Establishing the precision of regional and voxel-wise bone mineral density measurements at the knee using dual-energy...

Dual-energy quantitative computed tomography (DE-QCT) enables the quantitative measurement of bone mineral density (BMD), but the in vivo precision of these measurements remains largely unexplored. Voxel-based morphometry (VBM) provides a framework for assessing BMD precision at the voxel level, offering spatial detail that is not available using conventional region-based methods. This study aimed to evaluate the absolute and relative precision of DE-QCT-based BMD measurements at regional and voxel levels for future studies of knee osteoarthritis.

Thirty healthy participants (20 female; mean age = 35.7 ± 12.5 years) underwent two same-day bilateral DE-QCT knee scans with repositioning between scans. Scans were calibrated to provide quantitative measures of BMD, and the femur, tibia, and patella were segmented for analysis. Regional precision was assessed by comparing scan-rescan mean BMD values using root-mean-square standard deviation (SD_RMS) and root-mean-square percent coefficient of variation (CV_RMS), with least significant change (LSC) calculated for both metrics. Voxel-wise precision was evaluated using the same metrics following spatial normalization and Gaussian smoothing with 5 mm and 8 mm full-width-at-half-maximum (FWHM) kernels.

Regional precision was high, with pooled precision estimates across the three bones yielding low LSC values, ranging from 7.08-9.97 mg HA/cm³ (LSC SD_RMS) and ≤ 3.03% (LSC CV_RMS). At the voxel level, LSC SD_RMS was generally low and relatively uniform throughout each bone, with mean values ranging 6.02-6.58 mg HA/cm³ at 5 mm FWHM smoothing and 4.83-5.40 mg HA/cm³ at 8 mm FWHM smoothing across the three bones. Voxel-wise CV_RMS exhibited greater spatial variability, with the highest values observed in the femoral and tibial metaphyses.

Using DE-QCT, we provided spatial maps of reproducibility for voxel-wise BMD measurements in the distal femur, proximal tibia, and patella. At the voxel level, SD_RMS provides a more stable and interpretable precision metric than CV_RMS and is recommended for voxel-based comparisons.