Summary

In this study we introduce a trochlea dysplasia classification system built upon 3D curvature analysis displaying better interrater reliability than existing 2D classifiers.

Abstract

Introduction

Patellofemoral trochlear dysplasia is typically classified using two-dimensional (2D) plain radiographs or transverse plane axial cuts on magnet resonance imagining (MRI) or computerized tomography (CT), which cannot capture the complex three-dimensional (3D) morphology of the joint fully. These 2D-classifications tend to be unreliable (e.g., in a recent Level 1 Evidence study, reliability of the commonly used Dejour 2-D classification was estimated with a Fleiss kappa of 0.12-0.2, interpreted as “slight”). 3D-surface geometric analysis, such as that of curvature, can capture the entire morphology of a dysplastic trochlea and thereby offer comprehensive understanding of the joint’s entire topography and thereby likely improve surgical planning. This study aims to assess curvature analysis of 3D-surface geometry as a tool to classify trochlear dysplasia.

Methods

3D-models of distal femora were generated from high-resolution CT scans of patients with histories of recurrent patellar dislocation with at least two reported dislocation events. Control patients with no history of patellofemoral disease were acquired from the New Mexico Decedent Image Database and used for comparison. Patients younger than 15 years or with congenital diseases were excluded. The cohorts were age-sex matched to achieve similar male/female ratio and age range. The curvature of the distal femora was calculated using an open-source Python library (PyMeshLab), and the trochlear minimal curvature was used to classify the surface as locally concave (groove-shaped), flat, or convex, and was overlayed on 3D-renderings of the joint. The 3D-images were categorized into three groups by two independent blinded raters. Additionally, raters interpreted the curvature maps qualitatively. The significance of differences between controls and patients was tested with a chi-squared test with

Significance

level of 0.05. The minimum sample size was determined using a priori power calculation with a power level of 0.80. Demographic differences were tested with a Fisher’s exact test. Interrater reliability was calculated using Fleiss kappa and classified according to the literature. This study was deemed exempt by our institutions institutional research board (IRB).

Results

Curvature maps were generated for 33 patients with patellar instability (female/male:24/9,age:24.1±9.2years) and 36 controls (f/m:25/11,age:25.0±9.2years) and classified according to the proposed system. No significant differences in demographics were found. Significant differences in classification were found between the two groups (Class A/B/C: patient 2/13/18; control 32/4/0 (p<0.001)). Interrater reliability was 0.86 [95%CI 0.82-0.90], which is regarded as almost perfect. Qualitatively the proximal trochlea of controls was mostly concave, while patients could be separated to groups with a flat or convex proximal trochlea while also understanding the entire tracking path.

Conclusion

This study demonstrates how 3D-curvature mapping can be used to better understand trochlear dysplasia. This method shows significant differences between the patient cohort and matched controls. The interrater reliability is "almost perfect”. Patients were primarily separated in trochlea convex (destabilizes patella) or flat (neutral effects on patella stability), while controls were primarily concave (i.e., morphology stabilizes). 3D-curvature mapping allows understanding of the full complex trochlea groove.