Three-phase Dynamic Contrast-enhanced CT-based Clinical-radiomics Model To Predicted The Malignant Potential,mitotic Count,and Ki-67 Expression Of Preoperative Small Intestinal Stromal Tumor

Objective:A radiomic model based on enhanced CT was established to predict the malignant potential of small intestinal stromal tumor,mitotic count and Ki-67 expression before surgery,providing a feasible noninvasive scheme for guiding clinical diagnosis and postoperative evaluation of SIST patients,and realizing visualization of clinical application of imaging omics in SIST patients.Materials and Methods:The clinical and imaging data of 81 and 26 SIST patients confirmed by surgical pathology from the Second Affiliated Hospital of Nanchang University and the Affiliated Hospital of Jiujiang University from January 2019 to December 2022 were collected,and the patients were randomly divided into training group(74 cases)and verification group(33 cases)according to the ratio of 7:3.Radiomics features were extracted from the images of the enhanced arterial phase,venous phase and equilibrium phase of the patients before surgery.The Intraclass Corrtlation Coefficient(ICC)and Least Absolute Shrinkage and Selection Operator(LASSO)were used for radiomics feature screening and dimensionality reduction.Univariate and multivariate logistic regression were used to feature CT and clinical features.Multivariate logistic regression analysis was used to construct seven models,including arterial phase radiomics model,venous phase radiomics model and equilibrium phase radiomics model,CT-clinical feature model(referred to as clinical model),clinical-arterial phase radiomics model,clinical-venous radiomics model and clinical-equilibrium radiomics combination model,and verified the models in the validation group.The receiver operator characteristic curve(ROC),area under the curve(AUC),Akaile information criterion(AIC)and Delong test were used to evaluate model performance,decision curve analysis,DCA)to evaluate the clinical validity of the model.Results:In the constructed model,the three clinical imaging radiomics combination models had higher AUC values compared to the clinical model and the three single imaging radiomics models,and all had statistical differences.The clinical venous phase imageradiomics model has the best predictive performance,good Goodness of fit and simplicity in predicting the malignant potential,mitotic count and Ki-67 expression of small intestinal stromal tumors.In predicting the malignant potential of small intestinal stromal tumors,the clinical venous phase imaging radiomics model has a good AUC value,with a training group AUC of 0.961 and a validation group AUC of 0.938.The Delong test showed that there was no statistical difference between the clinical venous phase imaging radiomics model,the clinical arterial phase imaging radiomics model,and the clinical equilibrium phase imaging radiomics model.However,in AIC,the clinical venous phase imaging radiomics model had the smallest AIC value(42.568),The results suggest that the clinical venous phase imaging model has the best simplicity and Goodness of fit.Similarly,in predicting the mitotic count of small intestinal stromal tumors,the clinical venous imaging radiomics model showed good AUC values,with a training group AUC of0.858,a validation group AUC of 0.921,and a minimum AIC value of 70.271.In predicting the expression of Ki-67 in small intestinal stromal tumors,the clinical venous imaging radiomics model had a good AUC value,with a training group AUC of 0.860 and a validation group AUC of 0.829.The Delong test showed a statistical difference between the clinical venous imaging radiomics model and the clinical arterial imaging radiomics model(p=0.028).Conclusions:1.The clinical-radiomics model constructed based on phase III dynamic enhanced CT was better than the single radiomics model and clinical model in predicting the malignant potential of small intestinal stromal tumor,mitotic count and Ki-67 expression,respectively.2.Based on the enhanced CT clinical-venous phase radiomics model,the performance of predicting the malignant potential,mitotic count and Ki-67 expression of small intestinal stromal tumor was better than that of other combined models,and the AUC values were 0.961,0.858 and 0.860.