Radiomics model based on contrast-enhanced CT images to predict pre-treatment regional lymph node metastasis in rectal cancer
10.3969/j.issn.1002-1671.2023.12.017
- VernacularTitle:基于增强CT构建影像组学模型预测直肠癌患者治疗前区域淋巴结转移
- Author:
Jiaxuan LIU
1
;
Xi LU
;
Ying DU
;
Lingling SUN
Author Information
1. 中国医科大学附属第四医院放射科,辽宁 沈阳 110032
- Keywords:
rectal cancer;
regional lymph node metastasis;
radiomics;
machine learning;
nomogram
- From:
Journal of Practical Radiology
2023;39(12):1985-1989
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the value of a radiomics model based on contrast-enhanced CT at the arterial and venous phases in predicting the regional lymph node metastasis in rectal cancer.Methods The data of 303 patients with pathologically proven rectal cancer were analyzed retrospectively,and 1 648 CT radiomics features of the primary tumors were extracted from the arterial and venous phases,respectively.Four machine learning models were used including radiomics score(Rad-score),K-nearest neighbor(KNN),multi-layer perceptron(MLP)and support vector machine(SVM),and the model with the best diagnostic performance was selected.Taking the predicted probability of the best machine model in the arterial and venous phases as the input value,logistic regression algorithm was used to further construct a combined nomogram.Results The SVM model had been proved to be the most efficient among the four machine learning models,whether in arterial or venous phases.The combined nomogram based on ASVM and VSVM prediction probability further improved the ability to predict regional lymph node metastasis.The area under the curve(AUC)of the train group was 0.860,and that of the validation group was 0.801,which were higher than those of ASVM model(train group,AUC=0.822;validation group,AUC=0.777)and VSVM model(train group,AUC=0.832;validation group,AUC=0.786).Conclusion The arterial-venous phases radiomics combined nomogram model based on enhanced CT images performs well in predicting rectal cancer patients'pre-treatment lymph node metastasis.
