A Novel Nomogram Based on Orbital MRI Parameters for Early Prediction of Dysthyroid Optic Neuropathy in Thyroid-Associated Ophthalmopathy

Purpose: This study aimed to assess orbital magnetic resonance imaging (MRI) differences across various severity levels of thyroid-associated ophthalmopathy (TAO) and compare them to healthy controls. Additionally, it investigated the correlation and diagnostic potential of multiple MRI parameters for dysthyroid optic neuropathy (DON). Methods: Patients with TAO were classified according to the EUGOGO severity scale into mild TAO (66 orbits), moderate-to-severe TAO (59 orbits), and DON (21 orbits). A control group of 25 healthy orbits was also included. Patients underwent comprehensive ophthalmic and orbital MRI examinations, with parameters measured on the Carestream GCRIS 2.1 workstation. Generalized estimating equation (GEE) was used to analyze the clinical and MRI data, comparing differences across groups. Univariate and multivariate binary logistic regression identified independent risk factors for DON. The predictive value of MRI parameters for DON was further evaluated using a nomogram, receiver operating characteristic (ROC) curves, and decision curve analysis (DCA). Results: The extraocular muscle area significantly differed between the control and mild TAO groups (P < 0.05), as well as between both groups and the DON group (P < 0.05). Significant variations in AMR-ON and ALR-ON were observed across the control, mild TAO, and moderate-to-severe TAO groups compared to the DON group (P < 0.05). Muscle Index (MI) and Crowding Index (CI) also showed notable differences between the control, mild-to-severe TAO, and DON groups (P < 0.05). Regression analysis revealed superior rectus muscle area (SRA) and AMR-ON as independent predictors of DON (P < 0.05). The SRA + AMR-ON nomogram demonstrated high predictive performance for DON, with an AUC of 0.906 (P < 0.001), sensitivity of 90.5%, and specificity of 82.4%. DCA further confirmed that the SRA + AMR-ON model outperformed MI and CI in predictive efficacy for DON. Conclusions: A reduction in AMR-ON and ALR-ON correlates with increased TAO severity, while an enlargement in extraocular muscle area, MI, and CI reflects worsening TAO. SRA and AMR-ON serve as independent predictors for DON, and the predictive model combining these parameters is highly effective in forecasting DON onset.