气道内超声与临床特征结合在肺外周病变良恶性概率中的诊断探讨

The diagnostic value of the endobronchial ultrasonography model to estimate the probability of malignancy in patients with peripheral pulmonary lesions

目的由患者肺外周病变的气道内超声图像及临床特征建立恶性概率估算模型,并探讨该模型临床诊断价值。方法收集我院2010年9月1日-2015年1月30日共150例肺外周病变患者入选本研究,其中135例获得明确的病理诊断结果,采集其气道内超声图像,分析、记录超声图像内部结构特征。结果经二分类多因素logistic回归分析,最终2个临床因素及5个超声特征纳入恶性概率估算模型,P=l/[l+e-(-2.986+1.993吸烟史+2.293CEA+1.552边界+l.616异质-2.011支气管充气影+1.748无回声)];绘制ROC曲线,计算曲线下面积(AUC)为0.926(95%CI:0.883-0.969),以P≥0.62为诊断界值,得出此方程的灵敏度为69/77=89.6%,特异度为42/58=72.4%,准确率为82.2%。结论 1气道内超声对肺外周病变具有较高的良恶性诊断价值,是一种有效的影像学检查方法。2建立肺外周病变恶性概率估算模型,超声图像与临床特征相结合,更能提高对周围型肺癌诊断的准确率。

Objective To use the endobronchial ultrasonographic features of peripheral lung cancer with clinical data and establish a model to estimate the probability of malignancy in peripheral pulmonary lesions（PPLs）, and evaluate its clinical diagnostic value. Methods Between September 1st 2010 and May 30th 2015, endobronchial ultrasonography （EBUS） were performed in 150 patients with peripheral pulmonary lesions. At last, the EBUS images of the 135 cases were en- rolled, who had a definite pathological diagnosis. Analyse and record the characters of endobronchial images. Results According to the result of binary multivariable logistic regression analysis, we concluded two clinical data and five distinct EBUS image patterns which contribute to predicting the presence of malignancy. The equation of malignancy probability for any patient was： P= 1/I＋e^ （2.9861.993smorking＋2.293CEA＋1.552borderliine＋1.616heterogeneity 2.011air bmoncbortem＋1.745anechoic area）; the area under the ROC curve （AUC）was 0. 926（95% CI..0. 883-0. 969） ,with P≥0. 62 for the diagnosis of community values,it is concluded that the sensitivity of this equation was 69/77=89.6% and specificity was 42/58=72.4% ,the accuracy rate was 82.2%. Conclusion （1） Endobronchial ultrasonography in diagnosis of benign and malignant pulmonary peripheral lesions have a high value, and is an effective method of imaging. （2）Combining image results with the clinical data to establish binary logistic models to predict the malignant probability could raise the accuracy.