King's College London

Research portal

Accuracy, Linearity and Precision of Spine QCT vBMD Phantom Measurements for Different Brands of CT Scanner: A Multicentre Study

Research output: Contribution to journalArticlepeer-review

Yingwei Zhao, Kai Li, Yangyang Duanmu, Ling Wang, Xiaoming Xu, Yong Zhang, Jing Tang, Yujing Zhang, Zhenlin Li, Karen Hind, Glen M. Blake, Xiaoguang Cheng

Original languageEnglish
JournalJOURNAL OF CLINICAL DENSITOMETRY
DOIs
Accepted/In press2021

Bibliographical note

Funding Information: This work was supported in part by the Beijing Natural Science Foundation-Haidian Primitive Innovation Joint Fund (grant no. L172019 ) and through a grant from the Beijing Bureau of 215 Program (No . 2013-3-033; 2009-02-03 ). Publisher Copyright: © 2021 The International Society for Clinical Densitometry Copyright: Copyright 2021 Elsevier B.V., All rights reserved.

King's Authors

Abstract

We describe a multicenter study using the European Spine Phantom (ESP) to compare the accuracy, linearity and precision of QCT measurements of spine vBMD between different brands of scanner, different models of the same brand and identical units of the same model. Ten scans of the same ESP with repositioning were performed on forty CT scanners from five manufacturers in different hospitals across China, all calibrated with the Mindways QCT system. The three ESP vertebral bodies simulating low (L1), medium (L2) and high (L3) vBMD and their average (L1-3 vBMD) were compared with phantom values. Linearity was assessed using the standard error of the estimate derived from linear regression. Precision errors were expressed as the standard deviation of the ten measurements on each scanner. Median (IQR) vBMD over all forty CT scanners compared with phantom values were: L1: 52.2 (49.9−56.4) vs 51.0; L2: 104.4 (101.2−108.6) vs 102.2; L3: 201.4 (195.0−204.9) vs 200.4; L1-3: 119.3 (116.6−123.2) vs 117.9 mg/cm3. Statistically significant differences in L1-3 vBMD were found between different brands (p= 0.005) and between different models of the same brand and identical units of the same model (both p< 0.001). Cross-calibration using linear regression gave a good fit for all forty systems with a median standard error of the estimate of 1.7 mg/cm3. The median precision error for L1-3 vBMD was 0.61 mg/cm3. Statistically significant differences in spine vBMD measurements between different scanners reinforce the importance of cross-calibration in multi-center studies. Cross-calibration can be reliably performed using linear regression equations.

View graph of relations

© 2020 King's College London | Strand | London WC2R 2LS | England | United Kingdom | Tel +44 (0)20 7836 5454