Objective To develop and validate a nomogram based on CT features for predicting the mitotic index of gas-trointestinal stromal tumors(GIST)preoperatively.Methods The clinical and imaging data of 100 patients with GIST were retrospectively analyzed.According to the pathological data,the patients were divided into a low mitotic index group(63 cases)and a high mitotic index group(37 cases).The association of CT characteristics with mitotic index was analyzed using univariate and multivariate logistic regression analysis.A nomogram was developed for predicting the mitotic index.The predictive accuracy of the nomogram was determined by the area under the ROC curve(AUC),calibration curves and Hosmer-Lemeshow test.Results CT imaging features including tumor shape,the maximum diameter of tumor,ulceration,necrosis,the percentage of necrosis,enlarged vessels feeding or draining the mass(EVFDM),enhancement degree,en-hancement characteristic,metastasis and peripheral invasion showed significant differences between groups in univariate a-nalysis(all P<0.05).Only the maximum diameter of tumor,necrosis,ulceration and enhancement characteristic were in-dependent risk factors for high mitotic index of GIST.The nomogram model including these four independent factors a-chieved good AUC of 0.892(95%CI:0.829～0.956).The calibration curve and the results of Hosmer-Lemeshow test(P = 0.516)demonstrated good agreement between the predictions by the nomogram and the actual results.Conclusion The nomogram model consisting of the maximum diameter of tumor,ulceration,necrosis and enhancement characteristichas high value in predicting the mitotic index of GIST patientsand is helpful for clinicians to evaluate the recurrence and metastasis potential of GIST more comprehensively so as to make more accurate clinical decisions.
NETL
NSTL
万方数据
