Inter-reader Agreement and Diagnostic Accuracy of Imaging in Hepatic Cystic Lesions

Purpose: Pre-operative imaging poorly distinguishes simple liver cysts (SLC) from biliary cystic tumors (BCT), with high sensitivity but low specificity for BCT. We extended this question to a cohort that also includes echinococcal cysts, a third diagnostic category not addressed in prior imaging-accuracy studies, and assessed inter-reader agreement.

Methods: 77 patients who underwent surgery for a hepatic cystic lesion, with complete pre-operative imaging review by at least one of two independent, blinded readers, were retrospectively identified. Readers recorded lesion features (wall irregularity, septations, intracystic components, biliary dilatation, lymphadenopathy, segmental extent) and an overall diagnostic impression, compared against final pathology (SLC, BCT, or echinococcal cyst) using chi-square/Kruskal-Wallis tests, Cohen kappa, and standard diagnostic-accuracy metrics.

Results: Final pathology showed 51 SLC, 22 echinococcal cysts, and 4 BCT. SLC patients were significantly older than echinococcal or BCT patients (median 66 vs. 48.5 vs. 54.5 years, p=0.001). A suspicious intracystic component was the strongest imaging discriminator for both readers (p=0.0001 and p=0.0035), most frequent in echinococcal cysts, consistent with daughter cysts. Multisegment involvement, assessed from one reader`s data, was significantly associated with diagnosis (p=0.021). For BCT, imaging showed high negative predictive value (96–98%) but low positive predictive value (8–14%); the higher-performing reader achieved 69% sensitivity and 90% PPV for SLC, versus 12% and 60% for the other. Inter-reader agreement was fair for overall impression (κ=0.27) but moderate-to-substantial for individual features (κ 0.33–0.74).

Conclusion: Pre-operative imaging reliably excludes BCT but poorly confirms it, consistent with prior literature, now extended to a cohort including echinococcal cysts. Intracystic component is the most robust discriminating feature across all three cyst types. Inter-reader variability, greatest for overall impression, suggests standardized reporting criteria could improve diagnostic consistency.