Comparison of Two AI-Based Tools for CT-Based Fissure Completeness Scoring in Severe Emphysema

Background Automated CT-based quantification of the fissure completeness score (FCS) can evaluate patients with severe emphysema for endobronchial valve treatment. AI-based FCS quantifying tools have been approved by the European Medical Device Regulation (MDR) or made open-source available for scientific use. These tools may demonstrate reduced sensitivity and specificity, because of diversity and quality of underlying training data. We compared the results of two MDR-approved tools for FCS quantification. Methods Retrospective two-center study, FCS was assessed by MDR-approved tools: Thirona LungQ (Thirona, Nijmegen, Netherlands) and Coreline Aview (Coreline, Seoul, Republic of Korea). First-time evaluation patients for endoscopic lung volume reduction receiving a thoracic CT scan (2024 – 06/2025) were included. FCS, lobe volume, and voxel density results from both tools were compared. Receiver operating characteristic (ROC) curves evaluated each method predicting fissure completeness, by Chartisassessment (pulmonX, Neuchatel, Switzerland). Results 63 patients (mean residual volume 241%±46.26) met the inclusion criteria. Thirona LungQ data were available from 63, Coreline Aview from 61/63 (96,8%) patients (technical reasons). For fissure completeness, correlations ranged from moderate to very strong, Spearman’s ρ values: 0.78 for the left major fissure, 0.89 for the right (all p0.97 all combinations). Chartis assessments were performed in left major (n=41), right upper (n=32), and right major (n=8; not used for AUC calculation). For the left major fissure, Aview reached an AUC of 0.729, LungQ of 0.789. In the right upper lobe, discrimination was lower for both methods, with AUCs of 0.643 (Aview) and 0.665 (LungQ). Conclusion Both tools provide highly comparable measurements for lobar volume and low attenuation voxels, with greater variability for FCS. Overall, the discriminative power of both tools is insufficient to predict fissure completeness without additional diagnostics. Larger datasets are needed to determine whether both tools combined may improve predictive performance. Publication History Article published online: 04 March 2026 © 2026. Thieme. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany