肺亚实性结节CT分类评估和定量测量方法的比较及定量特征对肺腺癌病理等级预测价值的研究

发布时间：2018-06-19 19:26

本文选题：肺腺癌 + 肺亚实性结节　；参考：《第二军医大学》2017年硕士论文

【摘要】：第一部分肺亚实性结节分类评估和定量测量方法的可重复性和准确性的研究【目的】对比不同窗宽窗位条件下不同维度的肺亚实性结节(SSN)分类评估和定量测量方法的可重复性和准确性,并探讨实性成分大小与病理等级的相关性。【方法】回顾性分析157例患者肺内手术证实为腺癌的159个SSN的术前胸部HRCT图像及病理资料。根据病理结果将SSN分为无实性成分组和有实性成分组。观察者1和观察者2分别在肺窗、纵隔窗及-300阈值半自动分割法条件下对所有SSN进行分类评估,分类结果采用Cohen’s Kappa检验分析分类评估的观察者间一致性。两位观察者意见不一的SSN由观察者3判定,最终分类结果采用配对卡方检验和ROC曲线分析分类评估的准确性。观察者1使用联影后处理工作站测量所有SSN的肺窗结节一维长径(1D-WNLW),肺窗结节二维长径(2D-WNLW),肺窗实性一维长径(1D-SCLW),肺窗实性二维长径(2D-SCLW),纵隔窗实性一维长径(1D-SCMW),纵隔窗实性二维长径(2D-SCMW),肺窗结节体积(3D-WNLW),肺窗实性体积(3D-SCLW),纵隔窗实性体积(3D-SCMW)和-300阈值实性体积(3D-SCT),观察者2随机抽取50个SSN并重复上述测量。两位观察者测量结果采用组内相关系数分析定量测量的观察者间一致性。观察者1的测量结果使用Spearman等级相关分析检验定量特征与病理等级的相关性。P0.05时差异有统计学意义。【结果】有实性成分组共54个SSN,包括32个AAH和22个AIS;无实性成分组共105个SSN,包括47个MIA和58个IAC。不同窗宽窗位及不同维度条件下SSN分类评估和定量测量方法的可重复性均较高(0.71k0.9,0.71ICC)。使用-300阈值半自动分割法进行分类评估和定量测量的可重复性最高(Kappa=0.831,ICC=0.983),且分类评估的准确性最高(Sens=85%,Spec=61%,PPV=81%,NPV=67%,AUC=0.750)。所有定量特征均与病理等级存在中度正相关关系(0.4≤r0.7)。【结论】在不同窗宽窗位及不同维度条件下SSN的分类评估及定量测量的诸多方法中,-300阈值半自动分割法的可重复性和准确性最高。SSN及其实性成分的大小与病理等级呈正相关关系。第二部分第一节肺亚实性结节及其实性成分不同维度大小对病理等级的预测价值【目的】探讨不同窗宽窗位下肺亚实性结节(SSN)及其实性成分不同维度的大小对病理等级的预测价值。【方法】回顾性分析125例患者肺内病理为肺腺癌的127个SSN的术前HRCT图像及病理资料。根据术后5年生存率的不同,将所有SSN分为两组,A组包含AIS和MIA;B组包含IAC。由一名有5年影像诊断经验的医师使用联影后处理工作站测量所有SSN的肺窗实性一维长径(1D-SCLW)、肺窗实性一维长径(2D-SCLW)、纵隔窗实性一维长径(1D-SCMW)、纵隔窗实性二维长径(2D-SCMW)、肺窗结节一维长径(1D-WNLW)、肺窗结节二维长径(2D-WNLW)和-300阈值实性体积(3D-SCT)。使用Mann-Whitney U检验进行两组定量特征的差异性分析。采用ROC曲线检验定量特征对病理等级的诊断效能。对所有定量特征进行单因素Logistic回归分析,所得有统计学意义的定量特征纳入多因素Logistic回归进行分析,得到病理等级的独立预测因素,P0.05时差异有统计学意义。【结果】A组共69个SSN,包含22个AIS和47个MIA;B组包含58个IAC。B组的1D-SCLW、2D-SCLW、1D-SCMW、2D-SCMW、1D-WNLW、2D-WNLW和3D-SCT显著大于A组(P0.0001)。通过ROC曲线分析得出,在以上7种CT定量特征中,3D-SCT对病理等级的诊断效能最高(AUC=0.887,敏感度:81%,特异度:93%);1D-SCLW、2D-SCLW、1D-SCMW、2D-SCMW、1D-WNLW、2D-WNLW和3D-SCT的最优阈值分别为17.50mm、14.75 mm、9.50 mm、7.75 mm、0.50 mm、1.25 mm和139.00 mm3。多因素logistic回归分析结果表明,3D-SCT是SSN病理等级的独立预测因素(OR=4.978,95%CI=1.430~17.331,P=0.012)。当3D-SCT≥139.00 mm3时,高度提示IAC(AUC=0.887,敏感度:81%,特异度:93%)。【结论】在不同窗宽窗位下SSN及其实性成分不同维度的大小中,-300HU阈值实性成分体积是病理等级的独立预测因素,其诊断效能优于SSN及其实性成分的一维和二维长径。第二部分第二节肺亚实性结节实性成分体积及其占比对病理等级的预测价值【目的】对比不同窗宽窗位下肺亚实性结节(SSN)体积和实性成分体积及其占比对病理等级的预测价值。【方法】回顾性分析我院125例患者肺内手术证实为腺癌的127个SSN的术前HRCT图像及病理资料。根据术后5年生存率的不同,将所有SSN分为两组,A组包含AIS和MIA;B组包含IAC。由一名有5年影像诊断经验的医师使用联影后处理工作站测量所有SSN的肺窗结节体积(3D-WNLW)、肺窗实性体积(3D-SCLW)、纵隔窗实性体积(3D-SCMW)和-300阈值实性体积(3D-SCT),并计算肺窗实性占比(P-SCLW);纵隔窗实性占比(P-SCMW);-300阈值实性占比(P-SCT)。使用组内相关系数(ICC)检验定量测量的观察者间一致性。采用Mann-Whitney U检验比较两组间定量特征的差异。采用单因素Logistic回归分析对所有定量特征进行初筛,所得有统计学意义的定量特征纳入多因素Logistic回归进行分析,得到病理等级的独立预测因素。采用ROC曲线评价独立预测因素的诊断效能。P0.05时差异有统计学意义。【结果】A组共69个SSN,包含22个AIS和47个MIA;B组包含58个IAC。B组的3D-WNLW、3D-SCLW、3D-SCMW、3D-SCT、P-SCLW、P-SCMW和P-SCT显著大于A组(P0.0001)。单因素Logistic回归分析显示,3D-WNLW、3D-SCLW、3D-SCMW、3D-SCT、P-SCLW、P-SCMW和P-SCT均有统计学意义(P0.0001),多因素Logistic回归分析发现,仅P-SCT是SSN病理等级的独立预测因素(OR=1.093,95%CI:1.047~1.141,P0.0001)。当P-SCT≥6.00%时,高度提示IAC(AUC=0.846,敏感度:79%,特异度:75%)。【结论】在不同窗宽窗位下SSN体积和实性成分体积及其占比中,P-SCT是SSN病理等级的独立预测因素,能够有效区分IAC和AIS-MIA,与SSN及其实性成分的体积相比,能够为手术方式的选择提供更有价值的参考依据。
[Abstract]:The first part of the study on the reproducibility and accuracy of the taxonomy and quantitative measurements of the pulmonary nodules (objective) to compare the repeatability and accuracy of the classification and quantitative measurements of pulmonary subsolid nodules (SSN) with different dimensions under different window wide window positions, and to explore the correlation between the size of the solid components and the pathological grade. [Methods] the preoperative chest HRCT images and pathological data of 157 patients with adenocarcinoma confirmed by intrapulmonary surgery were retrospectively analyzed. According to the pathological results, the SSN was divided into the non solid component group and the solid component group. The observer 1 and the observer 2 were divided into all SSN under the condition of the lung window, the mediastinum window and the semi automatic segmentation of the -300 threshold. Class evaluation, Cohen 's Kappa test was used to analyze the inter observer consistency of the classification assessment. The two observer disagreed SSN was determined by the observer 3. The final classification results were evaluated by the paired chi square test and the ROC curve analysis. The observer 1 measured all the SSN lung window nodes using the combined post processing workstation. One dimension length diameter (1D-WNLW), two-dimensional long diameter (2D-WNLW) of pulmonary window nodules, solid one dimension diameter (1D-SCLW), solid two-dimensional long diameter (2D-SCLW) of the window, solid dimension of mediastinal window (1D-SCMW), solid two-dimensional long diameter (2D-SCMW) in the mediastinum window, volume of pulmonary window (3D-WNLW), real volume of lung window (3D-SCLW), solid volume of mediastinal window (3D-SCMW) and -300 threshold Value real volume (3D-SCT), the observer 2 randomly selected 50 SSN and repeated the above measurements. The two observer results were measured by intra group correlation coefficient, and the consistency between the observers was measured by the intra group correlation coefficient. The observer 1 measured the correlation between the quantitative characteristics and the pathological grade by Spearman correlation analysis. There was a statistical difference between the observer and the pathological grade. [results] a total of 54 SSN, including 32 AAH and 22 AIS, 105 SSN, including 47 MIA and 58 IAC. windows with different window width and different dimensions, SSN classification evaluation and quantitative measurement method with higher repeatability (0.71k0.9,0.71ICC). Classification using -300 threshold semi-automatic segmentation method is used for classification. The highest repeatability (Kappa=0.831, ICC=0.983) of evaluation and quantitative measurement (Kappa=0.831, ICC=0.983) was the highest (Sens=85%, Spec=61%, PPV=81%, NPV=67%, AUC=0.750). All quantitative characteristics had moderate positive correlation with pathological grades (0.4 < < r0.7). [Conclusion] the classification evaluation of SSN under different window width and different dimensions and Among the many methods of quantitative measurement, the repeatability and accuracy of the -300 threshold semi-automatic segmentation method is the highest.SSN and the size of its actual components is positively correlated with the pathological grade. The second part of the first section of the pulmonary nodules and the different dimensions of its actual components to the pathological grade of the premeasured value [Objective] to discuss the different window wide window position The predictive value of the size of the lower pulmonary subsolid nodules (SSN) and the size of their real components to the pathological grades. [Methods] a retrospective analysis of the preoperative HRCT images and pathological data of 127 SSN in lung adenocarcinoma in 125 patients was reviewed. According to the 5 year survival rate, the SSN was divided into two groups, the A group contained AIS and MIA, and B group contained IAC. A physician with 5 years of imaging diagnostic experience measured the solid one dimension length (1D-SCLW) of all SSN lung windows, solid one dimension (2D-SCLW), solid one dimension length (1D-SCMW) of the mediastinal window, solid two-dimensional long diameter (2D-SCMW) of the mediastinum window, one dimension diameter (1D-WNLW) of the window of the lung window (1D-WNLW), and the two-dimensional length of the pulmonary window nodules (2D-WNL). W) and -300 threshold real volume (3D-SCT). Use the Mann-Whitney U test to analyze the difference between the two groups of quantitative characteristics. Use the ROC curve to test the diagnostic efficiency of the quantitative characteristics for the pathological grade. All quantitative features are analyzed by single factor Logistic regression, and the quantitative characteristics of statistical meaning are included in the multiple factor Logistic regression. There were 69 SSN of SSN in group A, including 22 AIS and 47 MIA, and B group contained 1D-SCLW of IAC.B group, 2D-SCLW, 1D-SCMW, 2D-SCMW. 3D-SCT was most effective in the diagnosis of pathological grade (AUC=0.887, sensitivity: 81%, specificity: 93%); the optimal threshold for 1D-SCLW, 2D-SCLW, 1D-SCMW, 2D-SCMW, 1D-WNLW, 2D-WNLW and 3D-SCT were 17.50mm, 14.75 mm, 9.50 mm, 0.50, 1.25, 1.25 and 139. The independent predictor (OR=4.978,95%CI=1.430~17.331, P=0.012). When 3D-SCT is more than 139 mm3, IAC (AUC=0.887, sensitivity: 81%, specificity: 93%). [Conclusion] the volume of -300HU threshold value is an independent predictor of pathological grade in the size of different dimensions of SSN and its actual components at different window wide window positions. One and two dimensional length of effectiveness superior to SSN and its actual components. The volume of solid components in the second part of the second nodular pulmonary nodules and its predictive value to the pathological grade [Objective] to compare the volume of SSN volume and the volume of solid components under different window wide window positions and the predictive value of its proportion to the pathological grade. A retrospective analysis of the preoperative HRCT images and pathological data of 125 patients with adenocarcinoma confirmed by intrapulmonary surgery in 125 patients. According to the 5 year survival rate, all SSN were divided into two groups, the A group included AIS and MIA, and the B group included a 5 year imaging doctor with a 5 year postprocessing workstation to measure all the lungs of all SSN. Window nodule volume (3D-WNLW), pulmonary window volume (3D-SCLW), mediastinal real volume (3D-SCMW) and -300 threshold real volume (3D-SCT), and calculated the ratio of real lung window (P-SCLW); mediastinal window real occupying ratio (P-SCMW); -300 threshold real ratio (P-SCT). The consistency of quantitative measurement with intra group correlation coefficient (ICC) was used. Mann-Whi Tney U test compared the differences in quantitative characteristics between the two groups. Using single factor Logistic regression analysis, all quantitative characteristics were screened. The quantitative characteristics were statistically analyzed by multiple factor Logistic regression, and the independent predictors of pathological grade were obtained. The diagnostic effectiveness of independent predictors by ROC curve was used to evaluate the diagnostic efficiency.P0.05 [results] there were 69 SSN in group A, including 22 AIS and 47 MIA, and B group including 3D-WNLW, 3D-SCLW, 3D-SCMW, 3D-SCT, P-SCLW. 0001), multiple factor Logistic regression analysis found that only P-SCT was an independent predictor of SSN pathological grade (OR=1.093,95%CI:1.047~1.141, P0.0001). When P-SCT was more than 6%, IAC (AUC=0.846, sensitivity: 79%, specificity: 75%). [Conclusion] P-SCT is SSN disease in the volume and proportion of SSN and real components at different window wide window positions. The independent predictors of the grade can effectively distinguish between IAC and AIS-MIA, which can provide a more valuable reference for the choice of surgical methods compared with the volume of SSN and its actual components.
【学位授予单位】：第二军医大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R734.2;R730.44

【参考文献】