基于治疗前MR-DWI影像组学预测肺癌化疗疗效的初步研究

目的:探讨基于磁共振扩散加权成像(DWI)的影像组学对肺癌化疗疗效的预测价值.方法:回顾性搜集连续30例经病理证实的肺癌患者的病例资料,根据第二周期化疗后肿瘤最大径退缩率、按RECIST标准将患者分为治疗有效组(16例)和无效组(14例).提取所有患者的化疗前ADC图像(b=600、800和1000s/mm2),应用影像组学方法,在每种b值的ADC图像上提取病灶的19985个特征,采用Lasso进行降维和建模.采用受试者工作特征曲线(ROC)计算三种b值模型预测化疗疗效的诊断效能,并采用DeLong检验比较三种曲线的曲线下面积(AUC).结果:30例中有效组16例、无效组14例.基于b=600s/mm2的ADC图像的影像组学特征所建立的模型,其AUC、诊断敏感度和特异性分别为0.875、0.895和0.750;基于b=800s/mm2的ADC图像,其相应的AUC、诊断敏感度和特异度分别为0.924、0.947和0.938,基于b=1000s/mm2的ADC图像,相应的AUC、诊断敏感度和特异度分别为0.918、0.895和0.875.三种b值的AUC差异无统计学意义(P＞0.05).结论:基于MR-DWI的影像组学可在治疗前对肺癌化疗疗效作出准确预测.

DWI-based radiomics for predicting response of lung cancer to chemotherapy:a pilot study

Objective:To investigate the value of radiomics for DWI in prediction of treatment response of lung cancer to chemotherapy.Methods:Thirty patients with lung cancer confirmed by pathology were enrolled retrospectively.According to the RECIST,all patients were divided into good response group (GR) and poor response group (PR) based on tumor maximum diameter shrinkage rate after the second cycle of chemotherapy.ADC imaging (b =600,800 and 1000s/mm2) were collected,and radiomics features were further extracted and analyzed.In total,19985 radiomics features were extracted in the ADC imaging for each patient,respectively.The absolute shrinkage and selection operator (Lasso) was adopted in feature selection and used in the process of building the classifier model.Receiver operating characteristic (ROC) curve was used to evaluate the capability of the three model to predict GR.Results:The AUC,sensitivity and specificity of the radiomics model based on the ADC imaging (b=600s/mm2) for GR prediction were 0.875,0.895 and 0.750.The AUC,sensitivity and specificity of the radiomics model based on the ADC imaging (b=800s/mm2) for GR prediction were 0.924,0.947 and 0.938.The AUC,sensitivity and specificity of the radiomics model based on the ADC imaging (b=1000s/mm2) for GR prediction were 0.918,0.895 and 0.875.The AUCs among the three groups had no difference (P＞0.05).Conclusion:It is possible to predict response of lung cancer to chemotherapy based on the radiomics of pretreatment DWI.