基于直肠系膜脂肪多参数MRI影像组学列线图预测直肠癌壁外静脉侵犯  

A study on predictive value of the radiomics nomogram based on multiparametric MRI of the mesorectalfat for extramural venous invasion in rectal cancer

作　　者：陈高松 高强 朱建伟 CHEN Gao-song;GAO Qiang;ZHU Jian-wei(Department of Radiology,Taihe Hospital of Traditional Chinese Medicine,Anhui University of Chinese Medicine,Anhui 236600,China)

机构地区：[1]安徽中医药大学附属太和县中医院影像科,安徽阜阳236600

出　　处：《放射学实践》2025年第4期515-522,共8页Radiologic Practice

摘　　要：目的:建立并验证基于直肠癌系膜脂肪的多参数MRI放射组学列线图对预测直肠癌病理性壁外静脉侵犯(EMVI)的价值。方法:回顾性将2017年11月-2024年3月在本院经病理证实的189例直肠癌患者纳入本研究,其中,EMVI阳性组56例,EMVI阴性组133例。所有患者于术前一周内行MRI检查,检查序列包括T 2WI、DWI和DCE-T 1WI。自每例患者3个序列的图像分别提取影像组学特征,并采用提最小绝对收缩和选择算子(LASSO)回归分析筛选最佳特征、建立并计算影像组学标签值(Rad-score)。然后分别基于每个序列的最佳组学特征、MRI常规征像及磁共振EMVI评分(mrEMVI)构建影像组学模型、MRI常规征像模型、mrEMVI模型及联合模型,绘制联合模型的列线图。采用Delong检验比较各模型AUC的差异,并采用ROC曲线、校准曲线和决策曲线分析(DCA)在训练集和验证集中评估列线图的效能和临床实用性。结果:EMVI阳性组与阴性组之间mrT分期、肿瘤横径、Radscore值及mrEMVI评分的差异均有统计学意义(P<0.05)。联合模型的诊断效能最佳,在训练集和验证集中的AUC分别为0.918(95%CI:0.883~0.958)和0.889(95%CI:0.820~0.959)。校准曲线显示联合模型具有良好的校准度,决策曲线显示联合模型在预测直肠癌EVEI阳性时具有良好的临床实用性。结论:基于直肠癌系膜脂肪多参数MRI影像组学模型、MRI常规征像模型和mrEMVI构建的列线图对直肠癌EMVI具有良好的预测效能。Objective:To develop and validate a multiparametric MRI radiomics nomogram nomogram of the mesorectal adipose tissue for predicting pathological extramural venous invasion(EMVI)in rectal cancer.Methods:A total of 189 patients with rectal cancer pathologically confirmed in our hospital from November 2017 to March 2024 were retrospectively included in this study.Among them,there were 56 cases in the EMVI-positive group and 133 cases in the EMVI-negative group.All patients underwent MRI examination within one week before surgery,and the examination sequences included T 2WI,DWI,and DCE-T 1WI.Radiomics features were respectively extracted from the images of the three sequences of each patient,and the least absolute shrinkage and selection operator(LASSO)regression analysis was used to screen the optimal features,establish and calculate the radiomics score(Radscore).Then,radiomics models,conventional MRI features model,mrEMVI models,and combined models were constructed respectively based on the optimal omics features of each sequence,the conventional MRI features,and the magnetic resonance EMVI score(mrEMVI).The Delong test was used to evaluate the performance among various models.A combined model nomogram was constructed according to the radiomics model,the clinical model,and the mrEMVI model.The receiver opera-ting characteristic(ROC)curve and calibration curve were used to analyze and evaluate the efficacy of the nomogram in the training set and the validation set respectively,and the decision curve was used to evaluate the clinical practicality of the combined model.Results:There were statistically significant differences in mrT staging,transverse tumor diameter,Radscore value,and mrEMVI score between the EMVI-positive group and the EMVI-negative group(P<0.05).The diagnostic efficacy of the combined model was the best,and the areas under the curve(AUC)in the training set and the validation set were 0.918(95%CI:0.883~0.958)and 0.889(95%CI:0.820~0.959),respectively.The calibration curve showed that the combined model

关 键 词：直肠肿瘤 壁外血管侵犯 直肠系膜脂肪 磁共振成像 影像组学 

分 类 号：R445.2[医药卫生—影像医学与核医学] R735.3[医药卫生—诊断学]