Objective:To explore the predictive value of nomogram based on CT radiomics and deep learning features for the short-term efficacy of radiotherapy in oesophageal cancer.Methods:The clinical and imaging data of 137 patients with esophageal squamous carcinoma were retrospectively an-alyzed.Radiomics and deep learning features were extracted from CT images.The least absolute shrinkage and selection operator methods were used to reduce the dimension of radiomics features and deep learning features,respectively,and the radiomics score(Radscore)and deep learning score(Deepscore)were calculated.Multivariate logistic regression analysis was used to establish a prediction model and draw a nomogram.The calibration,diagnostic efficiency and clinical value of the nomogram were evaluated.Results:Six radiomics features were selected to calculate the Radscore,and six deep learning features were selected to calculate the Deepscore.The results of multivariate logistic regres-sion showed that Radscore,Deepscore and TNM staging were independent predictors of the combined model.The area under the curve(AUC)of the combined prediction model in the training set was 0.904,which was higher than that of the clinical model(AUC=0.662)and the radiomics model(AUC=0.814),and the differences in the AUCs were statistically significant(P＜0.001 and P=0.004).In the validation set,the AUC of the combined model was 0.938,which was higher than that of the clini-cal model(AUC=0.644)and radiomics model(AUC=0.852).The difference in AUC between the combined model and the clinical model was statistically significant(P＜0.001),while the difference in AUC between the combined model and the radiomics model was not statistically significant(P=0.091).Decision curve analysis showed that the combined prediction nomogram had good clinical prac-ticability within the threshold ranges of 0.1～0.9 and 0.97～0.99.Conclusion:CT radiomics combined with deep learning features can better predict the short-term efficacy of radiotherapy for esophageal cancer.
维普
万方数据