Logistic回归分析高分别率CT的纹理特征对孤立性肺结节诊断价值
|
摘要
目的通过对孤立性肺结节CT图像纹理参数提取进行多变量分析,建立用于早期肺癌诊断的数据模型。方法选取泸州市中医医院72例经手术或穿刺活检病理证实的肺结节CT影像资料,其中恶性结节42例,良性结节30例。分析肺结节的形态学特征(大小、边界、边缘、有无分叶征、毛刺征、血管集束征、胸膜牵拉征、空气支气管征、空泡征、空洞),用FireVoxel软件自动提取肺结节CT图像的8种纹理参数,包括平均值、方差、偏度、峰度、能量、自相关、对比度及熵值。比较良性结节组和恶性结节组形态学特征及纹理参数的差异,以具有统计学差异的参数为自变量,进行多参数Logistic回归分析。构建ROC曲线分析Logistic回归模型的诊断效能。结果多因素Logistic回归分析筛选出边界光整(OR=6.894,P <0.05)、毛刺征(OR=1.542,P <0.05)、分叶征(OR=3.846,P <0.01)、胸膜牵拉征(OR=8.467,P <0.001)、能量(OR=8.972,P <0.01)及熵(OR=4.578,P <0.001)为恶性肺结节患者的独立预测因子。根据独立预测因子建立的ROC曲线下面积为0.894,敏感度及特异度分别为93.43%和84.18%。结论 CT图像纹理分析结合影像学特征对早期肺癌有一定的预测价值。
Objective Base on the texture parameters of the solitary pulmonary nodule CT image,a multivariate analysis was performed to establish a data model for predicting early lung cancer.Methods Seventy-two cases of pulmonary nodules diagnosed by surgery or biopsy were selected in this study.There were42 cases of malignant nodules and 30 cases of benign nodules.The morphological characteristics of pulmonary nodules were analyzed,including size,border,margin,lobulation sign,burr sign,the vascular bundle sign,pleural pull sign,air bronchogram,vacuole sign,empty hole and etc;The texture parameters of pulmonary nodule CT image were extracted,including mean,variance,skewness,kurtosis,energy,autocorrelation,contrast and entropy.Independent sample t-test or chi-square test were used to compare the differences in morphological characteristics and texture parameters between benign and malignant nodules group.The parameters with statistical differences were independent variables,and a multi-parameter logistic regression analysis was performed.The ROC curve was constructed to analyze the diagnostic performance of the logistic regression model.Results Multivariate logistic regression analysis screened out the borderline(OR =6.894,P <0.05),burr sign(OR=1.542,P <0.05)and leaf symbiosis(OR=3.846,P<0.01).Pleural pull sign(OR =8.467,P <0.001),energy(OR =8.972,P <0.01)and entropy(OR =4.578,P <0.001)for independent predictors of patients with malignant pulmonary nodules.The area under the ROC curve established based on independent predictors was 0.894.Sensitivity and specificity were93.43%and 84.18%,respectively.Conclusion CT image texture analysis combined with imaging features has certain predictive value for early lung cancer.
Objective Base on the texture parameters of the solitary pulmonary nodule CT image,a multivariate analysis was performed to establish a data model for predicting early lung cancer.Methods Seventy-two cases of pulmonary nodules diagnosed by surgery or biopsy were selected in this study.There were42 cases of malignant nodules and 30 cases of benign nodules.The morphological characteristics of pulmonary nodules were analyzed,including size,border,margin,lobulation sign,burr sign,the vascular bundle sign,pleural pull sign,air bronchogram,vacuole sign,empty hole and etc;The texture parameters of pulmonary nodule CT image were extracted,including mean,variance,skewness,kurtosis,energy,autocorrelation,contrast and entropy.Independent sample t-test or chi-square test were used to compare the differences in morphological characteristics and texture parameters between benign and malignant nodules group.The parameters with statistical differences were independent variables,and a multi-parameter logistic regression analysis was performed.The ROC curve was constructed to analyze the diagnostic performance of the logistic regression model.Results Multivariate logistic regression analysis screened out the borderline(OR =6.894,P <0.05),burr sign(OR=1.542,P <0.05)and leaf symbiosis(OR=3.846,P<0.01).Pleural pull sign(OR =8.467,P <0.001),energy(OR =8.972,P <0.01)and entropy(OR =4.578,P <0.001)for independent predictors of patients with malignant pulmonary nodules.The area under the ROC curve established based on independent predictors was 0.894.Sensitivity and specificity were93.43%and 84.18%,respectively.Conclusion CT image texture analysis combined with imaging features has certain predictive value for early lung cancer.
引文
[1]迟淑萍.CT图像纹理分析在鉴别良恶性肺结节中的价值[J].实用放射学杂志,2016,32(11):1789-1792.
[2] CHOI W,OH J,ALAM R S,et al.Radiomics analysis of pulmonary nodules in low dose ct for early detection of lung cancer:th-ab-201-10[J].Medical Physics,2017,44(6):3288.
[3] MCKEE B J,REGIS S M,MCKEE A B,et al.Performance of ACR Lung-RADS in a clinical CT lung screening program[J].J Am Coll Radiol,2015,12(3):273-276.
[4] PATZ E F,PINSKY P,GATSONIS C,et al.Overdiagnosis in low-dose computed tomography screening for lung cancer[J].JAMA Intern Med,2014,174(2):269-274.
[5]王建卫,吴宁,唐威,等.低剂量CT肺癌筛查检出肺癌的影像特征[J].中华放射学杂志,2015,23(5):336-339.
[6]张竹伟,华婷,徐婷婷,等.常规MRI纹理分析鉴别乳腺良、恶性病变的价值初探[J].中华放射学杂志,2017,51(8):588-591.
[7]陆青云,陈武飞.不同病理类型肺部磨玻璃结节的HDCT征象分析[J].医学影像学杂志,2017,27(6):1084-1087.
[8] XU C,YUAN Q,CHI C,et al.Computed tomography-guided percutaneoustransthoracicneedlebiopsyforsolitary pulmonary nodules in diameter less than 20mm[J].Medicine,2018,97(14):388-391.
[9]顾晓雯,崔磊.单、双指数模型扩散加权成像鉴别肺结节和肿块良恶性的研究进展[J].中华放射学杂志,2018,52(3):236-239.
[10] SUO S,CHENG J,CAO M,et al.Assessment of heterogeneity difference between edge and core by using texture analysis:differentiation of malignant from inflammatory pulmonary nodules and masses[J].Acad Radiol,2016,23(9):1115-1122.
[11] MCWILLIAMS A,TAMMEMAHI M C,MAYO J R,et al.Probability of cancer in pulmonary nodules detected on first screening CT[J].N Engl J Med,2013,369(10):910-919.
[12]喻微,陈天翔,续力云,等.表现为磨玻璃结节的孤立性肺结节诊断模型的建立[J].中国医学影像学杂志,2017,25(6):435-440.
[2] CHOI W,OH J,ALAM R S,et al.Radiomics analysis of pulmonary nodules in low dose ct for early detection of lung cancer:th-ab-201-10[J].Medical Physics,2017,44(6):3288.
[3] MCKEE B J,REGIS S M,MCKEE A B,et al.Performance of ACR Lung-RADS in a clinical CT lung screening program[J].J Am Coll Radiol,2015,12(3):273-276.
[4] PATZ E F,PINSKY P,GATSONIS C,et al.Overdiagnosis in low-dose computed tomography screening for lung cancer[J].JAMA Intern Med,2014,174(2):269-274.
[5]王建卫,吴宁,唐威,等.低剂量CT肺癌筛查检出肺癌的影像特征[J].中华放射学杂志,2015,23(5):336-339.
[6]张竹伟,华婷,徐婷婷,等.常规MRI纹理分析鉴别乳腺良、恶性病变的价值初探[J].中华放射学杂志,2017,51(8):588-591.
[7]陆青云,陈武飞.不同病理类型肺部磨玻璃结节的HDCT征象分析[J].医学影像学杂志,2017,27(6):1084-1087.
[8] XU C,YUAN Q,CHI C,et al.Computed tomography-guided percutaneoustransthoracicneedlebiopsyforsolitary pulmonary nodules in diameter less than 20mm[J].Medicine,2018,97(14):388-391.
[9]顾晓雯,崔磊.单、双指数模型扩散加权成像鉴别肺结节和肿块良恶性的研究进展[J].中华放射学杂志,2018,52(3):236-239.
[10] SUO S,CHENG J,CAO M,et al.Assessment of heterogeneity difference between edge and core by using texture analysis:differentiation of malignant from inflammatory pulmonary nodules and masses[J].Acad Radiol,2016,23(9):1115-1122.
[11] MCWILLIAMS A,TAMMEMAHI M C,MAYO J R,et al.Probability of cancer in pulmonary nodules detected on first screening CT[J].N Engl J Med,2013,369(10):910-919.
[12]喻微,陈天翔,续力云,等.表现为磨玻璃结节的孤立性肺结节诊断模型的建立[J].中国医学影像学杂志,2017,25(6):435-440.