Three-dimensional volume-rendering computed tomography for measuring humeral version

Abstract

Humeral version is highly variable in human beings. Accurate assessment of humeral version may allow for more anatomic reconstruction at shoulder arthroplasty. Two-dimensional (2D) computed tomography (CT) has been used to measure humeral version but has limitations of poor interobserver reproducibility and strict dependence on arm positioning during image acquisition. This study evaluated a new technique, 3-dimensional (3D) volume rendering, for measuring humeral version. Eight dried human humerus specimens were included in the study. Gold standard measurements of humeral version were obtained by use of metallic beads and fluoroscopy. The specimens were then scanned at CT in 2 different positions, 1 neutral to the table and 1 angled at 20 degrees . The image data sets were used to measure humeral version in each bone with both the standard 2D technique and the new 3D technique. Measurements were performed by 3 readers at 2 different time points. Readers were blinded to the gold standard results and each others' measurements. For all readers, 3D measurements averaged within 4.3 degrees of the gold standard. For 2 of the 3 readers, 3D measurements were more accurate than 2D measurements. For all 3 readers, intraobserver variability was better with the 3D technique. For all reader pairs, interobserver variability was better with the 3D technique. This study shows a 3D volume-rendering CT technique to measure humeral version accurately and consistently that is independent of patient positioning.