PREDICTING TROCHLEAR DYSPLASIA FROM 3-D CORTICAL BONE AND SHAPE PARAMETERS

Abnormal shape of the distal femur is the defining characteristic of trochlear dysplasia (TD), leading to abnormal contact forces and early degeneration at the patellofemoral joint, as well as an increased risk of lateral patellar dislocation. In severe clinical cases treatment with various surgical techniques aims to normalise trochlea morphology and medialise patellofemoral knee forces. 1) To evaluate differences in 3-D bone distribution at the distal femur between normal individuals and TD; and 2) to demonstrate how these parameters discriminate between these two groups. 20 healthy volunteers without radiological signs of TD and 20 individuals with Dejour type B or D dysplasia and patellar instability were recruited, both groups aged 18-45 years with no history of arthritis, connective tissue disease, or traumatic knee conditions requiring physical or operative treatment. The TD group were awaiting surgical trochleoplasty after failing conservative treatment. With a 50:50 proportion of left and right knees in the TD group, a random matched 50:50 split of left and right knees were chosen for comparison in the healthy volunteers. Weight bearing computed tomography (WBCT) of the target knee in 20°of flexion using a Carestream Onsight© scanner, with full imaging details. The distal femur was semi-automatically segmented using inhouse software Stradview. A canonical distal femur was registered to each individual to build a 3-D statistical shape model from the registrations using inhouse software wxRegSurf. Cortical bone mapping was also performed on all femurs to measure 3-D distribution of cortical thickness (CTh) and trabecular attenuation (TA), which were then transferred to the canonical femur. Statistical parametric mapping (SPM) tested the dependence of the 3-D distribution of the bony parameters on the presence of TD. Mean values were taken from within the largest SPM significant region of interest (ROI) for CTh and TA, and used along with sex, age and shape mode coefficients in receiver operating characteristic analysis to deliver area under the curve (AUC) in a leave-one-out cross-validation classifier predictive model for TD. Mean (± SD) age of all participants was 30.5 ± 5.2 years, with an even split of sexes in both groups. SPM showed significantly thicker CTh in the TD group along the outer lateral margin of the lateral trochlea by up to 0.6 mm (left figure, unmasked blue patch). TA was significantly lower by up to 100 units in the TD group at t