基于DCE-MRI及临床特征预测结直肠癌微卫星不稳定性的研究  

Study on the Prediction of Microsatellite Instability in Colorectal Cancer Based on DCE-MRI and Clinical Features

作　　者：施柳言 潘妮妮 赵建新 熊恋秋 马丽丽 何迪梁 赵致平 王莉莉[3] 赵莲萍[3] 黄刚[3] SHI Liuyan;PAN Nini;ZHAO Jianxin;XIONG Lianqiu;MA Lili;HE Diliang;ZHAO Zhiping;WANG Lili;ZHAO Lianping;HUANG Gang(First Clinical Medical School,Gansu University of Chinese Medicine;Department of Radiology,Gansu Wuwei Tumour Hospital;Department of Radiology,Gansu Provincial Hospita)

机构地区：[1]甘肃中医药大学第一临床医学院 [2]甘肃省武威肿瘤医院影像科 [3]甘肃省人民医院放射科

出　　处：《中国医学计算机成像杂志》2024年第4期473-484,共12页Chinese Computed Medical Imaging

基　　金：甘肃省青年科技基金计划项目(20JR5RA143)。

摘　　要：目的:探讨基于动态对比增强磁共振成像(DCE-MRI)联合临床特征构建的列线图模型在术前预测结直肠癌微卫星不稳定性(MSI)的价值。方法:回顾性收集甘肃省人民医院经病理证实为结直肠癌患者的临床资料和磁共振图像。选取小视野高分辨率T2加权成像(sFOV HR-T2WI)、DCE-MRI峰值期和T1加权延迟成像(T1-DELAY)勾画肿瘤最大层面及上下两层作为感兴趣区(ROI)提取影像组学特征并进行降维筛选,之后分别构建影像组学单序列及多序列联合模型并计算影像组学评分(Radscore)。后处理工作站中勾画肿瘤最大层面测量灌注参数速率常数(K_(ep))、转运常数(K^(trans))及容积分数(V_(e)),获取频数表并计算直方图特征,特征降维筛选后计算灌注参数评分(Dcescore)。单因素及多因素逻辑回归筛选MSI的临床独立预测因素,联合Radscore和Dcescore构建列线图模型预测结直肠癌微卫星状态。采用受试者工作特征(ROC)曲线、校准曲线和决策曲线评估各模型效能及临床价值。结果:单序列中基于DCE-MRI峰值期图像的模型效能最佳,训练集和验证集ROC曲线下面积(AUC)分别为0.913和0.900;多序列联合模型效能优于任何单序模型,训练集和验证集的AUC分别为0.968和0.962;列线图模型具有更高的诊断效能,训练集和验证集的AUC分别为0.985和0.982。结论:基于DCE-MRI和临床特征构建的列线图模型可以在术前有效地预测结直肠癌患者MSI状态,有可能为结直肠癌的治疗选择及预后评估提供一种无创的评估方法。Purpose:To explore the value of nomogram model based on dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)combined with clinical features in preoperative prediction of microsatellite instability(MSI)in colorectal cancer.Method:Clinical data and MRI images from patients diagnosed with colorectal cancer at Gansu Provincial People's Hospital were retrospectively collected.Small field-of-view high-resolution T2WI(sFOV HR-T2WI)and dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)images during peak and T1 weighted delayed imaging(T1-DELAY)images were used to delineate the maximum slice and upper and lower slices of the tumor as regions of interest(ROI)to extract radiomics features and perform dimensionality reduction radiomics score(Radscore)was calculated.In the post-processing workstation,the rate constant(K_(ep)),transfer constant(K^(trans)),and volume fraction of extravascular extracellular space(V_(e))of the measured perfusion parameters at the maximum slice of the tumor were delineated,a frequency table was obtained,and histogram features were calculated.After feature dimensionality reduction screening,the perfusion parameter score(Dcescore)was calculated.Clinical independent predictors for MSI were selected using univariate and multivariate logistic regression,and nomogram model combining Radscore and Dcescore was constructed to predict colorectal cancer microsatellite instability.The performance and clinical value of each model were evaluated using receiver operating characteristic(ROC)curves,calibration curves,and decision curves.Result:In the single sequence model,DCE-MRI peak images had the best performance,with an area under the ROC curve(AUC)of 0.913 and 0.900 for the training and validation sets.The performance of the multi sequence joint model was superior to any single sequence model,with AUCs of 0.968 and 0.962 for the training and validation sets.The nomogram model had higher diagnostic performance,with AUCs of 0.985 and 0.982 for the training and validation sets.Conclusion:The

关 键 词：影像组学 结直肠癌 微卫星不稳定性 动态对比增强磁共振成像 

分 类 号：R445.2[医药卫生—影像医学与核医学]