A CT Features-based Model for Predicting Histological Grade of Gastric Cancer
Objective To explore the application value of a CT features-based model in predicting the histological grade of gastric cancer.Methods We retrospectively collected clinicopathological and CT imaging data from 273 patients with gastric cancer to compare the differences in CT features between the poorly differentiated group and moderately/well differentiated group.Univariate and multivariate analysis was conducted to determine the risk factors for gastric cancer differentiation.We evaluated the predictive efficacy of the model using the receiver operating characteristic curve.Results There are statistically significant differences in the age,tumor location,CT value in venous phase images,arterial phase enhancement rate,arterial phase enhancement degree,T stage,N stage,and lymphovascular invasion between two groups(P＜0.05).The age,location,T stage,CT value in venous phase images,and arterial phase enhancement rate were identified as risk factors for gastric cancer differentiation.In the testing set,the joint model that used all these risk factors had a higher AUC than the qualitative model combined with location and T stage(0.787 vs.0.720)and the quantitative model combined with age,CT value in venous phase images and arterial phase enhancement rate(0.787 vs.0.759).Conclusion The age,location,T stage,CT value in venous phase images,and arterial phase enhancement rate were all risk factors for gastric cancer differentiation.The joint model that used all these risk factors can effectively predict the histological grade of gastric cancer.
万方数据
维普
