A nomogram model based on CT imaging features to predict the pathological risk classification of small intestinal stromal tumors
10.3760/cma.j.cn112149-20240514-00272
- VernacularTitle:基于CT影像特征的列线图模型预测小肠间质瘤病理危险度分级
- Author:
Ying XU
1
;
Weihua ZHI
;
Lu LI
;
Ze TENG
;
Huiqin ZHANG
;
Feng YE
;
Xinming ZHAO
Author Information
1. 国家癌症中心 国家肿瘤临床医学研究中心 中国医学科学院北京协和医学院肿瘤医院影像诊断科，北京100021
- Keywords:
Gastrointestinal stromal tumors;
Tomography, X-ray computed;
Pathological classification;
Prediction model
- From:
Chinese Journal of Radiology
2024;58(10):1063-1068
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To explore the value of the imaging nomogram model based on preoperative CT features of patients with small intestinal stromal tumor (SIST) in predicting pathological risk classification.Methods:This was a cohort study. The patients who were diagnosed as primary SIST by postoperative pathology in Cancer Hospital, Chinese Academy of Medical Sciences from January 2014 to October 2023 were retrospectively included. According to the modified 2008 National Institutes of Health classification criteria, the patients were divided into a pathological intermediate/high-risk group (86 cases) and a very low/low-risk group (56 cases). The features of preoperative enhanced CT images of SIST were analyzed, including tumor boundary, necrosis, intra-tumoral hemorrhage, intra-tumoral calcification, growth pattern, enhancement pattern, enhancement degree, enlarged vessels feeding or draining the mass (EVFDM), and tumor location. Patients were followed up to determine the recurrence-free survival (RFS). Univariate and multivariate logistic regression were used to screen the independent predictors of SIST with pathological medium/high-risk group. The independent predictors were combined to construct an imaging prediction model, and a nomogram was drawn. The receiver operating characteristic curve was used to evaluate the predictive efficacy of the model. The Kaplan-Meier method was used to draw the survival curve, and the log-rank test was used to compare the differences in RFS.Results:Univariate logistic regression results showed that tumor shape, necrosis, intra-tumoral hemorrhage, EVFDM, and tumor location were potentially related to medium/high-risk SIST. Multivariate logistic regression results showed that tumor shape ( OR=3.92, 95% CI 1.58-9.71, P=0.003), necrosis ( OR=4.60, 95% CI 1.91-11.09, P<0.001), and EVFDM ( OR=6.25,95% CI 1.74-22.47, P=0.005) were independent predictors of pathological intermediate/high-risk SIST. The area under the curve of the imaging predictive model combining the three predictors to predict the intermediate/high-risk SIST was 0.835 (95% CI 0.769-0.901), the sensitivity was 0.810, the specificity was 0.839, and the accuracy was 0.789. Taking the cut-off value (0.810) as the boundary value, the patients were divided into the high-risk group (74 cases) and the low-risk group (68 cases) according to the prediction results. The median RFS of the predicted high-risk group was poorer than that of the predicted low-risk group, and the difference was statistically significant ( χ2=5.20, P=0.023). Conclusions:The imaging nomogram model based on preoperative CT image features shape, necrosis, and EVFDM can effectively predict the pathological intermediate/high-risk SIST before surgery and has important predictive value for postoperative recurrence.
