The Value Of Predicting Human Epidermal Growth Factor Receptor 2 Status In Adenocarcinoma Of The Esophagogastric Junction In Different Models Of CT-based Radiomics

1.PurposeTo explore and evaluate the value of predicting HER2 status in adenocarcinoma of esophagogastric junction on CT-based radiomics nomogram model and SVM machine learning model.2.MethodThe clinical features and CT images of 101 patients with pathologically confirmed AEG were retrospectively analyzed and collected.All patients were divided into a training set of 70 cases and a validation set of 30 cases in a 7:3ratio.The region of interest(ROI)was manually delineated on the largest section of the tumor region in the CT portal phase images of all patients,and the radiomics features were extracted in the ROI,screening of the best radiomics features on the training set data using Lasso regression.The radiomics signature were constructed using the radiomics signature score(Rad-score)formula.Multivariate logistic regression analysis to screen independent risk factors for HER2 status in AEG patients.A radiomics-clinical nomogram model,a radiomics SVM model,and a radiomics-clinical SVM model were constructed by combining radiomics signatures and independent risk factors,respectively.And the predictive performance of the models was evaluated in the two sets.3.ResultAmong the 101 AEG patients,46 were HER2(+)and 55 were HER2(-).Multivariate Logistic regression analysis showed that T stage and Rad-score were independent risk factors for HER2 status.(1)The AUC of the radiomics-clinical nomogram model constructed based on T staging and Rad-score was 0.94(95% CI: 0.91-0.97)and 0.90(95% CI:0.84-0.95)in the training set and the validation set,respectively.The calibration curve of the nomogram model showed a good degree of calibration,and DCA indicated that the nomogram model had good clinical application value.(2)The AUC of the radiomics SVM model based on Rad-score in the training set and the validation set was 0.86(95% CI: 0.78-0.94)and 0.78(95%CI: 0.63-0.92),respectively.The AUC of the radiomics-clinical SVM model based on T staging and Rad-score was 0.91(95% CI: 0.85-0.98)and 0.87(95%CI: 0.75-0.99)in the training set and the validation set,respectively.There was no significant difference in AUC between the two SVM models assessed by Delong test(Z=-2.03,-1.25,P=0.42,0.20).4.Conclusion(1)The radiomics nomogram prediction model constructed based on T staging and radiomics signature has good diagnostic performance in predicting the HER2 status of AEG patients.It can be used for individualized prediction of preoperative HER2 status in AEG patients,and has good clinical application value.(2)Both the radiomics SVM prediction model and the radiomics-clinical SVM prediction model achieved good diagnostic performance in predicting the HER2 status of AEG patients.The discriminative performance of the radiomics-clinical SVM prediction model was better than that of the radiomics SVM prediction model.However,the Delong test showed that the performance of the radiomics SVM prediction model was equivalent to that of the radiomics-clinical SVM prediction model established by combining clinical risk factors.