| 胡奎,张娴,白彪胜,王紫君,蔡茜,刘玉林.CT影像组学模型预测靶向药物治疗表皮生长因子受体基因突变非小细胞肺癌患者预后[J].中国医学影像技术,2022,38(10):1491~1496 |
| CT影像组学模型预测靶向药物治疗表皮生长因子受体基因突变非小细胞肺癌患者预后 |
| CT radiomics model for predicting prognosis of targeted drug therapy for non-small cell lung cancer with epidermal growth factor receptor gene mutation |
| 投稿时间：2022-03-21 修订日期：2022-06-19 |
| DOI：10.13929/j.issn.1003-3289.2022.10.011 |
| 中文关键词: 癌,非小细胞肺 ErbB受体 体层摄影术,X线计算机 分子靶向治疗 影像组学 |
| 英文关键词:carcinoma, non-small-cell lung ErbB receptors tomography, X-ray computed molecular targeted therapy radiomics |
| 基金项目:湖北省卫生健康委员会重点支持项目（WJ2019Z015）。 |
|
| 摘要点击次数: 2555 |
| 全文下载次数: 492 |
| 中文摘要: |
| 目的 探讨基于增强CT构建的影像组学模型评估靶向药物治疗表皮生长因子受体（EGFR）基因突变非小细胞肺癌（NSCLC）后患者预后的可行性。方法 回顾性分析86例接受靶向治疗的EGFR基因突变NSCLC患者，按7[DK （]∶[DK）]3比例将其随机归入训练集[n=59，16例无进展生存期（PFS）≤6个月、43例PFS>6个月]和测试集（n=27，8例PFS≤6个月、19例PFS>6个月）。采用Itksnap软件于治疗前肺窗CT增强图像中勾画病变三维ROI，以开源pyradiomics包提取396个特征，以主成分分析（PCA）对特征进行降维，获得23个新的维度特征。根据训练集23个维度特征构建逻辑回归模型，观察其评估训练集和测试集患者预后的价值。结果 以维度特征构建的逻辑回归模型评估训练集、测试集患者预后（PFS≥6个月）的曲线下面积（AUC）分别为0.923、0.849；训练集和预测集的校准曲线与理想模型的对角线均较为接近。影像组学评分瀑布图显示，训练集的阴性预测率为92.30%、阳性预测率为91.30%，测试集分别为100%、86.36%；决策曲线显示，训练集和测试集在0.20~0.90阈概率范围内有很好的净获益。结论 基于CT增强图像的逻辑回归模型可用于预测靶向药物治疗EGFR基因突变NSCLC后患者预后。 |
| 英文摘要: |
| Objective To explore the feasibility of CT radiomic for predicting prognosis of non-small cell lung cancer (NSCLC) patients after targeted drug therapy using epidermal growth factor receptor (EGFR) mutation. Methods Data of 86 EGFR-mutant NSCLC patients who underwent targeted therapy were retrospectively analyzed. The patients were randomly divided into training set (n=59, including 16 cases with progression-free survival[PFS] ≤ 6 months and 43 cases with PFS>6 months) and test set (n=27, including 8 cases with PFS ≤ 6 months and 19 cases with PFS>6 months) at the ratio of 7:3. Based on enhanced chest CT lung window images before treatment, the Itksnap software was used to delineate 3D ROI, and the open source pyradiomics package was used to extract 396 features of tumors. Then the dimension was reduced with principal component analysis (PCA), and 23 new dimensional features were obtained. A logistic regression model was constructed based on 23 dimensional features derived from training set, and the predictive value of the model for the prognosis of EGFR-mutant NSCLC patients after targeted therapy in training set and test set was analyzed. Results The area under the curve (AUC) of patients' prognosis (PFS ≥ 6 months) in training set and test set was 0.923 and 0.849, respectively. The calibration curves of training set and test set were both close to the diagonal of the ideal model. The waterfall graph of imaging score showed that the negative prediction rate of training set was 92.30%, the positive prediction rate was 91.30%, and of test set was 100% and 86.36%, respectively. The decision curve showed that both the training set and test set had good net benefit in the threshold probability range of 0.20-0.90. Conclusion Logistic regression model based on enhanced CT images could be used to predict the prognosis NSCLC patients with EGFR mutation after targeted drug therapy. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
