3D MRI-based knee shape and its association with incident osteoarthritis and sex differences (data from OAI dataset)

Abstract

Listen

Translate article

Purpose: This study utilizes a 3D magnetic resonance imaging (MRI) based statistical shape modeling (SSM) method to analyze differences in bone shape of the knee joint that may be associated with incident OA. The purpose of this study is two-fold: (i) to investigate differences in knee bone shape between incident osteoarthritis (OA) cases and controls; and (ii) to investigate sex differences in bone shape. Methods: 20 subjects (10 women) without radiographic signs of OA at baseline (KL=0), but with incident OA at 4 years follow up (KL=2) (cases) were identified randomly from the Osteoarthritis Initiative dataset and matched for age, BMI, and gender in a ratio of 1:1 to 20 controls (10 women) who did not show radiographic evidence of OA over the examination period (KL=0 at baseline and at 4 years). All 40 subjects underwent right knee 3.0 Tesla MR scans. Baseline sagittal 3D double-echo steady-state (DESS) with water excitation images (field of view = 140 mm, number of slices = 160, slice thickness = 0.7 mm, TE/TR 4.7/16.3 mm, TR pixel spacing (0.37 mm, 0.46 mm)) were processed using an in-house Matlab program that segmented the femur and tibia semi-automatically from a collection of plotted markers. A vertex-to-vertex correspondence of the reference knee was mapped onto each candidate surface using a fully automatic landmark-matching algorithm, invariant of size and rotation. Principal Component Analysis (PCA) of the covariance matrix for the landmarks provided relevant shape information. Each case was described by 20 tibia and femur modes (F1-F20, T1-T20) that represent aspects of the 3D bone shape. The mean shape and±5 standard deviations of each mode are presented (Figure 1). Results: Cases and controls had similar mean age and BMI (cases: age = 58.1±6.6 years, BMI = 26.3±2.9 kg/m2, controls: age = 58.1±7.0 years, BMI = 26.4±2.9 kg/m2, p > 0.05). With respect to femoral modes, cases had significant differences in modes F4 (cases: −97.74±178.08, controls: 92.86±176.39 p = 0.002) and F8 (cases: −44.06±107.89, controls: 41.86±124.02, p = 0.027). Negative values in F4 indicated an increase in the length and width of the lateral condyle of the femur relative to the medial condyle, whereas negative F8 values suggested a deeper trochlear groove. For the tibia, differences were less pronounced. In mode T3, cases were borderline different from control knees (cases: −121.92±283.35, controls: 136.26±471.43, p = 0.052). In T4, cases also showed a trend towards significance (cases: −100.91±337.68, controls: 112.79±297.13, p = 0.053). Negative values represent elevation of the medial tibial plateau in T3, and a flatter posterior tibial slope in T4. With respect to our secondary aim, the analysis of bone shape between sexes with knee OA showed that femoral mode F7 was −40.1±134.5 for men, while it was 57.7±103.8 (p = 0.019) for women. Similarly, mean F13 was significantly different in men (−26.15±86.5) versus women (38.1±93.8, p = 0.0032). Negative values in F7 indicated a narrowed intercondylar notch and shallow trochlear groove, whereas positive values showed widened condyles and a deeper trochlear groove. Negative F13 values yielded both a deeper intercondylar notch and trochlear groove. Tibial mode T3 was also significantly different between men (99.7±435.8) and women (−176.4±256.2, p = 0.045). Shape description for T3 is as above. Conclusions: This study suggests that shape differences including a wider and longer lateral condyle (F4), deeper trochlear groove (F8), elevated medial tibial plateau (T3), and flatter posterior slope (T4), may be associated with incident OA. Also, tibial plateau elevation (T3), intercondylar notch width (F7/F13), and trochlear depth (F7/F13) are associated with sex differences. Future analyses of shape modes may allow for their use as target biomarkers in OA.