ABSTRACT
Objective:To construct a quantitatively diagnostic nomogram model by analyzing the clinical information of patients and the features of multi-modality ultrasound images of thyroid lesions, so as to preoperatively predict the malignant probability of suspicious thyroid nodules and provide effective references for clinical decision-making.Methods:A total of 933 patients, 1 121 thyroid nodules of C-TIRADS 3-5 categories, who underwent surgery in the Second Affiliated Hospital of Harbin Medical University from September 1, 2020 to June 10, 2021 were collected. The nodules were randomly divided into training ( n=897) and test groups ( n=224) in 8∶2 ratio. Finally, the diagnostic performance was evaluated by area under the curve (AUC). Results:①After preliminary screening by univariate analysis, multivariate analysis showed that age, echogenicity, orientation, echogenic foci, margin, posterior features, and elastic score were significantly correlated with benign and malignant nodules (all P<0.001), and the difference of halo between benign and malignant nodules was also statistically significant ( P=0.012). ②The AUC of nomogram was up to 0.903(95% CI=0.862-0.944) in the test set, and 0.889(95% CI=0.832-0.946) and 0.960(95% CI=0.925-0.994) in nodules with maximum diameter of ≤10 mm and of >10 mm respectively, which showed high diagnostic performance. Conclusions:The nomogram model could accurately differentiate malignant from benign thyroid nodules preoperatively, with the highest diagnostic performance for the nodules with maximum diameter of >10 mm, and effectively avoid the unnecessary fine-needle biopsy and surgical operation.