The Impact of Radiomics Image Analysis on Adult Hip Pathologies: A Scoping Review

Le résumé fourni par la source

Radiomics promises quantitative biomarkers extracted from routine hip imaging to support diagnosis, prognosis, and surgical planning, but current evidence is fragmented across pathologies, modalities, and computational pipelines. We conducted a scoping review following PRISMA-ScR and the Population-Concept-Context framework, including peer-reviewed original studies on adults (≥18 years) that applied radiomics or deep-radiomics to hip imaging (X-ray, CT, MRI, DEXA) with clinically relevant outcomes. PubMed (MEDLINE), Embase and Scopus (Elsevier) were searched from 1 January 2021 to 30 August 2025 and complemented by snowballing; screening and data charting were performed in duplicate. Given heterogeneity, findings were synthesized narratively by a priori clusters. In fragility/osteoporosis, opportunistic CT and radiograph-based models frequently achieved AUCs around 0.90-0.96, while DXA-radiomics added information beyond bone mineral density/FRAX and trabecular MRI provided complementary microarchitectural signals. For osteonecrosis of the femoral head, multisequence MRI enabled early diagnosis with AUCs > 0.94; radiomics differentiated transient bone marrow edema with AUCs~0.92-0.94 and predicted collapse using radiographs or MRI with AUCs~0.85-0.90, including automated pipelines with external validation around 0.85. In femoroacetabular impingement, 3D Dixon-MRI studies reported very high performance (~0.97-1.00) with preliminary multicenter generalizability and added value from periarticular soft-tissue features. In total hip arthroplasty, radiomics anticipated press-fit cup stability from preoperative radiographs (AUC~0.82) and predicted 6-month functional recovery using clinico-radiomic CT models (AUC~0.95). Across clusters, methodological robustness was variable (sample sizes, harmonization, leakage control, external/temporal validation, calibration, clinical utility). Radiomics for adult hip disorders shows tangible translational promise in opportunistic screening, complex differential diagnosis, and perioperative decision support, but broader clinical adoption will require multicenter datasets, IBSI-aligned standardization, transparent reporting of calibration and decision-curve analyses, and prospective validation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.