| Objective: To retrospectively evaluate the interobserver agreement of radiologists who used the ACR 2013 Breast Imaging Reporting and Data System(BI-RADS)lexicon to characterize and categorize ultrasonographic(US)features of breast masses.Methods: From August 2019 to October 2020,138 patients with breast masses were selected in the Department of breast surgery,the First Affiliated Hospital of Wannan Medical College,in which 152 masses were detected.Preoperative ultrasound examination was performed and postoperative pathological results were obtained.Siemens S2000 color Doppler ultrasound diagnostic instrument was used to select the breast examination mode.The type of background gland and the location,shape and orientation of the lesion were recorded.The echo pattern,margin,posterior echo and calcification were observed.The ipsilateral axillary lymph nodes were examined for swelling or abnormal echo,and changes of surrounding tissues.Then adjust to CDFI mode and observe the blood flow signal of the lesions.The four groups of observers used double-blind method: they independently evaluated and classified the collected and stored images without knowing the clinical data,other imaging results and pathological results of the patients.The shape(regular / irregular),orientation(parallel/non-parallel),echo pattern(hypoechoic/ isoechoic/complex),margin(circumscribed /not circumscribed/indistinct/angular/ spiculated/microlobulated),posterior echo(no change/attenuation/enhancement / mixed change),calcification(microcalcification/large calcification/no calcification in the mass)and final classification(3/4A/4B/4C/5)were recorded according to the 2013 edition of BI-RADS dictionary.The ROC curve was drawn with pathological results as the gold standard.The best cut-off value for differential diagnosis of benign and malignant breast masses was determined,and the diagnostic efficiency of each observer was calculated;ROC curve was used to evaluate the difference of accuracy in the diagnosis of breast malignant lesions among the four groups;pair χ~2 test was used to evaluate the difference of sensitivity and specificity in the diagnosis of breast malignant lesions among the four groups of observers;χ~2 test was used to evaluate the difference between the actual positive predictive value(PPV)and the theoretical PPV of the four groups;Fliess,kappa test were used to evaluate the variability of the choice and classification of ultrasound descriptors among the four groups.Results: 1.The ROC curve was drawn with pathological results as the gold standard.The area under diagnostic curve(AUC)of the four groups were 0.716,0.775,0.821 and 0.883 respectively.The Youden index was 0.356,0.552,0.495 and 0.739 respectively.The cut-off points of benign and malignant lesions were between 4A and4 B.The diagnostic accuracy of the four groups was 68.42%,75.66%,78.29% and87.50% respectively;The sensitivity was 72.94%,82.35%,83.53% and 91.76%respectively;The specificity was 62.69%,67.16%,71.64% and 82.09% respectively.There was a significant difference in the accuracy between observer A and C,observer A and D,observer B and D(P<0.05);There was no significant difference between observer A and B,observer B and C,observer C and D(P> 0.05,P> 0.05,P=0.05).There were significant differences in sensitivity and specificity between observer A and observer D(P<0.05).2.The PPV of 4A diagnosed by four groups of observers were 35.48%,26.78%,23.33%and 11.66% respectively.The PPV of 4B were 57.14%,72.13%,69.64% and 85.29%,respectively.The PPV of 4C were 80.95%,80.77%,92% and 83.33% respectively.The overall PPV of the four groups were 54.67%,55.94%,53.90% and 52.21% respectively.The PPV of 5 were 80%,100%,100% and 100% respectively.3.The four groups of observers had good consistency in judging the shape of the breast mass(k=0.61).The consistency of orientation,echo type,posterior echo pattern and calcification was moderate(k=0.54,k=0.55,k=0.43,k=0.45),The consistency of margin was fair(k=0.21).The overall consistency of the evaluation classification was fair(k=0.27).The inter-observer consistency of 3,4a,4b,4C and 5 categories was moderate(k=0.41),moderate(k=0.44),poor(k=0.13),fair(k=0.22)and fair(k =0.21),respectively;The consistency between benign(category 3,4A)and malignant(category 4B,4C,5)was moderate(k=0.47).Conclusion: 1.The 2013 ACR BI-RADS has important clinical value in the diagnosis of benign and malignant breast lesions.2.The interobserver consistency of some BI-RADS lexicon is general.It shows that no matter the doctor’s working experience,the application of some lexicons still needs further training.At the same time,improving the ultrasound dictionary may reduce the interobserver variability of ultrasound diagnosis of breast lesions.3.The consistency of The 2013 ACR BI-RADS in the diagnosis of breast masses is fair among observers,especially for the sub classification of four types of lesions.It is recommended that inexperienced doctors in primary hospitals should not subdivide in clinical practice at present. |
PDF Full Text Request