MRI-Based Radiomics and Deep Learning of Thyroid Eye Disease: One Slice Speaks Volume

Motivation: Traditional multi-slice MRI segmentation is time-consuming and resource-intensive. This study investigates the potential of single-slice MRI for predicting intravenous glucocorticoid (IVGC) treatment response in thyroid eye disease (TED). Goal(s): To evaluate whether single-slice MRI can effectively predict treatment responses in TED patients compared to multi-slice models. Approach: A retrospective study of 127 TED patients treated with IVGC. Radiomics analysis and deep learning algorithm were applied to both single-slice and multi-slice MRI data. Results: Single-slice models demonstrated comparable performance to multi-slice models, suggesting single-slice MRI as a cost-effective alternative for clinical use of TED. Impact: This study highlights the feasibility of single-slice MRI as an efficient, cost-effective alternative to multi-slice segmentation for predicting IVGC treatment response in TED patients. It opens avenues for more accessible clinical applications, reducing time and resource requirements while maintaining performance.