions as seen using magnetic resonance (MR) imaging. All images, original and degraded, were individually segmented using a thresholding algorithm, and both the traditional morphometric and structural anisotropy measures were recomputed. The choice of slice direction was varied along the superior-inferior (axial), anterior-posterior (coronal), and medial-lateral (sagittal) directions to minimize the impact of the lower slice resolution on the architectural measures, It was found that traditional morphometric measures such as trabecular spacing and trabecular number showed weak resolution dependency; measures such as trabecular thickness, however, showed strong resolution dependency and required very high resolutions for precise measurement, In the case of the femur specimens, both structural anisotropy as well as the preferred orientation showed a strong resolution dependency. The resolution dependency of these parameters could be minimized for the femur and the vertebral body specimens if the slice direction was taken along the superior-inferior direction. (C) 1998 by Elsevier Science Inc, All rights reserved.