Enhancing Diagnostic Accuracy of American College of Radiology TI-RADS 4 Nodules: nomogram models based on MRI Morphological Features

Motivation: Thyroid nodules classified as ACR-TR4 present a diagnostic challenge due to their uncertain nature. Goal(s): This study aimed to develop and validate nomogram models using MRI morphological features to enhance diagnostic accuracy of ACR-TR4 thyroid nodules, thereby reducing unnecessary FNA. Approach: We analyzed 229 thyroid ACR-TR4 nodules. MRI morphological features were recorded, and nomogram and improved models were developed. The performance of models was assessed and compared with that of the ACR-TR4. Results: The nomogram showed robust discrimination performance, with an AUC 0.904 in validation cohort. Improved models indicated that unnecessary FNA and missed cancer rates were lower than those of ACR-TR4 system. Impact: MRI-based models demonstrated outstanding diagnostic performance for distinguishing benign from malignant ACR-TR4 thyroid nodules. Combined model, which utilizing restricted diffusion and reversed halo sign, holds promise for reducing the need for unnecessary FNA, while simultaneously minimizing risk of missed cancers.