中国中西医结合影像学杂志2025，Vol.23Issue(3)：297-302,6.DOI:10.3969/j.issn.1672-0512.2025.03.006
基于CT的临床影像组学联合模型对胃癌脉管侵犯的术前预测研究
Preoperative prediction of vascular invasion in gastric cancer using CT-based clinicoradiomics combined model
摘要
Abstract
Objective:To explore the preoperative prediction of vascular invasion in advanced gastric cancer by clinicoradiomics combined model based on CT.Methods:The clinical and CT imaging data of 200 patients with advanced gastric cancer confirmed by operation and pathology were retrospectively collected.All patients underwent conventional abdominal CT plain scan and three-phase enhanced scan before surgery.The target lesion(the largest lesion)was manually delineated layer by layer on the venous phase images,and the edge of the target lesion was fused to complete the three-dimensional volume of interest(VOI)of the lesion.And further image segmentation,feature extraction and statistical analysis were performed.According to pathological diagnosis,67 cases had vascular tumor embolus(the positive group)and 133 cases had not vascular tumor embolus(the negative group).Patients were randomly divided into the training set(139 cases)and the validation set(61 cases)at a ratio of 7∶3.The independent risk factors of vascular invasion in advanced gastric cancer were screened out from clinical indicators through univariate and multivariate analysis,and the clinical model was constructed.The variance threshold,SelectKBest and LASSO algorithm were used to select the optimal radiomics features.Based on these independent risk factors and selected radiomics features,a clinicoradiomics combined model was constructed.The ROC curve was used to evaluate model performance.Results:There were significant differences in tumor location,long diameter,pathological differentiation degree,lymph node metastasis,AJCC stage and Ki-67 between the two groups(all P<0.05),while there were no significant differences in age or gender(both P>0.05).Multivariate logistic analysis showed the pathological differentiation degree,AJCC stage and Ki-67 were the independent risk factors of vascular invasion(all P<0.05).1,688 radiomics features were extracted from CT venous phase images,and 12 optimal features were selected by LASSO algorithm to construct a radiomics model.Three clinical features were selected to construct the clinical model.Radiomics model and clinical model were combined to construct the clinicoradiomics combined model.The ROC curve showed that the combined model had good predictive efficacy in the training set(AUC of 0.939,precision of 0.871,sensitivity of 0.870,specificity of 0.871)and validation set(AUC of 0.741,precision of 0.705,sensitivity of 0.857,specificity of 0.625).The decision curve also showed that the combined model had good diagnostic accuracy and clinical utility.Conclusion:CT-based clinicoradiomics combined model has high efficacy in preoperative prediction of vascular invasion in advanced gastric cancer.关键词
进展期胃癌/脉管侵犯/临床影像组学/体层摄影术,X线计算机/列线图Key words
Advanced gastric cancer/Vascular invasion/Clinicoradiomics/Tomography,X-ray computed/Nomogram引用本文复制引用
舒恩芬,余卓,胡玉敏,周永进,毛卫波,高杨,纪建松..基于CT的临床影像组学联合模型对胃癌脉管侵犯的术前预测研究[J].中国中西医结合影像学杂志,2025,23(3):297-302,6.基金项目
丽水市科学技术局项目(2020SJZC064). （2020SJZC064）
