3D Patellar Shape is Associated with Patellar Dislocation: an Automated Coordinate Algorithm and Statistical Shape Modeling Analysis

Purpose: To establish an automated, landmark-based patellar coordinate system for standardized alignment, develop a patellar statistical shape model (SSM), and quantify 3D morphological variations associated with patellar dislocation (PD). Methods: Patellar surface models were reconstructed from CT/MRI scans of 54 participants (33 PD, 21 controls). An automated coordinate system was established and quantitatively validated. Demographic/morphometric risk factors were assessed using logistic regression. An SSM was built for the entire cohort, and principal component analysis (PCA) was used to extract major 3D shape modes. Between-group differences in PC scores were evaluated with multiple-testing control and covariate adjustment. A logistic regression classifier based on shape modes and demographics was evaluated using stratified 10-fold cross-validation. Results: The automated coordinate system showed high repeatability. Patellar linear dimensions and centroid size did not differ between groups and were not independent predictors. Two robust shape modes differentiated PD from controls: PC4 (thickness/facet morphology) and PC7 (facet-edge morphology). A cross-validated classifier showed good in-cohort discrimination (mean AUC ≈ 0.91). Conclusion: In this cohort, PD was associated with localized 3D articular-surface shape patterns, characterized by a prominent medial facet, a flattened posterolateral facet, and accentuated facet margins, without corresponding differences in linear dimensions. The automated coordinate system and SSM provide a reproducible approach for quantitative patellar phenotyping. These shape modes may deepen understanding of PD pathomechanics and provide a quantitative basis for future, externally validated risk modeling in diverse populations.