Prediction model based on MRI morphological features for distinguishing benign and malignant thyroid nodules

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Thyroid nodules require accurate preoperative diagnosis to avoid unnecessary biopsies.
• The Thyroid Imaging Reporting and Data System (TI-RADS) has limitations in specificity for distinguishing benign from malignant nodules.
• Improving diagnostic specificity is crucial for patient management and reducing healthcare costs.

Purpose of the Study:

• To develop and validate a predictive model using MRI morphological features for improved preoperative diagnosis of thyroid nodules.
• To enhance the specificity of thyroid nodule diagnosis compared to existing methods like TI-RADS.
• To reduce the rate of unnecessary thyroid biopsies.

Main Methods:

• Retrospective analysis of 825 pathologically confirmed thyroid nodules.
• Development and validation of a predictive model and nomogram incorporating MRI features using logistic regression.
• Evaluation of diagnostic efficacy, discrimination, calibration, and comparison with TI-RADS using AUC and NRI.

Main Results:

• MRI morphological features such as restricted diffusion and reversed halo sign were independent predictors of malignancy.
• The developed nomogram demonstrated high discrimination and calibration in both training (AUC=0.972) and validation (AUC=0.968) cohorts.
• The MRI-based model showed superior accuracy (0.947) and specificity (93.4%) compared to TI-RADS, with enhanced diagnostic performance (NRI>0).

Conclusions:

• The MRI-based predictive model offers superior diagnostic efficacy for benign versus malignant thyroid nodules compared to TI-RADS.
• The model achieves high sensitivity and specificity, capable of reducing overdiagnosis and unnecessary biopsies.
• This approach provides a more accurate tool for preoperative risk stratification of thyroid nodules.