Portal-venous phase CT peritumoral radiomics for predicting lymphovascular invasion of gastric adenocarcinoma
10.13929/j.issn.1003-3289.2025.06.017
- VernacularTitle:门静脉期CT瘤周影像组学预测胃腺癌脉管侵犯
- Author:
Xiao SUN
1
;
Rui DING
1
;
Li MA
1
;
Wenling LI
1
;
Huairong ZHANG
1
;
Li ZHU
1
Author Information
1. 宁夏医科大学总医院放射科,宁夏银川 750004
- Publication Type:Journal Article
- Keywords:
gastric neoplasms;
neoplasm invasiveness;
radiomics;
tomography,X-ray computed
- From:
Chinese Journal of Medical Imaging Technology
2025;41(6):928-932
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore value of peritumoral radiomics models for predicting lymphovascular invasion(LVI)of gastric adenocarcinoma based on portal-venous phase enhanced CT images.Methods Totally 351 patients with gastric adenocarcinoma were collected and randomly divided into training set(n=246)and test set(n=105)at a ratio of 7∶3.ROI of tumors were manually delineated on portal-venous phase enhanced CT images,then radiomics features of tumoral areas and peritumoral areas(1,3 and 5 mm expanded from lesions)were extracted,respectively.Clinical-CT,tumoral and peritumoral(1 mm,3 mm,5 mm)radiomics and comprehensive model nomograms were established,and their predictive performances for LVI were compared.Calibration curve was used to evaluate the consistency between the predicted and actual LVI of gastric adenocarcinoma,while decision curve analysis(DCA)was used to assess the net clinical benefit of each model.Results The area under the curve of clinical-CT,tumoral,peritumoral(1 mm,3 mm,5 mm)radiomics models and the comprehensive model for predicting LVI in training set was 0.741,0.732,0.713,0.728,0.708 and 0.755,respectively,which in test set was 0.748,0.725,0.759,0.724,0.704 and 0.764,respectively.The comprehensive model demonstrated the highest prediction efficiency,also good calibration and clinical practicability.Conclusion Portal-venous phase CT peritumoral radiomics models could be used to predict LVI of gastric adenocarcinoma.Combining with CT features,tumoral and optimal peritumoral radiomics features could further improve predictive efficiency.
