邢倩,谷小磊,朱海涛,李晓婷,崔湧,孙应实.CT影像组学预测结直肠癌肝转移术后患者1年内无进展生存期[J].中国医学影像技术,2022,38(7):1035~1040
CT影像组学预测结直肠癌肝转移术后患者1年内无进展生存期
CT radiomics for predicting progression free survival within one year in patients with colorectal cancer liver metastases after surgical resection
投稿时间:2022-01-26  修订日期:2022-04-14
DOI:10.13929/j.issn.1003-3289.2022.07.017
中文关键词:  结直肠肿瘤    肿瘤转移  体层摄影术,X线计算机  影像组学
英文关键词:colorectal neoplasms  liver  neoplasm metastasis  tomography,X-ray computed  radiomics
基金项目:国家自然科学基金(61520106004)。
作者单位E-mail
邢倩 北京大学肿瘤医院暨北京市肿瘤防治研究所医学影像科 恶性肿瘤发病机制及转化研究教育部重点实验室, 北京 100142  
谷小磊 北京大学肿瘤医院暨北京市肿瘤防治研究所医学影像科 恶性肿瘤发病机制及转化研究教育部重点实验室, 北京 100142  
朱海涛 北京大学肿瘤医院暨北京市肿瘤防治研究所医学影像科 恶性肿瘤发病机制及转化研究教育部重点实验室, 北京 100142  
李晓婷 北京大学肿瘤医院暨北京市肿瘤防治研究所医学影像科 恶性肿瘤发病机制及转化研究教育部重点实验室, 北京 100142  
崔湧 北京大学肿瘤医院暨北京市肿瘤防治研究所医学影像科 恶性肿瘤发病机制及转化研究教育部重点实验室, 北京 100142 yong.cui@outlook.com 
孙应实 北京大学肿瘤医院暨北京市肿瘤防治研究所医学影像科 恶性肿瘤发病机制及转化研究教育部重点实验室, 北京 100142  
摘要点击次数: 1149
全文下载次数: 341
中文摘要:
       目的 评价CT影像组学模型预测结直肠癌肝转移(CRLM)患者术后1年无进展生存期(PFS)的价值。方法 回顾性分析147例CRLM术前末次CT资料,将其分为训练集(n=100)及验证集(n=47),依据临床预后分为预后良好(PFS≥12个月)及预后不佳(PFS<12个月)。手动分割CT所示肝转移灶,提取及选择特征后,基于训练集数据构建影像组学模型,以多因素logistic回归构建临床模型及联合模型。绘制受试者工作特征曲线,计算曲线下面积(AUC),评价并比较各模型预测CRLM患者术后PFS的效能。结果 共选出7个特征用于构建影像组学模型。临床模型中,原发灶N分期、有无基因突变及有无术后化疗是预测CRLM患者术后PFS的独立因素;联合模型中的独立因素还包括影像组学评分。影像组学模型、联合模型预测训练集CRLM患者术后PFS的AUC均大于临床模型(0.89、0.93及0.67,P均<0.05);其在验证集的AUC依次为0.77、0.78及0.56,前二者的效能优于临床模型(P<0.05)。结论 影像组学模型及联合模型预测CRLM患者1年内PFS的效能均较好。
英文摘要:
      Objective To explore the value of preoperative CT radiomics model for predicting progression free survival (PFS) within one year in patients with colorectal cancer liver metastases (CRLM) after surgical resection. Methods The last time CT data of 147 CRLM patients were retrospectively analyzed. The patients were divided into the training set (n=100) and the validation set (n=47), also into good prognosis group (PFS≥12 months) and poor prognosis group (PFS<12 months) according to clinical prognosis. After manually segmenting the liver lesions, extracting and selecting radiomics features, the radiomics model was constructed based on the training set. Then multivariable logistic regression was used to build the clinical model, radiomics model and combining clinic-radiomics model. The receiver operating characteristic curves were drawn, and the areas under the curves (AUC) were calculated to evaluate the efficacy of the models for predicting PFS within one year in patients with CRLM. Results Seven radiomics features were screened for constructing radiomics model. N stage of primary tumor, the presence of gene mutation and postoperative chemotherapy were all independent predictors for postoperative PFS in patients with CRLM for clinical model, while for combined model, the above indexes and radiomics scores were all independent predictors. AUC of radiomics model and combined model for predicting postoperative PFS in patients with CRLM in training set was higher than that of clinical model (0.89, 0.93 and 0.67, both P<0.05), while in validation set was 0.77, 0.78 and 0.56, respectively, and the efficacy of combined model was superior to that of clinical model (P<0.05). Conclusion Both radiomics model and combined model had good efficacy for predicting PFS of CRLM patients within one year after surgical resection.
查看全文  查看/发表评论  下载PDF阅读器