Low agreement of visual rating for detailed quantification of pulmonary emphysema in whole-lung CT.

BACKGROUND: Multidetector spiral computed tomography (CT) has opened the possibility of quantitative evaluation of emphysema extent in the whole lung. Visual assessment can be used for such a purpose, but its reproducibility has not been established. PURPOSE: To assess agreement of detailed assessment of pulmonary emphysema on whole-lung CT using a visual scale. MATERIAL AND METHODS: Thirty patients with chronic obstructive pulmonary disease underwent whole-lung inspiratory CT. Four chest radiologists rated the same 22 ± 2 thin sections using a visual scale which defines a range of emphysema extent between 0 and 100. Two of them repeated the rating two months later. Inter- and intra-operator agreement was evaluated with the Bland and Altman method. In addition, the percentage of emphysema at -950 Hounsfield units in the whole lung was determined using fully automated commercially available software for 3D densitometry. RESULTS: In three of six operator pairs and in one of two intra-operator pairs the Kendall τ test showed a significant correlation between the difference and the average magnitude of visual scores. Among different operators the half-width of 95% limits of agreement (95% LoA) was wide ranging between a score of 14.2-27.7 for an average visual score of 20 and between 18.5-36.8 for an average visual score of 80. Within the same operator the half-width of 95% LoA ranged between a score of 10.9-21.0 for an average visual score of 20 and between 25.1-30.1 for an average visual score of 80. The visual scores of the four radiologists were correlated with the results of densitometry (P < 0.001; r = 0.65-0.81). CONCLUSION: The inter- and intra-operator agreement of detailed assessment of emphysema in the whole lung using a visual scale is low and decreases with increasing emphysema extent.