A combined radiomics-clinical nomogram for assessing MRI-defined femoral Head collapse status in osteonecrosis of the femoral Head.

BACKGROUND: Osteonecrosis of the femoral head (ONFH) frequently progresses to femoral head collapse, resulting in irreversible joint dysfunction and the need for total hip arthroplasty. Reliable assessment of magnetic resonance imaging (MRI)-defined collapse status remains limited with conventional assessment methods. PURPOSE: To develop and validate a multi-modal nomogram integrating MRI-based radiomics signatures and clinical risk factors for the assessment of MRI-defined collapse status in patients with osteonecrosis of the femoral head (ONFH). METHODS: This retrospective study analyzed 149 ONFH patients (99 non-collapsed, 50 collapsed), partitioned into training (n = 104) and testing (n = 45) sets. Radiomics features were extracted from MRI-based necrotic lesion segmentations. A Radscore, derived via least absolute shrinkage and selection operator (LASSO) regression, was integrated with selected clinical variables to construct a combined nomogram. RESULTS: The combined clinical-radiomics model demonstrated favorable discriminative performance, achieving an AUC of 0.940 (95% CI: 0.858-0.994) in the testing set, which outperformed the clinical-only model (AUC = 0.589) and the radiomics-only model (AUC = 0.818). Using the cutoff determined by the Youden index in the training cohort, the combined model exhibited a sensitivity of 80.0%, specificity of 90.0%, and overall accuracy of 86.7% in the testing cohort. Calibration curves showed acceptable agreement between model-predicted probabilities and observed MRI-defined collapse status. CONCLUSION: The integrated clinical-radiomics nomogram showed promising performance for non-invasive assessment of femoral head collapse status in patients with ONFH. This quantitative approach may assist individualized collapse-status stratification, but external validation and prospective clinical workflow studies are required before routine clinical implementation.