Purpose
We aimed to diagnose the benign or malignant of large thyroid nodules by quantitative analysis of diffusion-weighted imaging (DWI).
Methods
82 thyroid nodules were investigated retrospectively and divided them into benign (n = 62) and malignant groups (n = 20). DWI data were acquired, and apparent diffusion coefficients (ADCs) were calculated. Univariate and multivariate logistic regression were conducted to identify independent predictors, and develop a prediction model. We performed receiver operating characteristic (ROC) analysis to determine the optimal threshold of risk factors, and constructed combined threshold models. Diagnostic performance metrics, including area under the ROC curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the ratio of unnecessary fine-needle aspiration biopsy (UFNAB) of all models were calculated and compared with the American College of Radiology Thyroid Imaging Reporting and Data System (ACR-TIRADS) result.
Results
Two independent predictors of malignant nodules were identified by multivariate analysis: DWI signal intensity ratio (DWISIR, P = 0.007) and minimum ADC (ADCmin, P < 0.001). At a cutoff value of 0.198, the multivariate prediction model had an AUC of 0.946. The combined threshold model of DWISIR and ADCmin had the highest specificity up to 100% and the lowest UFNAB rate of 0%.
Conclusion
Quantitative DWI demonstrated favorable malignant thyroid nodule diagnostic efficacy. Combined thresholds of DWISIR and ADCmin greatly reduced the UFNAB.