Pilot study using synthetic MRI for determining the disease activity of thyroid-associated ophthalmopathy

OBJECTIVE: To investigate the value of synthetic magnetic resonance imaging (SyMRI) of extraocular muscles (EOMs) in determining the disease activity of thyroid-associated ophthalmopathy (TAO). METHODS: Seventy-seven patients with TAO (31 active; 46 inactive) were prospectively enrolled. Orbital SyMRI was performed and post-processed to generate quantitative maps (T1 map, T2 map, and proton density [PD] map) and contrast-weighted images (double inversion recovery [DIR], short inversion recovery [STIR], and fluid-attenuated inversion recovery [FLAIR]). Maximum, mean, and minimum values of PD, T1 relaxation time (T1RT), and T2 relaxation time (T2RT) of EOMs, and the signal intensity ratio (SIR) of contrast-weighted images were calculated and compared between two groups. Multivariate logistic regression identified independent predictors of disease activity. Receiver operating characteristic curve analyses and DeLong tests evaluated and compared the diagnostic performance of significant parameters and combined models. RESULTS: Active TAO patients had higher values of both quantitative and semi-quantitative parameters than inactive patients, except for FLAIR-Min (all p < 0.05). DIR-Mean, PD-Mean, T1RT-Mean, and T2RT-Mean of EOMs were independently associated with active TAO (all p < 0.05). The combined model integrating four independent predictors showed higher diagnostic performance (AUC, 0.884; specificity, 89.13%; sensitivity, 77.42%) than DIR-Mean (AUC=0.802), PD-Mean (AUC=0.818), T1RT-Mean (AUC=0.835), or T2RT-Mean (AUC=0.824) alone in distinguishing active from inactive patients (all p < 0.05). CONCLUSION: SyMRI might be one promising imaging modality for determining the disease activity of TAO. ADVANCES IN KNOWLEDGE：: SyMRI quantifies PD, T1RT, and T2RT in a single scan with clinically acceptable time. Our study innovatively synthesized multiple contrast-weighted images, improving the efficacy of TAO staging.