Abstract
Objectives:
Femoroacetabular impingement (FAI) is a leading cause of hip pain in young adults and is recognized as a key risk factor in the subsequent development of hip osteoarthritis (OA). Recently, Kamenaga et al proposed a novel approach to induce an injury to the medial third of the proximal femoral epiphysis of an immature rabbit hip and confirmed that this results in a head-neck cam-type deformity that morphologically represents that observed in human FAI. Using this proposed animal model, Kamenaga et al. also confirmed OA progression secondary to cam-FAI deformity via 2-dimensional image analysis and histological analysis. However, 3-dimensional (3D) image analysis and biomechanical confirmation of an induced cam is lacking. Statistical shape modeling (SSM) implements particle-based modeling to express an average shape for a 3D object and to capture variability both within and across cohorts. This study aims to use SSM to generate 3D femurs from micro-computed tomography (CT) images to quantify morphological differences between femurs with and without induced cam-FAI deformity to take a crucial step towards validating this proposed animal model to human FAI.
Methods:
Six-week-old New Zealand White rabbits (n = 13) were subject to right hip surgery, with the left hip used as a control. An epiphyseal injury in the medial femoral head was created using a 1.6-mm drill. Hips were harvested bilaterally at 28 days after surgery and micro-CT images were taken. This study received approval from the hospital and university Institutional Animal Care and Use Committee. All 3D reconstructions were segmented and created in 3D Slicer (v5.2.2). The reconstructed surfaces were triangulated, and segmentation artifacts were removed by using median smoothing tools in 3D Slicer. 2 cam-FAI hips and 2 control hips were excluded due to incomplete femoral head 3D reconstruction. Eleven cam-FAI hips and 11 control hips were aligned in ShapeWorks using an iterative closest point algorithm to minimize root mean square distances between surfaces. 2048 particles were placed on each femur using a hierarchal splitting pattern with a gradient decent energy function that was determined using a system of repulsive forces until a steady state between particles was achieved. Each hip had constraints placed such that the particles spread over the same surface area of interest. The generalized Procrustes algorithm was applied during the placement of these particles to normalize with respect to scale and to align shapes with respect to rotation and translation. Group labels in ShapeWorks separated the particles into cam-FAI hips and control hips. Mean shapes were constructed as the mean location of the corresponding particles. The morphological differences between mean shapes were calculated using a Hotelling T2 test with the null hypothesis that the cam-FAI and control groups were drawn from the same distribution. Principal component analysis (PCA) was used to reduce the high dimensionality of variation across SSM correspondence into linearly uncorrelated modes that characterize the variation within the group of femur shapes. A parallel analysis was used to determine the number of statistically significant modes of variation. PCA was also applied to the cam-FAI group and control group separately to analyze variations within each group.
Results:
The Hotelling T2 test demonstrated significant differences between the cam-FAI and control mean shapes (p < 0.02). The cam-FAI mean shape protruded above the control mean by a maximum of 0.8 mm in the anterolateral head–neck junction (Figure 1). Morphologically, there are sustained protrusions of ~0.6 to 0.8mm over the anterosuperior aspect of the femoral neck and over the anteroposterior (AP) midline of the femoral head-neck junction. Maximum deviations between cam-FAI hips and the mean control hips were between 0.2-1.6 mm over the anterosuperior aspect of the femoral neck and over the AP midline of the femoral head-neck junction. The first 8 modes accounted for 90.4% of the cumulative variation among all femurs (Figure 2). Specifically, mode 1 captured 38.2% of the variation, mode 2 captured 16.5%, mode 3 captured 13.0%, mode 4 captured 7.0%, mode 5 captured 6.5%, mode 6 captured 3.5%, and mode 7 captured 3.3%. PCA performed on the cam-FAI and the control groups independently demonstrated similar areas of intra-group variation (Figure 3).
Conclusions:
This study further validates the rabbit FAI model presented by Kamenaga et al by detailing the specific morphological features of cam deformity and demonstrating statistical significance. Particularly, the morphological difference to the healthy control shows a bone protrusion over the anterosuperior aspect of the femoral neck and over the AP midline of the femoral head-neck junction, which is in line with clinical observations of cam-FAI. To our knowledge, this is the first study that has analyzed 3D statistical shape modeling of cam-FAI pathology using an animal model. Importantly, this model seems to follow the clinical scenario of FAI progression. So, defining and measuring the 3D morphology changes in this validated small animal model will provide a low-cost base for effective translational research on hip OA and could be used as a platform to understand in-depth the mechanisms of hip OA, test interventions, and help translate discoveries to clinical care.
