STATISTICAL SHAPE MODELING OF CAM-TYPE FEMOROACETABULAR IMPINGEMENT OF THE PROXIMAL FEMUR USING 3D MRI

Femoroacetabular impingement (FAI) is a hip disorder involving bony abnormalities that cause collisions of the femoral head-neck junction with the acetabular rim, leading to labral tears, cartilage damage, and early-onset osteoarthritis. Cam FAI is characterized by a focal bony prominence at the anterolateral head-neck junction, while Pincer FAI involves acetabular over-coverage. Current clinical assessment relies on 2D measurements such as the alpha angle for Cam FAI, which may not fully capture the 3D nature of these deformities. Statistical shape modeling (SSM) offers a data-driven approach to characterize complex 3D morphological differences between healthy and pathological anatomy.

To develop a principal component analysis (PCA)-based SSM pipeline for 3D MRI to identify and characterize shape features associated with Cam and Pincer FAI. This preliminary analysis focuses on Cam morphology of the femoral head-neck region using a balanced subset of the full cohort.

This study was approved by the IRB. We selected 10 asymptomatic controls (5F; age 32.0±4.9 years) and 41 patients with unilateral mixed-type FAI confirmed during arthroscopy (22F; age 36.2±7.9 years). All subjects received a 3T MRI scan (preoperative for patients) including a 3D VIBE sequence with Dixon water-fat separation of the entire pelvis (TR=10ms, TE=2.4/3.7ms, matrix=320 × 320 × 160, 1.0mm isotropic). In this preliminary study, we analyzed only the symptomatic hip that underwent arthroscopy, matching the side for controls. The femur was manually segmented with assistance from a musculoskeletal radiologist. Surface meshes of the femoral head-neck region were generated using discrete marching cubes, followed by windowed sinc and Taubin smoothing. The SSM pipeline consisted of three stages: (1) RBF-based non-rigid registration to a reference subject and alignment of 10 control meshes to obtain a mean shape representing normal anatomy; (2) a combined PCA model trained on all 20 subjects registered to the control mean; (3) projection onto PCA space, where patient and control weight distributions were compared using Welch's t-tests, Cohen's d effect sizes, FDR correction, and power analysis. Shape modes were visualized as signed surface displacement maps at ±2SD.

Five PCs explained 92.6% of total shape variance (Table 1). PC1-PC3 (86.0%) captured overall femoral head size, head shape, and neck contour variation. PC4 (4.0%) captured a focal outward displacement (up to 2.9mm) at the anterolateral head-neck junction consistent with cam morphology (d=0.71, p=0.132). Figure 1 shows the PC4 signed surface displacement maps at -2SD, mean, and +2SD, where the -2SD extreme reveals a clearly localized outward bump (red) at the head-neck junction corresponding to the cam deformity, while the +2SD extreme shows a smooth contour. FAI patients had lower PC4 scores than controls, indicating their shapes were oriented toward the Cam-like extreme. PC7 (1.14%) showed the strongest group difference (d=1.3, p=0.01, p_FDR=0.098). Power analysis indicated that 28 and 12 subjects per group would achieve 80% power for PC4 and PC7, respectively.

PCA-based SSM identified a localized shape mode (PC4) capturing Cam morphology at the femoral head-neck junction with a moderate-to-large effect size, though it did not reach statistical significance in this preliminary study. These findings suggest that clinically meaningful deformities may be captured by anatomically interpretable shape modes, even with limited statistical power in small cohorts. Ongoing work includes segmentation refinement to enable inclusion of the full cohort and contralateral hips to increase statistical power, as well as extending the framework to include the acetabulum to characterize pincer morphology and differentiate cam and pincer contributions in mixed-type FAI.