摘要
目的探讨基于CT软窗的影像组学对肺腺癌化疗疗效的评估价值。方法回顾性搜集2015年12月-2018年12月期间在安徽医科大学第二附属医院经病理证实的105例肺腺癌患者的影像及病例资料,根据半年化疗后的疗效,按RECIST标准将患者分为缓解组(35例)和未缓解组(70例),其中未缓解组分为稳定组(35例)及进展组(35例)。提取所有患者的化疗前CT软窗图像,使用ITK-SNAP软件对病灶手动分割,通过影像组学方法使用AK软件分析CT软窗图像分割获得的病灶,提取病灶的纹理特征,采用Lasso降维和RTree建模。采用受试者工作特征曲线(ROC)并建立肺腺癌预测模型决策曲线计算缓解组与未缓解组间比较的模型评估化疗疗效的诊断效能。结果基于缓解组与未缓解组的图像,提取出12个有意义的纹理特征,其中共生矩阵5个,游程矩阵7个,建立相应的影像组学模型,其训练组相应的AUC、诊断敏感度和特异度分别为0.80、0.68和0.80,验证组的AUC、诊断敏感度和特异度分别为0.74、0.70和0.81,2组的AUC值均在0.70~0.90之间,同样达到良好的诊断价值。影像组预测模型决策曲线在0.18~0.76较大的阈值范围内,具有良好的临床实用性。结论基于治疗前CT软窗的影像组学可在化疗前对肺腺癌的疗效作出较准确评估。
Objective To discuss the value of CT soft tissue window radionics in assessing the efficacy of pulmonary adenocarcinoma chemotherapy. Methods Images and case materials of 105 patients pathologically diagnosed with pulmonary adenocarcinoma in our hospital from December 2015 to December 2018 were retrospectively collected. After half-year chemotherapy, the patients were assigned to the response group(35 cases) and non-response group(70 cases) according to RECIST. The non-response group was divided into the stable group(35 cases) and progression group(35 cases). For all patients, their CT soft tissue images before chemotherapy were obtained, and the lesions were manually segmented using ITK-SNAP software. The focal features of those CT soft tissue images analyzed by AK analysis software were extracted using radionics methods, and then subjected to Lasso dimensionality reduction and RTree modeling. The model for comparison between the response group and the non-response group was calculated using the receiver operating characteristic curve(ROC) and image omics model to assess the diagnostic efficiency of chemotherapy efficacy. Results Based on the images of response group and the non-response group, 12 significant texture features were extracted, including 5 co-occurrence matrices(GLCM) and 7 run-length matrices(RLM). The training group’s AUC, diagnosis sensitivity and specificity were 0.80, 0.68 and 0.80, respectively. The validation group’s AUC, diagnosis sensitivity and specificity were 0.74, 0.70 and 0.81, respectively. The AUC values of both groups were between 0.7-0.9, which also achieved good diagnostic value. The decision curve of image omics model had a good clinical practicability in the range of 0.18-0.76. Conclusion Radionics based on the CT images before chemotherapy may help exactly assess the efficacy of pulmonary adenocarcinoma chemotherapy before treatment.
作者
余业洲
赵红
王龙胜
鲍芳
邹立巍
段绍峰
杨进
YU Ye-zhou;ZHAO Hong;WANG Long-sheng;BAO Fang;ZOU Li-wei;DUAN Shao-feng;YANG Jin(Department of Radiology,the Second Hospital of Anhui Medical University,Hefei,Anhui 230601,China)
出处
《中华全科医学》
2020年第4期623-626,701,共5页
Chinese Journal of General Practice
基金
国家自然科学基金项目(81400058)。
关键词
肺腺癌
计算机断层成像
影像组学
化疗
Pulmonary adenocarcinoma chemotherapy
Computed tomography
Radionics
Chemotherapy
