基于超声征象的决策树模型在预测胃肠道间质瘤危险度中的应用价值

背景与目的：胃肠道间质瘤（gastrointestinal stromal tumor,GIST）的生物危险度可分为极低危、低危、中危和高危,术前就危险度进行准确预判对临床制订诊疗方案至关重要。探讨基于超声征象的决策树模型在预测GIST危险度中的应用价值。方法：收集福建医科大学附属协和医院2016年12月—2020年12月收治的206例GIST患者,建立决策树模型,并判断模型预测的准确度。结果：低危、中危和高危3组间病灶长径、短径/长径（short diameter/long diameter,S/L）、肿瘤部位、回声高低、回声均匀性、边界、形态、是否坏死囊变和血流信号的差异均有统计学意义（P均<0.05）。以病灶长径、肿瘤部位、回声均匀性及形态为节点建立的决策树模型预测的准确度为72.33%,其中低危组和高危组预测的准确度可高达80.90%和93.90%。结论：以超声特征为主要指标构建的决策树模型在预测GIST危险度中具有较高的应用价值。

Application value of decision tree model based on ultrasonic signs in predicting the risk grade of gastrointestinal stromal tumors

Background and purpose: The biological risk of gastrointestinal stromal tumor (GIST) can be divided into very low-risk, low-risk, moderate-risk and high-risk. Accurate preoperative prediction of risk is very important for clinical diagnosis and treatment. This study explored the application value of decision tree model based on ultrasonic signs in predicting the risk grade of GIST. Methods: A total of 206 GIST patients treated in Union Hospital Affiliated to Fujian Medical University from December 2016 to December 2020 were collected. The decision tree model was established, and the prediction accuracy of the model was calculated. Results: There were statistically significant differences in long diameter, short diameter/long diameter (S/L), location, internal echo, echo uniformity, boundary, shape, cystic necrosis and blood flow signals among low-risk, moderate-risk and high-risk groups (all P<0.05). The prediction accuracy of the decision tree model based on the long diameter, location, echo uniformity and shape was 72.33%, and the prediction accuracies of the low-risk group and high-risk group were 80.90% and 93.90%, respectively. Conclusion: The decision tree model based on ultrasonic signs has high application value in predicting the risk grade of GIST.