RT Journal Article SR Electronic T1 Phantom-based evaluation of computed tomography parameters: Understanding the differences in automated emphysema scoring JF European Respiratory Journal JO Eur Respir J FD European Respiratory Society SP P269 VO 40 IS Suppl 56 A1 Johan Coolen A1 Frederik De Keyzer A1 Walter De Wever A1 Els Wauters A1 Wim Janssens A1 Marc Decramer A1 Johny Verschakelen YR 2012 UL http://erj.ersjournals.com/content/40/Suppl_56/P269.abstract AB PURPOSE:Automated lung emphysema (E) measurements vary strongly between examinations.We examined the effect of CT scanners, acquisition parameters, kernels and windowing on the software-based E scoring (S) in a phantom.MATERIAL AND METHODS:A human preserved torso in epoxy was used as phantom and was scanned on 5 different scanners using various settings of the following parameters: KV, mAs, care dose, slice/increment, window and kernel. For each of these data sets, the E was evaluated. The ES was performed using both 2D and 3D software. A multiple linear regression analysis (LRA) was used to evaluate the importance of each examined parameter.RESULTS:Measured E values ranged between 17.0% and 70.5% for 3D software evaluation. A similar range could be seen at 2D evaluation, ranging between 13.7% and 66.8%. The used kernel had the strongest impact on the measured ES, and a strong effect could also be seen for slice/increment, mAs, window and KV. However, the scanner-dependent parameters and the usage of the caredosis option proved to have only a minor impact on the measurement of E. The LRA found a very strong correlation between the measured values, and the estimated values based on the optimal regression formulas, with R² values of 0.828 and 0.772 for the right and left lungs on 3D evaluation, and 0.872 and 0.851, resp., on 2D evaluation.CONCLUSION:A fixed phantom allows assessing the influence of different scanners, acquisition parameters and evaluation techniques on the software-based ES. The current dataset indicates that scan parameters and the used kernels have the strongest effect, and that the induced differences can be estimated using multiple LRA.