Application of clinical-CT radiomics nomogram for preoperative prediction of colon cancer lymph node metastasis
10.3969/j.issn.1002-1671.2024.12.012
- VernacularTitle:临床-CT影像组学列线图术前预测结肠癌淋巴结转移应用研究
- Author:
Mingsong DONG
1
;
Xiaojin ZHANG
;
Jiajun XU
;
Xianfeng ZHU
;
Yong GUO
;
Xin DAI
;
Fei LIU
;
Hu ZHANG
Author Information
1. 芜湖市第二人民医院医学影像科,安徽 芜湖 241000
- Keywords:
colon cancer;
lymph node metastasis;
radiomics;
nomogram;
computed tomography
- From:
Journal of Practical Radiology
2024;40(12):1989-1993
- CountryChina
- Language:Chinese
-
Abstract:
Objective To construct clinical imaging model,radiomics model,and a combined model based on the above two for predicting lymph node metastasis(LNM)of colon cancer(CC),and to compare the diagnostic performance of each model.Methods The data from 328 CC patients confirmed by surgical pathology were analyzed retrospectively,including 156 with LNM.All patients were randomly divided into training group(229 cases)and validation group(99 cases)at a ratio of 7∶3.The difference of clinical imaging indicators were compared between groups and a clinical imaging model for diagnosing LNM was constructed.The tumor three-dimensional volume of interest(VOI)was used for radiomics feature extraction,and after dimensionality reduction and selection,8 features were obtained to construct the Radiomics score(Radscore).A combined model of clinical imaging indicators and Radscore was built.The diagnostic performance of each model for LNM was compared,and the calibration and clinical benefit of the optimal model were evaluated.Results There were statistical differences in clinical imaging indicators between the two groups:carcinoembryonic antigen(CEA),CA199,tumor long diameter,and lymph node short diameter(P＜0.05).The area under the curve(AUC)of the clinical imaging model,radiomics model,and combined model were 0.721,0.814,0.854(training group),and 0.744,0.732,0.808(validation group),respectively.The AUC of the combined model was the highest,and both the training and validation groups were higher than that of the clinical imaging model(P＜0.05).The combined model demonstrated higher calibration,with a clinical benefit from decision curve analysis(DCA)threshold range of 0.09 to 0.91.Conclusion The nomogram constructed based on clinical imaging indicators and CT radiomics holds high value in diagnosing LNM of CC.
