CTXA hip: the effect of partial volume correction on volumetric bone mineral density data for cortical and trabecular bone.

Abstract

Previous studies have reported concerns about the reliability of CTXA hip cortical vBMD measurements generated using partial volume (PV) correction (the "default" analysis, with cortical PV correction). To date, no studies have examined the results of the alternative ("new") analysis (with trabecular PV correction). This study presents in vivo and phantom data comparing the corrected and uncorrected data for cortical and trabecular bone respectively. We used the commercial QCTPro CTXA software to analyze CT scans of 129 elderly Chinese men and women and an anthropomorphic European Proximal Femur phantom (EPFP) and accessed data for two alternative scan analyses using the database dump utility. The CTXA software gives the user two methods of performing the PV correction: (1) a default analysis in which only cortical bone results are corrected; (2) a new analysis in which only trabecular bone results are corrected. Both methods are based on a numerical recalculation of vBMD values without any change in volume of interest (VOI) placement. In vivo, the results of the two analyses for integral bone were the same while cortical and trabecular results were different. PV correction of cortical bone led to a decrease of cortical volume for all four VOIs: total hip (TH), femoral neck (FN), trochanter (TR), and intertrochanter (IT) volumes were reduced on average by 7.8cm3, 0.9cm3, 2.5cm3, and 4.3cm3 respectively. For TR, where cortex was thinnest, average corrected cortical volume was negative (- 0.4± 1.3cm3). Corrected cortical vBMD values were much larger than uncorrected ones for TH, FN, and IT. Scatter plots of corrected cortical vBMD against cortical bone thickness showed that elevated results correlated with thinner cortices. When trabecular bone was corrected for the PV effect, trabecular volumes of TH, FN, TR, and IT were reduced on average by 7.9cm3, 0.8cm3, 2.6cm3, and 4.4cm3 respectively, while vBMD measurements were increased correspondingly. The trabecular volume and vBMD measurements of the two datasets both had highly positive correlations. For the EPFP, the PV-corrected FN data deviated from the nominal phantom value, but was closer for the TR and IT VOIs. Both corrected and uncorrected data overestimated trabecular vBMD, with the corrected results showing greater deviation from nominal values. The default and new CTXA analyses for volumetric data generate different results, both for cortical and trabecular bone. For cortical bone, the uncorrected results are subject to partial volume effects but the correction method of the default analysis overcorrects the effect leading to in part unreasonable results for cortical bone volume and BMD. For trabecular bone, the correction effects are smaller. CTXA volumetric data should be interpreted with caution.