MRI-based virtual radiographs of the pelvis for assessment of acetabular morphology

OBJECTIVES: We generated magnetic resonance imaging (MRI)-based virtual radiographs (VRs) from hips with/without intra-articular contrast and compared them with conventional anteroposterior pelvic radiographs (CR) regarding measurement bias, interobserver and intraobserver agreement, and diagnosis of acetabular deformities; we also assessed hip deformity prevalence on VR in asymptomatic participants. MATERIALS AND METHODS: Retrospective study of 50 patients (100 hips; mean age 34 years) with femoroacetabular impingement or hip dysplasia undergoing CR and unilateral magnetic resonance arthrography. Additionally, 53 asymptomatic participants underwent noncontrast MRI. Segmentation of the osseous pelvis was performed on T1-weighted volume interpolated breathhold examination (VIBE) Dixon MRI using 3D nnU-Net, pseudo-computed tomography generation, and cone-beam projection to generate VR. Acetabular coverage (lateral center edge angle (LCE), acetabular index (AI), anterior/posterior coverage, extrusion index (EI)), and acetabular version were compared between CR and VR. Interobserver agreement was assessed using intraclass correlation coefficients (ICC) and κ statistics; bias with Bland-Altman analysis. RESULTS: Interobserver agreement for VR ranged from moderate to almost perfect (ICC/κ 0.60-0.94), except for the AI. LCE showed no significant bias with ( - 0.2° ± 2.5, p = 0.663) or without contrast ( - 0.2° ± 2.2, p = 0.530). Systematic differences occurred for AI, EI, anterior/posterior coverage, and ischial spine sign; 57-75% of discrepancies were within CR interobserver variation. Diagnostic classification showed substantial agreement between CR and VR (κ = 0.62), comparable to CR interobserver agreement (κ = 0.70). In asymptomatic participants, 13% had acetabular dysplasia, 68% normal coverage, and 19% overcoverage. CONCLUSION: MRI-based VR provides acetabular morphology assessment comparable to CR and enables radiation-free evaluation in symptomatic patients and asymptomatic populations. KEY POINTS: Question Can MRI-based virtual radiographs reproduce conventional radiographic measurements of acetabular morphology, enabling radiation-free two-dimensional assessment in symptomatic patients and asymptomatic populations? Findings The lateral center-edge angle showed no significant bias between virtual and conventional radiographs; diagnostic classification of acetabular deformities agreed substantially (κ = 0.62), comparable to conventional radiograph interobserver agreement. Relevance Statement MRI-derived virtual radiographs can provide acetabular measurements comparable to conventional radiographs without ionizing radiation, supporting integration into three-dimensional surgical planning and enabling assessment in young patients and population-based cohorts where repeated radiographs are undesirable.