Objective To analyze the diagnostic value of artificial intelligence ultrasonic assisted diagnosis system (AI-UADS) combined with contrast-enhanced ultrasound (CEU) for American College of Radiology (ACR) Thyroid Imaging Reporting and Data System (TI-RADS) class 4 nodules. Methods A total of 84 patients with ACR TI-RADS class 4 thyroid nodules admitted to the hospital were enrolled between March 2018 and March 2022. All underwent AI-UADS and CEU examinations. The diagnostic value of the two methods and combined detection for benign and malignant ACR TI-RADS class 4 nodules was compared. Results There were 92 nodules in the 84 patients

including 17 malignant nodules (18.48%). CEU showed that there were 14 malignant nodules

its sensitivity

specificity

accuracy

positive predictive value

negative predictive value and kappa value in the diagnosis of malignant nodules were 82.35%

81.33%

81.52%

50.00%

95.31% and 0.509

respectively. AI-UADS showed that there were 15 malignant nodules

and its sensitivity

specificity

accuracy

positive predictive value

negative predictive value and kappa value were 88.24%

82.67%

83.70%

53.57%

96.88% and 0.567

respectively. The sensitivity

specificity

accuracy

positive predictive value

negative predictive value and kappa value of combined detection were 82.35%

93.33%

91.30%

73.68%

95.89% and 0.724

respectively. The specificity of combined detection was higher than that of CEU and AI-UADS (χ2=0.027

0.044

P<0.05). Conclusion Both AI-UADS and CEU have good diagnostic efficiency for benign and malignant ACR TI-RADS class 4 thyroid nodules

and the combined detection can effectively improve clinical specificity.