Objective To evaluate the agreement and clinical value of computed tomography (CT), magnetic resonance imaging (MRI),and Hertel exophthalmometer for measuring exophthalmos in patients with moderate-to-severe thyroid-associated ophthalmopathy(TAO).Methods A total of 78 patients (156 eyes) with moderate-to-severe TAO admitted to the Department of Radiation Oncology of the First Hospital of Shanxi Medical University from October 2018 to November 2024 were retrospectively included as the research subjects. Multimodal measurement methods were adopted: based on CT/MRI axial images, the vertical distances from the corneal apex to the line connecting the bilateral zygomatic orbital processes (Method 1) and the line connecting the medial and lateral orbital margins of one side of the orbit (Method 2) were measured respectively. In addition, based on sagittal images, the vertical distance from the corneal apex to the line connecting the anterior edge of the zygomatic frontal process (Method 3) was measured, and the ratio of the orbital eyeball area to the orbital area was innovatively calculated (GA/OA, Method 4). The Hertel exophthalmometer (Method 5) was used to measure the degree of exophthalmos. The Bland-Altman consistency analysis was used to evaluate the consistency between observers, and the Spearman and Pearson correlation coefficients were used to evaluate the correlation between the measurement results of the imaging modal methods (Methods 1-4) and the clinical modal methods (Method 5). The Wilcoxon signed-rank test and paired t test were used to analyze the differences in the measurement of proptosis under different modalities. Results Imaging modalities (Methods 1-3) demonstrated good inter-observer agreement. All imaging-based methods showed moderate correlations with clinical measurements (|r|=0.575～0.75, all P＜0.01). Measurements obtained using the same method were significantly higher in MRI than in CT modalities (all P＜0.05). Conclusion In patients with moderate-to-sever TAO, methods 1-3 under imaging modality have good consistency between different observers, which can be used for clinical needs for accurate evaluation and follow-up. Method 4 may not be usedas a conventional detection method due to its initations. In patients with moderately-severe TAO, since the difference with identical protocols between CT and MRI modalities, it is imperative to maintain consistency in imaging modality selection (either CT or MRI) throughout follow-up periods. This standardization is crucial for minimizing measurement bias and ensuring data comparability. Future work requires optimizing measurement standards via multicenter studies, enabling cross-modality conversion, establishing personalized calibration thresholds, and developing AI-assisted tools to enhance work efficiency