Integrated 18F-NOTA-FAPI-04 PET/MRI for quantitative assessment of disease activity in thyroid-associated ophthalmopathy.

PURPOSE: This study aimed to evaluate the value of integrated 18F-NOTA-FAPI-04 PET/MRI for CAS-defined clinical activity assessment in thyroid-associated ophthalmopathy (TAO) by investigating the correlation between PET- and multi-parametric MRI-derived features and their predictive performance. METHODS: A total of 33 TAO participants were prospectively enrolled and underwent intergrated 18F-NOTA-FAPI-04 PET/MRI. According to the Clinical Activity Score (CAS), Eyes were classified as active (CAS ≥ 3) or inactive (CAS<3). For patient-level analysis, participants were classified as inactive only when both eyes were inactive; otherwise, they were classified as active. PET-derived parameters (maximum standardized uptake value [SUVmax], mean standardized uptake value [SUVmean], metabolic tumor volume [MTV], and total lesion FAPI uptake [TLF]), as well as first-order features (mean, minimum, maximum, median, skewness, and kurtosis) from MRI-derived parametric maps (normalized T2-weighted imaging-derived parameter [T2WI], heterogeneity index [α], distributed diffusion coefficient [DDC], diffusion coefficient [D], true relative blood flow [rBF], mean diffusivity [MD], and mean kurtosis [MK]) were extracted. Repeatability was assessed using the Dice similarity coefficient (DSC) of VOIs delineation and the intraclass correlation coefficient (ICC) for quantitative parameters. Correlation analyses between PET and MRI features, comparisons between active and inactive group, and discriminative performance evaluations of PET, MRI, and combined PET/MRI models were performed at both patient and eye levels. Support vector machine (SVM) and receiver operating characteristic (ROC) analysis were performed to compare discriminative performance. RESULTS: VOIs delineation demonstrated good interobserver and intraobserver agreement, with DSC of 0.850 (95% CI: 0.815-0.886), 0.880 ± 0.068 (95% CI: 0.831, 0.929), and 0.901 (95% CI: 0.838, 0.965), respectively. All PET/MRI quantitative parameters showed high reproducibility (ICC > 0.907). PET/MRI correlations were generally weak at patient-level but substantially stronger at eye-level, particularly for associations between PET metrics and T2WI-, α-, D-, and MD- derived features. At the eye level, SUVmean, MTV, and TLF were significantly higher in active TAO than in inactive group, whereas SUVmax showed no significant difference. Several MRI-derived parameters, including T2WI, α, D, MD, and MK features, also differed significantly between groups. For discriminative performance, SUVmean showed the best performance among single PET parameters at the patient-level (AUC = 0.759), whereas MTV performed best at the eye level (AUC = 0.767). The combined PET/MRI model achieved the highest performance at both patient and eye levels (AUC = 0.844 and 0.859, respectively). CONCLUSION: Integrated 18F-NOTA-FAPI-04 PET/MRI provides complementary information on FAPI uptake and MRI-derived tissue microstructural changes in TAO. SUVmean / MTV (PET) and T2WI-, D-derived features (MRI) showed potential associations with CAS-defined disease activity. The combined PET/MRI model demonstrated improved discriminative performance for assessing active and inactive ; however, further validation in larger independent cohorts is required. CLINICAL TRIAL REGISTRATION: NCT07400224 (2/3/2026).