基于高分辨率T2WI的影像组学列线图预测直肠癌神经周围侵犯的研究  

High resolution T2WI image-based radiomics nomogram for prediction of perineural invasion in rectal cancer

作　　者：杨彦松 丁勇生[1] 张益飞[1] 张明珠[1] 郑桂华[2] 沈月红[1] 葛亚琼 YANG Yansong;DING Yongsheng;ZHANG Yifei;ZHANG Mingzhu;ZHENG Guihua;SHEN Yuehong;GE Yaqiong(Department of Radiology,Affiliated Tumor Hospital of Nantong University,Nantong 226361,Jiangsu Province,China;Department of Pathology,Affiliated Tumor Hospital of Nantong University,Nantong 226361,Jiangsu Province,China;GE Healthcare,Nanjing 210000,Jiangsu Province,China)

机构地区：[1]南通市肿瘤医院影像科,江苏南通226361 [2]南通市肿瘤医院病理科,江苏南通226361 [3]通用电气医疗,江苏南京210000

出　　处：《肿瘤影像学》2022年第2期130-138,共9页Oncoradiology

基　　金：南通市卫生和计划生育委员会科研课题专项面上项目(MA2020007);南通大学临床研究专项青年项目(2019LQ014)。

摘　　要：目的:探讨高分辨率磁共振成像(magnetic resonance imaging,MRI)影像组学列线图预测直肠癌神经周围侵犯(perineural invasion,PNI)的价值。方法:回顾并入组南通市肿瘤医院2016年12月—2020年12月直肠癌患者164例,在高分辨率T2加权成像(T2-weighted imaging,T2WI)斜轴位上逐层勾画病灶,提取影像组学特征。采用最大相关最小冗余对影像组学特征进行初步筛选,然后进行最小绝对收缩和选择算子(the least absolute shrinkage and selection operator,LASSO)回归分析降维,计算影像组学标签。通过单因素和多因素logistic回归分析临床特征、MRI影像学表现、影像组学标签与PNI的关系并构建预测PNI的模型。结果:PNI发生率约29.9%(49/164),PNI阳性组和PNI阴性组之间肿瘤最大径、组织分化程度、T分期、N分期、环周切缘(circumferential resection margin,CRM)状态、壁外血管侵犯(extramural vascular invasion,EMVI)状态差异均有统计学意义(P<0.05),其余指标差异无统计学意义。最终预测PNI的列线图包括组织分化程度、EMVI、影像组学标签,训练集曲线下面积(area under curve,AUC)为0.88(95%CI 0.82~0.95),验证集AUC为0.88(95%CI 0.74~1.00)。结论:基于高分辨率T2WI影像组学列线图能较好地预测直肠癌PNI。Objective:To explore the value of high-resolution magnetic resonance imaging(MRI)-based radiomics nomogram in predicting peripheral invasion(PNI)of rectal cancer.Methods:A total of 164 rectal cancer patients in Affiliated Tumor Hospital of Nantong University from December 2016 to December 2020 were retrospectively enrolled.The lesions were delineated on high-resolution oblique axial T2-weighted imaging(T2WI)layer by layer and radiomics features were extracted.Firstly,the radiomics features were initially screened by maximum correlation and minimum redundancy,then the least absolute shrinkage and selection operator(LASSO)regression analysis was performed to screen the features again and radiomics signature was calculated.Univariate analysis was conducted on the radiomic features,clinical risk factors and MRI findings.Multivariate logistic analysis was carried out to investigate the final feature subset and thus the predicting model was established.Results:PNI was present in 29.9%(49/164)of rectal cancer patients.There were statistically significant differences in tumor length,histological grade,MRI reported T stage,MRI reported N stage,circumferential resection margin status and extramural vascular invasion status between the PNI positive group and PNI negative group(P<0.05).While there were no statistically significant differences in other indicators.The predictive nomogram of PNI included the histological grade,extramural vascular invasion status and radiomics signature.The area under curve(AUC)was 0.88(95%CI 0.82-0.95)in the training cohort,which was 0.88(95%CI 0.74-1.00)in the validation cohort.Conclusion:T2WI-based radiomic nomogram could be helpful for the prediction of PNI preoperatively in rectal cancer patients.

关 键 词：直肠癌 影像组学 磁共振成像 神经周围浸润 列线图 

分 类 号：R735.37[医药卫生—肿瘤] R445.2[医药卫生—临床医学]