Application of an enhanced CT nomogram prediction model in differentiating gastric stromal tumor ≤5 cm in diameter from gastric schwannoma
Objective To evaluate the application value of an enhanced computed tomography(CT)no-mogram prediction model in differentiating gastric stromal tumor(GST)≤5 cm in diameter from gastric schwannoma(GS).Methods This retrospective study analyzed the clinical data and enhanced CT image characteristics of 84 patients with GST and 23 patients with GS confirmed by surgical pathology and immuno-histochemistry at Yijishan Hospital between January 2018 and June 2023.A total of 23 relevant variables were included.Significant variables(P＜0.05)were screened through univariate analysis and used to construct a multivariate prediction model,with a corresponding nomogram being drawn.The diagnostic performance of the model was evaluated using the receiver operating characteristic(ROC)curve and its area under the curve(AUC),and the diagnostic efficacies of different variables and the nomogram were compared using the DeLong test.The accuracy of the model was internally validated using the calibration curve and further validated through 2-fold cross-validation.Results Univariate analysis identified eight significant variables:tumor mor-phology,site of growth,mode of growth,heterogeneity in the venous phase(SHRTv),enhancement rate in the venous phase(ERTv),enhancement rate in the delayed phase(ERTd),enhancement degree in the venous phase(DEv),and enhancement degree in the delayed phase(DEd).Among these,tumor morphology,site of growth,and DEd were independent influencing factors for differential diagnosis.The AUC of the nomogram for differentiating GST from GS was 0.894(95％CI:0.818～0.969),with a sensitivity of 79.8％and a speci-ficity of 91.3％.The calibration curve indicated good consistency between the model predictions and actual ob-servations.Conclusion The nomogram prediction model based on enhanced CT features demonstrates high performance in differentiating GST≤5 cm in diameter from GS,and its reliability has been confirmed through internal and cross-validation.
万方数据
