Application Value Of Enhanced CT Based Radiomics In Prognosis Assessment Of Patients With Gastric Neuroendocrine Neoplasm

Objective:The traditional TNM staging system cannot reflect the biological heterogeneity of gastric neuroendocrine neoplasm(GNEN)and gastric neuroendocrine tumor(GNET)and gastric neuroendocrine carcinoma(GNEC)follow different TNM staging standards,which is insufficient to provide complete and accurate prognostic information and not convenient for clinical use.The purpose of this study is to investigate the efficacy of CT radiomics models of different phases in the prognosis assessment of GNEN patients,and to develop a radiomics signature corresponding to the radiomics model with the best prognostic efficacy.The further objective is to establish and validate a combined radiomics-clinical model combining radiomic signature with independent clinicopathological risk factors for individualized prediction of the prognosis of GNEN patients,and to quantitatively compare the prognostic efficacy of the model with the clinical model alone,radiomic model and traditional TNM staging.Materials and methods:A retrospective study of 182 patients with GNEN who underwent dual-phase enhanced scanning at two tertiary health-care institutions between August 2011 and December 2020 was conducted.Patients were randomly divided into the training group and the validation group in a 7:3 ratio.ITK-SNAP software was used to delineate the three-dimensional region of interest(ROI)of arterial phase and venous phase respectively,and after extracting the radiomics features,LASSO-Cox regression analysis was used to screen the features and establish the arterial phase radiomics model,the venous phase radiomics model and the arteriovenous phase combined radiomics model respectively.The radiomics model with the best prognostic performance among the three models was selected for subsequent analysis by calculating Harrell’s concordance index(C-index).Time-dependent receiver operator characteristics curve(ROC)was used to evaluate the accuracy of the radiomics model in predicting overall survival(OS)at different follow-up times.The radiomics signature,radiomics score(R-score),was developed based on the radiomics model with optimal prognostic performance.The potential association of radiomics signature with OS was first assessed in the training group and then validated in the validation group.The patients were divided into high-risk and low-risk group with the median R-score of the training group as the cut-off value.The Kaplan-Meier method and the Log-Rank test were used to estimate the OS and compare the survival curves.The independent clinicopathological risk factors associated with prognosis were selected by univariate and multivariate Cox regression analysis.We integrated radiomics signature with independent clinicopathological risk factors to construct a combined radiomics-clinical model and visualized it in the form of a nomogram.Using clinicopathological factors to construct a simple clinical model was also displayed in the form of a nomogram.C-index values were calculated to compare the prognostic value of the combined model,the clinical model,radiomic model,and traditional TNM staging.Calibration curve was used to analyze the agreement between the predicted and actual values of the combined model,and decision curve was used to evaluate the clinical efficacy of the various models.Results:The arteriovenous phase combined radiomics model had the best performance in predicting OS,and its C-index values in the training group and the validation group were 0.803[95%confidence interval(CI)0.741～0.864,P<0.001]and 0.751(95%CI 0.661～0.841,P<0.001)respectively,better than the independent arterial phase radiomics model(C-index values were 0.784 and 0.747 respectively)and the venous phase radiomics model(C-index values were 0.756 and 0.723 respectively).Time-dependent ROC analysis showed that the area under the curve(AUC)values of the arteriovenous phase combined radiomics model to predict OS at 18 months,24 months,and 30 months in the training group was 0.829,0.838 and 0.909,respectively,all greater than 0.800;The AUC values was also higher in the validation group,0.835,0.839,and 0.864,respectively.The combined radiomics model of the arteriovenous phase included 14 radiomics features,and the corresponding radiomics signature was significantly associated with OS in the training group[hazard ratio(HR)=5.015;95%CI 3.602～5.982,P<0.001],this result was confirmed in the validation group(HR=6.829;95%CI 2.661～17.527,P<0.001).GNEN patients could be successfully divided into high and low prognostic risk groups with R-score median-0.4939 as the cut-off value,and the difference between the two groups was statistically significant(P<0.001).After univariate and multivariate Cox regression analysis,it was found that the arteriovenous phase combined radiomics signature was an independent risk factor for prognosis.The combined radiomics-clinical model combining this radiomics signature and independent clinicopathological risk factor(sex,age,treatment methods,T stage,N stage,M stage,tumor boundary,Ki67,CD56)had the best prognostic performance,its C-index value was 0.882(95%CI 0.848～0.916)in the training group and 0.827(95%CI 0.765～0.888)in the validation group,both of which were better than the clinical model alone(C-index values were 0.861 and 0.796,respectively),radiomics model(C-index values were 0.803 and 0.751,respectively),and traditional TNM staging system(C-index values were 0.870 and 0.801,respectively).The calibration curve showed that there was significant consistency between the predicted survival rate and the actual survival rate of the combined radiomics-clinical model,and the decision curve analysis verified that the combined radiomics-clinical model had high clinical benefit and application value in clinical practice.Conclusions:1.The arteriovenous phase combined radiomics model was more effective than the independent phase radiomics model in predicting OS of GNEN patients,and the corresponding radiomics signature could successfully divide the prognosis of GNEN patients into high-risk group and low-risk group.2.This study established and validated a combined radiomics-clinical model to predict the OS of GNEN patients based on radiomics signature and independent clinicopathological risk factors and visualized it in the form of a nomogram.The model was expected to serve as a potential tool for individualized prognostic assessment of GNEN patients.3.The combined radiomics-clinical model outperformed the clinical model alone,radiomics model and traditional TNM staging system in predicting OS,and demonstrated the incremental value of the radiomics signature for individualized assessment of OS in GNEN patients,which could provide effective information for clinical decision making.