Although there have been a few reports regarding the effects of IR on CT-based quantitative analysis of emphysema,14–17 it remains unclear whether IR can provide truly “better” measurements than conventional FBP methods. In this study, by using three different dose settings, we demonstrated that there were increases in LAA% or decreases in 15th percentile results due to artifacts at lower tube current settings when using the FBP method and these errors were clearly reduced or controlled when using IR (AIDR3D). Better stabilized and more consistent measurements at different dose settings or for patients with different body habitus may enable researchers to select lower radiation dose settings and use automatic exposure control in future COPD studies. Although it remains difficult to explain why AIDR3D can stabilize measurements, the following explanation is plausible for understanding this phenomenon. It is known that when using conventional FBP methods, sharper reconstruction kernels result in higher LAA% or lower 15th percentile values as compared to using standard kernels.5,6
These sharp kernels, which are frequently applied to chest CT for edge-enhancement and better acutance of small pulmonary structures, involve much greater image noise than do standard kernels.5,13 When using these sharp kernels, artifact density dispersion (density heterogeneity) that is caused by added image noise directly causes an increase in LAA% and a decrease in the 15th percentile, which results in overestimating emphysema severity.5 This relationship regarding image noise differences between sharp and standard kernels can be expanded to that between FBP and IR. Because IR methods greatly reduce image noise, LAA% and 15th percentile are less affected by image noise on scans when using IR, as compared to those when using FBP. Further, even when a significant amount of image noise is anticipated, such as when scanning at a lower tube current or scanning larger patients, IR still powerfully reduces image noise (Figure 2). For these situations, conventional FBP gradually loses control over image noise and, thus, emphysema overestimation caused by greater density heterogeneity appears at lower tube current settings or for larger patients, similar to when using sharper kernels. However, the increase in image noise for these situations is much less when using IR, which probably provides for better stabilized, more consistent measurements.