Subject motion during high-resolution peripheral quantitative computed tomography (HR-pQCT) causes image artifacts that affect morphological analysis of bone quality. The aim of our study was to determine effectiveness of techniques for quality control in the presence of motion in vivo including automated and manual approaches. First, repeatability of manual grading was determined within and between laboratories. Given proper training using a standardized scale and training images (provided by the manufacturer), we found that manual grading is suitable for repeatable image quality grading within and across sites (ICC > 0.7). Both a new automated technique providing motion measures based on projection moments, and traditional manual grading (1 = best, 5 = worst) were subsequently used to assess subject data for motion in N = 137 image pairs (scan/re-scan) from the Canadian Multicentre Osteoporosis Study (CaMos) Calgary cohort. High quality image pairs were selected and measurement precision was estimated by calculating the coefficient of variation (CV). Consistent with previous data, density parameters (e.g. total bone mineral density) are more robust than structural (e.g. trabecular number) or finite element parameters (e.g. failure load). To obtain acceptable measurement precision, images should not exceed a manual grading of 3 (on a scale from 1 to 5) or an automatic (εT) grading of 1.2. Automatic and manual grading provide comparable quality control, but the advantage of the automated technique is its ability to provide a motion value at scan time (providing a basis for real time decision regarding re-scan requirements), and the assessment is objective. Notably, automatic motion measurement can be performed retrospectively based on original scan data, and is therefore well suited for multi-center studies as well as any research where objective quality control is paramount.
HR-pQCT; Subject motion artifacts; Morphological analysis; Bone quality parameters; Image quality