摘要
目的探讨MRI影像组学模型、临床模型和综合模型对局部进展期直肠癌(LARC)新辅助放化疗(nCRT)疗效的预测价值。资料与方法回顾性收集2017年1月—2021年12月在河南中医药大学第一附属医院行nCRT后行根治性手术的140例LARC患者的临床病理和影像资料,其中病理完全缓解(pCR)108例、无病理完全缓解(npCR)32例,以7∶3随机分为训练组99例和验证组41例。所有患者治疗前均行直肠MRI检查,收集、提取并筛选患者的:①临床特征,包括年龄、性别、癌胚抗原、糖类抗原199、血管通透性参数(K_(trans)、K_(ep)、V_(e))等;②MRI影像组学特征;构建临床模型、影像组学模型及影像组学标签与临床特征相结合的综合模型。采用受试者工作特征曲线下面积(AUC)评估临床、影像组学和综合模型的预测效能,采用决策曲线分析法评价3种模型的临床获益情况,并构建疗效预测的诺模图。结果训练组中,pCR和npCR患者的K_(ep)差异有统计学意义(t=3.862,P<0.0001);验证组中,pCR和npCR患者的K_(trans)和Ve差异有统计学意义(t=2.415、2.552,P均<0.05)。训练组中,综合模型预测效能最佳(AUC=0.922,95%CI 0.846~0.964),敏感度为93.4%,特异度为86.9%;验证组中,综合模型的预测效能稍低于影像组学模型(AUC=0.840,95%CI 0.708~0.944),敏感度为93.3%,特异度为63.6%。综合模型在训练组和验证组中具有较好的校准能力;决策曲线分析显示当风险阈值为2%~96%时,采用综合模型的临床获益高于影像组学模型或临床模型。结论基于MRI影像组学特征的综合模型对LARC患者的nCRT疗效反应具有较高的预测效能,且优于影像组学模型和临床模型。
Purpose To investigate the value of MRI radiomics model,clinical model and integrated model for the efficacy prediction in patient with locally advanced rectal cancer(LARC)after neoadjuvant chemoradiotherapy(nCRT).Materials and Methods A total of 140 patients in the First Affiliated Hospital of Henan University of Chinese Medicine from January 2017 to December 2021 with LARC who received nCRT and radical surgery were retrospectively enrolled.All patients with LARC included 108 patients with pathologic complete response(pCR)and 32 patients with no pathologic complete response(npCR).The data of clinicopathology and MRI were collected and all patients were randomly divided into two groups,including training group(n=99)and validation group(n=41)in a ratio of 7:3.All the patients underwent rectal MRI before treatment,and(i)clinical features of patients were all collected and screened,such as age,gender,carcinoembryonic antigen,carbohydrate antigen 199,and vascular permeability parameters,including volume transfer constant(K_(trans)),interstitium-to-plasma rate constant(K_(ep)),extravascular extracellular space volume fraction(V_(e)),etc;and(ii)MRI radiomics features were also extracted to construct clinical model,radiomics model and comprehensive model combining radiomics label and clinical features.The area under receiver operating characteristic curve(AUC)was used to evaluate the prediction efficiency of the clinical model,radiomics model and integrated model and decision curve analysis was used to evaluate the clinical benefits from three models and a nomogram for prediction was further constructed.Results In the training group,there were significant differences in K_(ep) between pCR and npCR groups(t=3.862,P<0.0001),while in the validation group,there were significant differences in K_(trans) and V_(e) between pCR and npCR groups(t=2.415,2.552,all P<0.05).In the training group,the integrated model demonstrated the best predictive performance(AUC=0.922,95%CI 0.846-0.964),and sensitivity and specificity were 93.4%an
作者
周彦汝
张岚
韩鼎盛
庞志峰
ZHOU Yanru;ZHANG Lan;HAN Dingsheng;PANG Zhifeng(MRI Department,the First Affiliated Hospital of Henan University of Chinese Medicine,Zhengzhou Key Laboratory of Intelligent Analysis and Utilization of Traditional Chinese Medicine Information,Henan University of Chinese Medicine,Zhengzhou 450000,China)
出处
《中国医学影像学杂志》
CSCD
北大核心
2022年第9期881-888,共8页
Chinese Journal of Medical Imaging
基金
河南省科技攻关计划(212102310732)
国家中医临床研究基地科研专项课题(2019JDZX2074)
2021SKY影像科研基金(Z-2014-07-2101)。
关键词
磁共振成像
影像组学
直肠癌
局部进展期
新辅助放化疗
综合模型
Magnetic resonance imaging
Radiomics
Rectal cancer
Locally advanced
Neoadjuvant chemoradiotherapy
Integrated model
