ABSTRACT
Objective To assess the value of multimodal ultrasonography for diagnosing thyroid nodules—atypia of undetermined significance (AUS) of thyroid imaging reporting and data system (TI-RADS) categories 3 to 5. Methods A total of 90 AUS thyroid nodules in TI-RADS 3-5 categories from 88 patients underwent conventional ultrasonography, ultrasound elastography, superb microvascular imaging, and multimodal ultrasonography at the same time. With fine needle aspiration biopsy results as the gold standard, the methods were compared in terms of the sensitivity, specificity, accuracy, false positive rate (FPR), false negative rate (FNR), and area under the receiver operating characteristic curve (AUC) for diagnosing thyroid nodules. Results There were no significant differences between patients with benign and those with malignant thyroid nodules in terms of sex, age, and nodule locations (all P > 0.05), but the proportion of thyroid nodules ≤ 1 cm in diameter was significantly higher for malignant thyroid nodules than for benign thyroid nodules (χ2=9.610, P=0.002). Compared with benign nodules, malignant nodules were significantly more frequent to have low-level echoes or very low-level echoes, a blurred margin, a vertical diameter/horizontal diameter ratio of > 1, and microcalcifications or no calcifications (all P < 0.05). An ultrasound elastography score of ≥ 3 and type III vascularity on superb microvascular imaging indicated a higher possibility of malignant thyroid nodules (both P < 0.001). The multivariable logistic regression analysis showed that the size, echogenicity, margin, and vertical diameter/horizontal diameter ratio, and superb microvascular imaging type of thyroid nodules were not significant markers for benign or malignant thyroid nodules (all P > 0.05), while microcalcifications/no calcifications and an ultrasound elastography score of ≥ 3 were independent risk factors for malignant AUS nodules (both P < 0.05). The diagnostic sensitivity, specificity, accuracy, FPR, and FNR of conventional ultrasonography for AUS nodules were 91.30%, 71.40%, 62.70%, 28.60%, and 8.70%, respectively; the values for ultrasound elastography were 85.50%, 66.70%, 52.20%, 33.30%, and 14.50%, respectively; the values for superb microvascular imaging were 66.70%, 76.20%, 42.90%, 23.80%, and 33.30%, respectively; and the values for multimodal ultrasonography were 75.20%, 92.50%, 67.70%, 24.80%, and 7.50%, respectively. For distinguishing between benign and malignant AUS nodules, the AUC values of conventional ultrasonography, ultrasound elastography, superb microvascular imaging, and multimodal ultrasonography were 0.866, 0.745, 0.774, and 0.918, respectively. Conclusion Multimodal ultrasonography shows better diagnostic efficacy for AUS nodules of TI-RADS 3-5 compared with conventional ultrasonography, ultrasound elastography, and superb microvascular imaging, which can facilitate the malignancy risk stratification and management of AUS thyroid nodules.