人工智能辅助评分联合剪切波弹性成像在乳腺结节鉴别诊断中的应用

Application of artificial intelligence assisted scoring combined with shear wave elastography in the differential diagnosis of breast nodules

目的:探讨人工智能辅助评分(artificial intelligence assisted scoring,AIAS)联合剪切波弹性成像(shear wave elastography,SWE)在乳腺结节鉴别诊断中的价值。方法:纳入90例患者共91个乳腺结节进行AI和SWE检查,记录结节的AIAS和弹性模量最大值(E_(max))、平均值(E_(mean))及标准差(Esd)。以AIAS和最佳弹性模量值为自变量,病理结果为因变量作二元Logistic回归分析,计算联合诊断预测因子并建立回归方程。以病理结果为金标准,绘制所有参数的受试者工作特征(receiver operating characteristic,ROC)曲线,计算并比较其曲线下面积(area under curve,AUC),获得最佳诊断临界值和净重新分类指数(NRI),评估AIAS、SWE及二者联合诊断乳腺结节良恶性的效能差异。结果:乳腺恶性病灶的AIAS、E_(max)、E_(mean)、Esd值均较良性增大(P均<0.001)。AIAS截断值≥0.6分时,AI诊断乳腺癌的灵敏度、特异度、准确率较高。E_(max)、E_(mean)、Esd的AUC分别为0.913、0.836、0.903(P均<0.05),其中E_(max)曲线下面积最大,截断值为87.4 kPa时诊断效能最佳。AIAS、E_(max)及联合诊断的AUC分别为0.911、0.913、0.943(P均<0.05)。根据NRI结果,联合诊断对AIAS和E_(max)均为正改善(P均<0.05)。结论:AIAS和SWE对乳腺结节均具有较好的鉴别诊断能力。与单独应用相比,二者联合诊断可提高乳腺恶性肿瘤的诊断效能。

Objective:To investigate the value of artificial intelligence-assisted scoring combined with shear wave elastography in the differential diagnosis of breast nodules.Methods:A total of 91 breast nodules were included in 90 patients for AI and SWE examination,and the AIAS and elastic modulus maximum(E_(max)),mean(E_(mean))and standard deviation(Esd)of the nodules were recorded.A binary logistic regression analysis was performed with AIAS and optimal elastic modulus values as independent variables and pathological findings as dependent variables to calculate joint diagnostic predictors and establish regression equations.The receiver operating characteristic(ROC)curves were plotted for all parameters using the pathological results as the gold standard.The area under curve(AUC)was calculated and compared to obtain the optimal diagnostic threshold and net reclassification index(NRI)to evaluate the difference in the efficacy of AIAS,SWE and their combined diagnosis of benign and malignant breast nodules.Results:AIAS,E_(max),E_(mean),Esd of malignant breast lesions were all increased compared to benign(all P<0.001).When the AIAS cut-off value is≥0.6 points,AI has higher sensitivity,specificity and accuracy in the diagnosis of breast cancer.The AUCs of E_(max),E_(mean) and Esd were 0.913,0.836 and 0.903 respectively(all P<0.05).The AUC of E_(max)was the largest,and the diagnostic efficiency was the best when the cut-off value was 87.4 kPa.The AUCs for AIAS,E_(max)and combined diagnosis were 0.911,0.913 and 0.943 respectively(all P<0.05).According to the NRI results,the combined diagnosis was positively improved for both AIAS and E_(max)(all P<0.05).Conclusion:Both AIAS and SWE have good differential diagnostic ability for breast nodules.The combined diagnosis of both can improve the diagnostic efficacy of breast malignancies compared with their application alone.