A Micro-computed Tomography Database and Reference Implementation for Parallel Computations of Trabecular Thickness and Spacing

Abstract

We provide a well-defined accurate brute force implementation without approximations as a reference to support future development of correct and fast algorithms to measure the trabecular thickness and spacing. Using artificial ellipsoid examples, the systematic error is shown to be related to the voxelization, which can be reduced by upsampling. Furthermore, we collected a database of 40 three-dimensional micro-computed tomography images of trabecular bone regions of a sheep femur. With the coverage of a broad range of trabecular thickness and spacing, the database is suitable for assessing the correctness of algorithms for the measurements of trabecular thickness and spacing. The bone data is shared publicly in the online repository ‘Figshare’ and the brute force code is freely available in Code Ocean. Scientists are encouraged to implement faster algorithms, e.g., graphic processing unit-accelerated, with accurate or controlled approximate behaviour to compute trabecular thickness and spacing. Especially, the provided data can be re-employed together with the reference algorithm to evaluate their accuracies.

Keywords

• trabecular thickness
• trabecular spacing
• threedimensional
• volume-based thickness
• brute force