Clinical Application Value Of Artificial Intelligence Revised American College Of Radiology Thyroid Imaging Reporting And Data System In Thyroid Nodules

ObjectiveTo explore the clinical application value of artificial intelligence revised ACR TI-RADS(AI TI-RADS)in thyroid nodules,to evaluate its diagnostic value and unnecessary fine-needle aspiration(FNA)of thyroid nodules,and compare it with ACR TI-RADS proposed by American College of Radiology in 2017 and ATA guideline proposed by American Thyroid Association in 2015.MethodsA total of 1925 thyroid nodules ultrasonic images were selected from 1269 patients who underwent thyroid ultrasound examination and ultrasound guided fine needle aspiration in the Department of Ultrasound of our hospital from January 2018 to October 2019.The ultrasound pictures of all nodules were analyzed by two doctors with more than 5 years of superficial ultrasound diagnosis experience,and all nodules were classified according to the classification criteria of AI TI-RADS,2017 ACR TI-RADS and 2015 ATA guideline respectively.The ROC curves were plotted to compare the diagnostic value and unnecessary fine-needle aspiration rate of the three risk stratification systems.Results1.The area under the ROC curve(AUC)of AI TI-RADS was 0.899,the sensitivity was 83.9%,and the specificity was 87.6%.The AUC,sensitivity and specificity of ACR TI-RADS were 0.886,80.3%and 86.5%,respectively.The AUC,sensitivity and specificity of ATA guideline were 0.863,90.9%and 78.6%,respectively.There were statistically significant differences in AUC among the three risk stratification systems,of which AI TI-RADS was the highest and ATA guideline was the lowest(P<0.05).The sensitivity differences among the three systems were statistically significant,of which ATA guideline was the highest,ACR TI-RADS was the lowest(P<0.05).There was no significant difference in specificity among the three systems(P>0.05).2.In the<10mm group,there was no statistically significant difference in AUC of the three risk stratification systems(P>0.05).In the≥10mm group,there were statistically significant differences in AUC of the three systems,of which AI TI-RADS was the highest(P<0.05).3.The unnecessary fine-needle aspiration rates of AI TI-RADS,ACR TI-RADS and ATA guideline were 9.1%,12.5%and 22.3%,respectively,and there were statistically significant differences of the three systems,of which AI TI-RADS was the lowest(P<0.05).There were no significant difference in the malignant detection rate among the three systems(P>0.05).Conclusion1.The diagnostic value of AI TI-RADS for thyroid nodules is higher than that of ACR TI-RADS and ATA guideline.The ATA guideline has the highest sensitivity.2.The diagnostic value of the three risk stratification systems in thyroid nodules more than 1cm are better than that for nodules less than 1cm.3.AI TI-RADS can significantly reduce the unnecessary fine-needle aspiration rate for benign nodules,and without reducing the detection rate for malignant nodules.