The application of deep learning combined with C-TIRADS in the risk stratification management of thyroid nodules classified as 4a
Objective To explore the application of deep learning combined with the Chinese thyroid imaging reporting and data system(C-TIRADS)in the risk stratification management of thyroid nodules classified as 4a.Methods A total of 179 patients with thyroid nodules treated at Shaanxi Provincial People's Hospital from December 2018 to October 2022 were included,divided into benign(n=76)and malignant groups(n=103)based on pathological results.All patients underwent ultrasound examination and were diagnosed using C-TIRADS guidelines and deep learning.Multiple factor Logistic regression analysis was used to obtain independent predictive indicators;the accuracy of predictive variables was assessed using the ROC curve.Results Multiple factor Logistic analysis showed that the structural,directional,edge,echo,focal strong echo,and age characteristics of thyroid nodule images are independent indicators for predicting the nature of thyroid nodules(P<0.05).With pathological results as the gold standard,the complete consistency rate of deep learning combined with C-TIRADS with pathological results was 96.65%,and the Kappa value was 0.932,indicating good consistency;the consistency rate,specificity,and positive predictive value of the combined diagnosis for thyroid nodules classified as 4a were significantly higher than those of deep learning and C-TIRADS(P<0.05).The sensitivity and negative predictive value of the combined diagnosis for the disease were higher than those of deep learning(P<0.001),but the difference with C-TIRADS was not statistically significant(P>0.05).ROC curve analysis showed that the AUCs for C-TIRADS,deep learning,and combined diagnosis were 0.873,0.819,and 0.967,respectively;compared with Az=0.5,the differences were all statistically significant(P<0.001).Conclusion C-TIRADS has a high sensitivity in the risk stratification management of thyroid nodules classified as 4a,and combined with deep learning for auxiliary diagnosis,it can accurately distinguish between benign and malignant thyroid nodules,with high diagnostic efficacy.
thyroid nodulesChinese thyroid imaging reporting and data systemdeep learningmalignant riskguidelinesdiagnosis
万方数据
维普
