Automated Sinus MRI Phenotyping with PARASIDE: Construct-Specific Validation against Independent Manual Annotations and Radiological Assessments in a Population-Based Cohort

Manual sinus magnetic resonance imaging (MRI) annotations are valuable but difficult to scale. We evaluated PARASIDE, an nnU-Net-based framework for automated sinus compartment segmentation on T1-weighted MRI, against independent manual and radiological assessments from the Study of Health in Pomerania. We linked 12,867 sinus sides from 3,217 participants. Analyses defined before outcome modelling evaluated soft-tissue fraction for healthy versus abnormal sides, inferior soft-tissue centroid position for basal versus apical maxillary opacification, and total frontal volume for aplasia/hypoplasia. Soft-tissue fraction discriminated abnormal sides in frontal (area under the receiver operating characteristic curve [AUC] 0.850, 95% confidence interval [CI] 0.834–0.866) and maxillary sinuses (AUC 0.828, 95% CI 0.818–0.839). The topographic marker discriminated basal from apical opacification (AUC 0.746 left; 0.782 right). Total frontal volume discriminated historical aplasia/hypoplasia ratings (AUC 0.988); a previously established near-absence threshold was highly specific but insensitive. Historical maxillary volumetric masks showed strong overlap and volume association for total and aerated compartments, whereas frontal comparisons reflected a predefined caudal boundary. The historical polyposis rating category remained weakly separable, and surgery-related features corresponded more closely to visible postoperative morphology than to self-reported surgery. PARASIDE reproduced anatomically measurable MRI phenotypes, whereas morphology-specific constructs required dedicated reference standards.