Text

Objectives and questions: Borderline dysplasia (BDDH) shows variable morphology and remains a treatment challenge. Surgical decision-making relies on instability markers such as femoral head decentration (FHD); however, the three-dimensional (3D) morphology of borderline hips remains undefined. This study aimed to determine whether BDDH hips with and without FHD differ in lunate surface morphology, labral morphology, and cartilage quality assessed by artificial intelligence (AI)-based MR segmentation.

Material and methods: A total of 307 consecutive hips undergoing MR arthrography between 2019 and 2025 were included in the current study. Based on lateral center-edge angle (LCEA), hips were classified as normal (25–33°, control), BDDH (20–25°), or dysplasia (DDH) (<20°, pathological reference). The primary comparison was within the BDDH cohort (n = 77), subdivided into hips with and without FHD (n = 27 and n = 50). Radiographic evaluation included 17 parameters assessing coverage (LCEA, acetabular index, extrusion index, and femoral head coverage), instability signs (e.g., upsloping sourcil, Shenton line), femoral torsion, acetabular version, and all three retroversion signs. 3D analysis was performed using an in-house validated AI-based algorithm for automatic segmentation of hip cartilage and labrum, including volumetric delayed gadolinium-enhanced MRI (dGEMRIC) for assessing cartilage quality. Cartilage thickness, surface area, volume, labral volume, surface and cross-sectional area, and dGEMRIC were measured at each clock-face position in absolute and relative terms (area/volume relative to femoral head), yielding 170 variables. Group comparisons were performed using one-way ANOVA and independent t tests and Fisher’s exact tests.

Results: Demographics did not differ between BDDH hips with and without FHD, including age (31 ± 10 vs 31 ± 10 years; p = 0.975) and sex (63% vs 52% female; p = 0.471). BDDH hips with FHD showed increased femoral torsion (34.1 ± 10.6° vs 25.7 ± 15.0°, p = 0.013) and a higher prevalence of a positive Shenton line (11% vs 0%, p = 0.040); no other radiographic parameters differed significantly. Morphological 3D parameters did not differ between BDDH hips with and without FHD (all p ≥ 0.144). However, BDDH hips with FHD demonstrated lower dGEMRIC values than hips without FHD (p < 0.001–0.045), particularly in postero-inferior and superior peripheral regions corresponding to the direction of decentration. Across DDH, BDDH, and normal hips, most 3D parameters differed significantly; relative cartilage area was lowest in dysplastic hips and highest in normal hips (p < 0.001).

Discussion and conclusions: BDDH hips with and without FHD showed identical 3D cartilage and labral morphology. FHD was associated with increased femoral torsion and lower dGEMRIC values due to contrast distribution between cartilage layers (decentration). These findings indicate that decentration in BDDH relates primarily to femoral torsion rather than acetabular morphology and may support consideration of derotational femoral osteotomy.