基于Nomogram模型鉴别肺腺癌病理亚型的临床价值
The clinical value of distinguishing pathological subtypes of lung adenocarcinoma based on Nomogram maps
王朝晖 1岳军艳1
作者信息
- 1. 新乡医学院第一附属医院放射科 河南 新乡 453100
- 折叠
摘要
目的 探讨基于最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)回归分析构建Nomogram模型预测原位腺癌(AIS)、微浸润腺癌(MIA)及浸润性腺癌(IAC)的价值.方法 选取本院 97 例经手术病理证实且病理亚型明确的肺腺癌患者,将AIS和MIA归为第 1 组,IAC为第 2 组,比较两组患者年龄、性别、吸烟史、长径、短径及免疫组化Ki-67 等临床医学特征差异,采用 3D Slicer软件进行图像分割,特征提取与选择,通过LASSO算法对特征进行降维,筛选影像组学特征构建预测模型.再采用R软件的rms工具包构建Nomogram模型,计算ROC曲线下面积(AUC),以评价Nomogram模型鉴别肺磨玻璃结节病理亚型的效能.结果 1)性别、吸烟史、长径、短径及免疫组化Ki-67 等临床医学特征均差异无统计学意义(P>0.05);2)筛选 7 个 CT 影 像 组 学 特 征:平面度、大依赖低灰度强调、小波变换LHL第十百分位、小波变换HLL第十百分位、小波变换最小值、小波变换均值及小依赖低灰度强度比较,差异均有统计学意义(P均<0.05);3)基于CT影像组学特征建立预测肺磨玻璃结节病理亚型的Nomogram模型,训练集中AUC为 0.863,准确率为 87.9%,灵敏度为 67.9%,特异度为 91.1%;验证集中 AUC为 0.792,准确率为 75.0%,灵敏度为66.7%,特异度为 90.5%,可见此Nomogram模型具有较好的预测效能.结论 对于预测肺腺癌浸润程度,Nomogram 模型具有明显优势,可作为一种鉴别手段.
Abstract
Objective To explore the value of constructing a Nomogram model based on LASSO regression analysis for pre-dicting adenocarcinoma in situ(AIS),minimally invasive adenocarcinoma(MIA),and invasive adenocarcinoma(IAC).Meth-ods In the First Affiliated Hospital of Xinxiang Medical University,Henan Province,97 patients with adenocarcinoma of the lung confirmed by surgery and pathology and with definite pathological subtype were retrospectively analyzed.AIS and MIA were classified as group one,IAC as group two.The differences of clinical medical characteristics between the two groups,such as age,gender,smoking history,long diameter,short diameter and immunohistochemical Ki-67,were compared.The characteristics were extracted by 3Dslicer software.By using the LASSO algorithm to reduce the dimensionality of features and to filter out imag-ing omics features,a prediction model was constructed.Then,the rms toolkit of R software was used to draw the nomogram and to calculate the area under the curve(AUC)of the receiver operating characteristic to evaluate the efficacy of the nomogram in identi-fying the pathological subtypes of ground glass nodules of the lung.Results 1)Clinical medical characteristics such as gender,smoking history,history of primary tumors,long and short diameters,and immunohistochemical Ki-67 were not statistically sig-nificant(P>0.05);2)This study screened 7 CT imaging omics features in terms of original_shape_Flatness,original_gldm_Large Dependence Low Gray Level Emphasis,wavelet LHL_first order_10 Percentile,wavelet HLL_first order_10 Percentile,wavelet HLL_first order_Minimum,wavelet HHL_firstorder_Mean,and wavelet HHH_gldm_Small Dependence Low Gray Level Empha-sis.Based on this,a predictive model for differentiating pathological subtypes of pulmonary ground glass nodules was established(P>0.05);and 3)According to the nomogram of predicting pathological subtypes of pulmonary ground glass nodules based on the characteristics of CT imaging,the AUC in the training set was 0.863,the accuracy was 87.9%,the sensitivity was 67.9%,and the specificity was 91.1%.The validation set AUC was 0.792,with an accuracy of 75.0%,sensitivity of 66.7%,and specificity of 90.5%,indicating that this column chart had good predictive performance.Conclusion Nomogram model has obvious advan-tages in predicting the invasion degree of adenocarcinoma of the lung and can be used as a means of differentiation.
关键词
肺磨玻璃结节/最小绝对收缩和选择算子/Nomogram模型/病理亚型/体层摄影术,X线计算机Key words
Pulmonary ground glass nodules/Least absolute shrinkage and selection operator/Nomogram model/Pathologi-cal subtype/Tomography,X-ray computed引用本文复制引用
基金项目
河南省医学科技攻关计划联合共建项目(LHGJ20200487)
出版年
2024
NETL
NSTL
万方数据