探讨灰阶超声（GSU）、超声造影（CEUS）及超声引导下经皮肺穿刺在周围型肺结核诊断中的应用价值，建立周围型肺结核的Logistic回归诊断模型。

回顾性分析2018年1月至2019年12月在浙江大学医学院附属杭州市胸科医院就诊并经病理及Gene X-pert MTB/RIF检查证实的周围型肺结节患者61例，共61个结节。将患者分为结核病组43例，非结核病组18例。所有患者均行GSU、CEUS、超声引导经皮肺穿刺术，记录病灶形态、内部回声、支气管征、造影到达时间（AT）、造影剂到达病灶与邻近肺组织时间差、灌注模式、增强程度等超声特征及活检组织完整性，并进行单因素分析。然后将各项参数纳入Logistic多因素回归分析，建立周围型肺结核的Logistic回归诊断模型，并绘制ROC曲线，计算该模型预测周围型肺结核的诊断效能。

结核病组与非结核病组的病灶形态、造影剂AT、造影剂到达时间差、灌注模式参数比较，差异均有统计学意义（χ²=6.811、5.770、5.960、5.728，P均＜0.05）。二分类Logistic回归分析显示，年龄、结节形态、造影剂到达时间差、灌注模式4个变量进入Logistic回归模型，回归方程式为：Logit（P）=-3.565+1.868X₂+2.469X₃+1.734X₇+2.650X₈。回归模型预测周围型肺结核的ROC曲线下面积为0.911，以Logit（P）≥0.50为截断值，其预测准确性为87.3%，敏感度、特异度、阳性预测值、阴性预测值分别为94.6%、72.2%、87.5%、86.7%。

以周围型肺结节患者年龄、病灶形态、造影剂到达病灶与邻近肺组织时间差、灌注模式为纳入因素的Logistic回归模型有助于诊断周围型肺结核。

To explore the application of gray-scale ultrasound (GSU), contrast-enhanced ultrasound (CEUS), and ultrasound-guided percutaneous lung puncture in the diagnosis of peripheral pulmonary tuberculosis, and establish a Logistic regression model for the evaluation of peripheral pulmonary tuberculosis.

A retrospective analysis was performed on 61 patients with 61 peripheral pulmonary nodules confirmed by pathology and Gene X-pert MTB/RIF examinations who were admitted to the Affiliated Hangzhou Chest Hospital,Zhejiang University School of Medicine from January 2018 to December 2019. The patients were divided into either a tuberculosis group (43 cases)or anon-tuberculous group (18 cases). All patients underwent GSU, CEUS, and ultrasound-guided percutaneous lung puncture, and the lesion morphology, internal echo, bronchial signs, contrast arrival time (AT), time difference between arrival of the contrast agent to the lesion and adjacent lung tissue, perfusion pattern, enhancement degree, biopsy tissue integrity, and other ultrasound characteristics were recorded to perform univariate analysis. Parameters with significance in the univariate analysis were then incorporated into Logistic multivariate regression analysis to establish a Logistic regression diagnosis model. ROC curve analysis was performed tocalculate the performance of this model to predict peripheral tuberculosis.

Univariate analysis showed that lesion morphology, AT, time difference between arrival of the contrast agent to the lesion and adjacent lung tissue, and perfusion pattern differed significantly between the tuberculosis group and the non-tuberculosis group (χ²=6.811,5.770,5.960,5.728,P＜0.05). Two-category logistic regression analysis showed that four variables, namely, age, nodule morphology, contrast medium arrival time difference, and perfusion pattern, entered the Logistic regression model. The regression equation is Logit (P)=-3.565+1.868X₂+2.469X₃+1.734X₇+2.650X₈. The area under the ROC curve of the regression model for predicting peripheral pulmonary tuberculosis was 0.911. With Logit(P)≥0.50 as the cutoff value, its prediction accuracy, sensitivity, specificity, positive predictive value, and negative predictive value were 87.3%, 94.6%, 72.2%, 87.5%, and 86.7%, respectively.

The Logistic regression model, which incorporates the age of patients with peripheral pulmonary nodules, the shape of the lesion, the time difference between the arrival of the contrast agent to the lesion and the adjacent lung tissue, and the perfusion pattern, is helpful for the diagnosis of peripheral pulmonary tuberculosis.