Effect of iterative reconstruction techniques on the volume measurement accuracy of pulmonary ground-glass density nodule on low-dose CT / 中华放射学杂志

ABSTRACT

Objective To evaluate the potential of three commercial CT iterative reconstruction methods (ASIR,MBIR and iDose4) for the image quality and accuracy of volumetric measurement of ground-glass opacity (GGO) nodules on low-dose chest computed tomography (CT) using a phantom study.Methods A phantom containing GGO nodules was examined in volumetric mode on 2 multi-detector CT scanners at low dose.CT images were reconstructed with adaptive statistical iterative reconstruction (30％ ASIR),model-based iterative reconstruction (MBIR),iDose4 L3,L6 reconstruction algorithm.Then two radiologists measured CT value of the region of interests (ROI) on reconstruction images,and calculated average noise and contrast to noise ratio (CNR) of reconstruction images.And they applied the lung nodules evaluation software to measure the volume of GGO nodules on the reconstruction images,and calculated the absolute percentage error (APE) of nodule volume measurement.The difference of noise and CNR at different reconstructions was compared with ANOVA test,and Kruskal-Wallis H ranking test was used to the difference of GGO measurements at different reconstructions.Results The average noises of CT image reconstructed with the standard FBP algorithm,30％ ASIR algorithm,MBIR algorithm were 40.38 ± 4.49,30.06 ± 20.17,10.55 ± 1.56,and the CNRs were 5.00 ± 2.36,8.43 ± 3.89,20.27 ± 7.35,respectively.Compared to CT image reconstructed with the standard FBP algorithm and 30％ ASIR algorithm,CT image noise reconstructed with MBIR algorithm decreased by 73.9％ and 64.9％,while the CNR increased by 75.3％ and 58.4％ with effective dose of 0.36 mSv in HDCT,respectively(F =13.053,9.459,P ＜0.01).When effective dose was equal to 0.64 mSv in iCT,the image noise reconstructed with iDose L6 algorithms decreased by 36.5％ compared with the standard FBP algorithm,and CNR increased by 78.4％.The reconstruction algorithm,NI and tube current were no significant difference in lung nodule volume measurement of absolute percentage error(P ＞ 0.05).Conclusions Three kinds of iterative reconstruction algorithms are able to reduce the image noise in low-dose CT and improve the image quality.