A Validated Model for the Imaging Diagnosis of Cystic Lung Disease

Abstract

Rationale and Objectives: Cystic lung disease may be accurately diagnosed by imaging interpretation of specialist radiologists, without other information. We hypothesized that with minimal training non-specialists could perform similarly to specialist physicians in the diagnosis of cystic lung disease. Methods: 72 cystic lung disease cases and 25 cystic lung disease mimics were obtained from three sources: 1) a prospective acquired diffuse lung disease registry, 2) a retrospective search of medical records and 3) teaching files. Cases were anonymized, randomized and interpreted by 7 diffuse lung disease specialists and 15 non-specialist radiologists and pulmonologists. Clinical information other than age and sex was not provided. Prior to interpretation, non-specialists viewed a short PDF training document explaining cystic lung disease interpretation. Results: Correct first choice diagnosis of 85%-88% may be achieved by high-performing specialist readers and 71%-80% by non-specialists and lower-performing specialists, with mean accuracies in the diagnosis of LAM (91%, p < 0.0001), BHD (93%, p < 0.0001), PLCH (89%, p < 0.0001) and LIP (92%, p < 0.0001). A strategy based on cyst appearance: simple cysts (LAM), peri-septal cysts (BHD), bizarre-shaped cysts (PLCH) and vascular indented cysts (LIP) gave non-specialists accuracies of 90% (p < 0.0001), 94% (p < 0.0001), 92% (p < 0.0001) and 88% (p < 0.0001), respectively, for these diagnoses. Cystic lung abnormalities caused by diseases other than LAM, BHD, PLCH and LIP are rarely accurately diagnosed by imaging alone. Conclusion: With specific but limited training, non-specialist physicians can diagnose cystic lung diseases from CT appearance alone with similar accuracy to specialists, correctly identifying approximately 75% of cases.