基于增强CT特征的列线图模型在预测胃肠道间质瘤危险度分级中的价值

A nomogram model based on enhanced CT features for risk-grade prediction of gastrointestinal stromal tumors

目的探讨基于增强CT图像特征的列线图模型在预测胃肠道间质瘤(GIST)危险度分级中的价值。方法回顾2017年1月至2021年6月湖州市中心医院行增强CT检查并经内镜或手术病理检查证实为GIST的135例患者资料。根据病理分级将患者分为潜在恶性组74例(极低危险度43例,低危险度31例)和恶性组61例(中危险度24例,高危险度37例)。对两组间的增强CT特征进行统计学分析,将差异有统计学意义的特征纳入多因素logistic回归分析,筛选出预测GIST危险度分级的独立危险因素,并构建列线图预测模型。结果两组患者的肿瘤最大径、部位、生长方式、边界、形态、液化坏死、强化方式、静脉期CT值比较差异均有统计学意义(均P<0.05)。多因素logistic回归分析结果显示,肿瘤最大径(OR=2.636,95%CI:1.180~5.890,P<0.05)、形态(OR=0.055,95%CI:0.005~0.570,P<0.05)、液化坏死(OR=0.042,95%CI:0.004~0.434,P<0.05)是预测GIST危险度分级的独立危险因素。利用该3个特征构建术前预测GIST危险度分级的列线图模型,其AUC为0.952,灵敏度为0.950,特异度为0.833,校准曲线与标准曲线拟合度良好。结论基于增强CT征象的列线图模型对预测GIST危险度分级有较高的价值,可为临床提供一种比较精确的术前量化预测方法。

Objective To explore the value of a nomogram model based on enhanced CT features for predicting the risk-grade of gastrointestinal stromal tumor(GIST).Methods Clinical and imaging data of 135 patients with GIST confirmed by biopsy or surgery in Huzhou Central Hospital from January 2017 to June 2021 were retrospectively analyzed.According to the histopathological grading,there 74 cases of potential malignancy,including 43 cases of very low risk and 31 cases of low risk,and 61 cases of malignancy,including 24 cases of moderate risk and 37 cases of high risk.Enhanced CT features were compared between the potentially malignant and malignant patients.The independent predictors for the risk-grade of GIST were identified by multivariate logistic regression analysis,and a nomogram model for risk-grade prediction of GIST was established.Results There were significant differences in tumor maximum diameter,CT value of venous phase,location,growth pattern,boundary,shape,liquefaction necrosis and enhancement way between the potentially malignant and malignant groups.Multivariate logistic regression analysis showed that tumor maximum diameter(OR=2.636,95%CI:1.180-5.890,P<0.05),shape(OR=0.055,95%CI:0.005-0.570,P<0.05)and liquefaction necrosis(OR=0.042,95%CI:0.004-0.434,P<0.05)were independent predictors for risk-grade of GIST.The area under the receiver operating characteristic curve(AUC)of the established nomogram model was 0.952,the sensitivity was 0.950,the specificity was 0.833,and the calibration curve fitted well with the ideal curve.Conclusion The established nomogram model based on enhanced CT features is effective in predicting the risk-grade of gastrointestinal stromal tumors and can provide a more accurate preoperative quantitative prediction method in clinical practice.