CT low attenuation area quantification versus visual scoring for early pulmonary emphysema detection in lung cancer screening
Hailan Liu, Daiwei Han, Marjolein A Heuvelmans, Mario Silva, Jan Willem C Gratama, Yifei Mao, Carlijn van der Aalst, Jaeyoun Yi, Zhaoxiang Ye, Harry J de Koning, Matthijs OudkerkAbstract
Objectives
To compare low-attenuation analysis (LAA) with visual scoring and explore their application in the detection of emphysema in low-dose CT (LDCT) scans.
Methods
A random cohort of 352 CT examinations from health-check participants was selected from a lung cancer screening(LCS) dataset. Scans were evaluated for emphysema by four readers according to latest Fleischner criteria, and emphysema was quantified by LAA with thresholds from -900 HU to -980 HU. Fleiss’ Kappa assessed inter-reader agreement on visual detection and severity classification. Cases with full agreement on emphysema presence were used to compare LAA quantification and visual detection. Spearman correlation test measured their correlation and Cohen’s Kappa evaluated the agreement between LAA%-measured emphysema and visual detection.
Results
Agreement among four observers was found in 266 cases on emphysema presence based on visual evaluation (K = 0.738, 0.737–0.739). Concordance between emphysema presence using a LAA% cutoff of 6% and binary visual classification was 61.3% (163/266, Cohen’s Kappa:0.173. Spearman analysis showed a correlation between LAA% and visual scores (P < 0.05).
Conclusions
LAA emphysema quantification shows in comparison with visual scoring limited performance for discrimination of normal and emphysematous lung parenchyma in a Low dose CT Lung cancer screening dataset.
Advances in knowledge
Full agreement among four readers regarding visual emphysema classification was achieved in only three-quarters of cases. Limited concordance was found between LAA%-measured emphysema and visual detection.