Quantitative Radiographic Assessment of Medial Tibial Rim Osteophyte Morphology in Knee Osteoarthritis: Association with Kellgren-Lawrence Grade

<title>Abstract</title> <p> Objectives To develop and evaluate an ImageJ/Fiji-based quantitative radiographic workflow for characterizing localized medial tibial rim contour morphology in knee osteoarthritis and to determine whether curvature-derived descriptors are associated with consensus-defined Kellgren–Lawrence (KL) grade after accounting for analyzable contour length. Methods In this retrospective single-center imaging methods study, posteroanterior weight-bearing knee radiographs obtained between January 2023 and December 2024 were analyzed in a curated, grade-stratified cohort of 100 knees, with 20 knees in each KL grade from 0 to 4. The cohort was intentionally curated as a consensus-confirmed methodological sample rather than as an unselected clinical population. Medial tibial rim contours were manually traced within a prespecified anatomic window using ImageJ/Fiji, and curvature metrics were extracted from spline-smoothed contours. Mean curvature (κ _mean ) and curvature variance [Var(κ)] were prespecified as primary curvature-derived descriptors. Realized analyzable contour length (L _segment ) was included as a quality-control covariate. Associations with KL grade were evaluated using analysis of covariance, with complementary rank-based analyses. Intra-observer repeatability was assessed using intraclass correlation coefficients (ICCs). Results Curvature metrics were successfully extracted in all 100 knees. Intra-observer repeatability was good for κ _mean , Var(κ), and L _segment , with ICCs of 0.81, 0.76, and 0.88, respectively. L _segment differed significantly across KL grades, F(4,95) = 11.42, <bold>P</bold>  &lt; 0.001, partial η² = 0.19. After adjustment for L _segment , KL grade remained significantly associated with κ _mean , F(4,94) = 5.50, <bold>P</bold>  &lt; 0.001, partial η² = 0.19, and Var(κ), F(4,94) = 3.16, <bold>P</bold>  = 0.018, partial η² = 0.12. Within these L _segment -adjusted models, L _segment was not statistically associated with either curvature endpoint. Adjusted κ _mean increased from KL grade 0 to grade 2 and declined modestly at higher grades, whereas adjusted Var(κ) showed a broader upward shift across higher KL grades. Pairwise discrimination was strongest between KL grade 0 and higher grades, whereas separation among KL grades 1–4 was limited. Conclusions In a curated, grade-stratified methodological cohort, ImageJ/Fiji-derived curvature descriptors of medial tibial rim osteophyte morphology were feasible to extract and were associated with consensus-defined KL grade after adjustment for analyzable contour length. These metrics should be interpreted as adjunctive localized contour descriptors rather than validated diagnostic, prognostic, or KL-grading biomarkers. Larger multicenter studies with inter-observer, longitudinal, and external validation are required before clinical or research deployment. </p>