An anthropomorphic phantom study on the effect of midvertebral slice placement and region-of-interest positioning on the reproducibility of single-energy quantitative CT (QCT) of the spine.

Abstract

The purpose of our study was to develop an anthropomorphic phantom with a 3D external reference system capable of geometrically describing the region of interest (ROI) of single-energy quantitative CT (QCT) scans and to study the reproducibility of ROI placement (volume) and bone mineral density (BMD) after operator-defined and algorithm-supported midvertebral slice (MVS) placement. In three vertebrae (L1-3) of 10 human cadaveric spines placed in a water phantom, MVSs were defined by an operator and an algorithm-supported technique on lateral digital CT radiographs, and QCT scans were performed accordingly. The measurements were repeated once after repositioning the phantom on the CT table. ROIs of the trabecular bone were determined with a standard technique. The percentage of bone volume was calculated for one ROI not covered by the repetition (volume mismatch percent). Reproducibility with algorithm-supported MVS placement was superior to that of operator-defined positioning with regard to volume mismatch (mean +/- SD): 10.6+/-8.4 vs. 7.9+/-5.3%; and mean of paired BMDs (mean of three vertebral bodies): 2.7 vs. 1.5% (p < 0.05). The ROI volume mismatch of repeated QCT scans, which is approximately 10% of ROI volume, can be quantified with an external reference system. Automated placement is superior to the manual technique and should be used in clinical practice.