基于放疗前增强定位CT的影像组学预测非小细胞肺癌脑转移分次立体定向放疗的近期疗效  

Prediction of short-term efficacy of fractionated stereotactic radiotherapy for brain metastases from non-small-cell lung cancer based on radiomics of pre-radiotherapy enhanced CT simulation

作　　者：夏铀铀 武丹 范育伟 周冲[2] Xia Youyou;Wu Dan;Fan Yuwei;Zhou Chong(Department of Radiation Oncology,Lianyungang Hospital Affi liated to Xuzhou Medical University(The First People's Hospital of Lianyungang),Lianyungang 222061,China;Department of Radiation Oncology,Xuzhou Central Hospital,Xuzhou 221009,China)

机构地区：[1]徐州医科大学附属连云港医院(连云港市第一人民医院)肿瘤放疗科,江苏连云港222061 [2]徐州市中心医院肿瘤放疗科,江苏徐州221009

出　　处：《实用肿瘤杂志》2022年第5期465-469,共5页Journal of Practical Oncology

基　　金：吴阶平医学基金(320.6750.2020-10-73);徐州市卫生健康委青年医学科技创新项目(XWKYHT20200024)。

摘　　要：目的探讨基于放疗前增强定位CT影像组学特征预测非小细胞肺癌(non-small-cell lung cancer,NSCLC)脑转移立体定向放疗近期疗效的价值。方法回顾性分析2016年1月至2019年12月连云港市第一人民医院接受脑分次立体定向放疗(fractionated stereotactic radiation therapy,FSRT)的52例NSCLC脑寡转移患者。采用神经系统肿瘤脑转移反应评估(The Response Assessment in Neuro-Oncology Brain Metastases,RANO-BM)标准将患者分为治疗反应组(完全缓解+部分缓解;n=38)及治疗无反应组(疾病稳定+疾病进展;n=14)。提取患者脑转移瘤放疗前增强定位CT影像组学特征,应用ITK-SNAP软件勾画肿瘤信息,MaZda软件提取影像组学特征,LASSO回归降维法处理高维数据,构建模型预测疗效。绘制受试者工作特征(receiver operating characteristic,ROC)曲线的曲线下面积(area under the curve,AUC)评价模型效能。结果对纳入的52例患者提取287项影像组学特征,包括三维直方图、灰度水平共生矩阵、自动回归模型、小波转换、绝对梯度和运行长度矩阵六大类。剔除共线性变量后使用LASSO回归降维后筛选出3个最优组学特征:S(1,0)Correlat、S(2,2)InvDfMom和S(4,0)Contrast。构建回归方程:y=0.89583×S(1,0)Correlat+0.99934×S(2,2)InvDfMom+0.10925×S(4,0)Contrast-25.64。其效能指标为:AUC为0.71(95%CI:0.64~0.90),敏感度为75.4%,特异度为72.6%,约登指数为0.48。结论放疗前增强定位CT影像组学特征具有较好地预测NSCLC脑转移FSRT疗效的价值。Objective To evaluate the value of using radiomics of enhanced CT simulation before radiotherapy to predict the short-term efficacy of stereotactic radiotherapy for brain metastases(BM)from non-small-cell lung cancer(NSCLC).Methods A retrospective analysis was performed on 52 NSCLC patients with oligo-BM treated with fractionated stereotactic radiation therapy(FSRT)in The First People’s Hospital of Lianyungang,from January 2016 to December 2019.According to The Response Assessment in Neuro-Oncology Brain Metastases(RANO-BM)criteria,patients were classified into the response group(complete response+partial response,n=38)and non-response group(stable disease+progressive disease,n=14).The CT image features of patients with BM before radiotherapy were extracted,the tumor regions of interest was delineated using ITK-SNAP software,and the radiomics features were extracted by MaZda software.LASSO regression was used to construct a model to predict lesion response.The area under the curve(AUC)of receiver operating characteristic(ROC)curve was drawn to evaluate the effectiveness of the model.Results Two-hundred-and-eighty-seven radiomics features were extracted from the 52 patients enrolled,including six categories:3D histogram,gray level co-occurrence matrix,automatic regression model,wavelet transform,absolute gradient and running length matrix.After eliminating collinear variables,LASSO regression was used to reduce dimensionality and three optimal omics features were screened out:S(1,0)Correlat、S(2,2)InvDfMom,and S(4,0)Contrast.A regression equation was contructed:y=0.89583×S(1,0)Correlat+0.99934×S(2,2)InvDfMom+0.10925×S(4,0)Contrast-25.64.The equation had an AUC of 0.71(95%CI:0.64-0.90),a sensitivity of 75.4%,a specificity of 72.6%,and a Youden’s index of 0.48.Conclusions Enhanced CT simulation-based radiomics features before radiotherapy have a good predictive value of FSRT efficacy for BM from NSCLC.

关 键 词：非小细胞肺癌 脑转移 分次立体定向放疗 影像组学 

分 类 号：R734.2[医药卫生—肿瘤]