基于CT图像的决策树模型在腹内疝诊断和鉴别诊断中的价值  

Value of decision tree model based on CT images in the diagnosis and differential diagnosis of internal hernia

作　　者：王子 王嵩[2] WANG Zi;WANG Song(Department of Radiology,Longhua Hospital Affiliated of Shanghai University of Traditional Chinese Medicine,Shanghai 200032,China)

机构地区：[1]上海核工程研究设计院股份有限公司,上海200233 [2]上海中医药大学附属龙华医院放射科,上海200032

出　　处：《中国中西医结合影像学杂志》2023年第2期184-187,共4页Chinese Imaging Journal of Integrated Traditional and Western Medicine

基　　金：上海市自然科学基金项目(19ZR1457800)。

摘　　要：目的:分析总结腹内疝的MSCT征象,构建并验证诊断腹内疝的决策树模型。方法:回顾性收集42例经手术证实的腹内疝患者(腹内疝组)和76例其他肠道疾病患者(对照组),并按照7∶3的比例将118例随机划分为训练集(n=82)和验证集(n=36)用于决策树模型的构建。由2名影像医师分析图像,分别记录6种CT征象:肠管异位聚集征、鸟嘴征、蘑菇征、血管纠集征、靶征和有无肠梗阻。计算各征象的频数和频率,并将各个征象纳入决策树模型的不同节点,然后采用ROC曲线的AUC评估各个征象的诊断性能。结果:42例腹内疝患者的CT征象出现频率从高到低依次为肠梗阻(85.7%)、鸟嘴征(76.2%)、肠管异位聚集征(64.3%)、血管纠集征(52.4%)、蘑菇征(33.3%)、靶征(26.2%)。在决策树模型中,血管纠集征和蘑菇征的联合征象具有最高的诊断价值(验证集AUC 0.74,95%CI 0.57~0.91)。结论:该决策树模型在腹内疝和其他疾病鉴别诊断中具有重要的应用价值,可为临床决策提供可靠依据。Objective:To analyze and summarize the MSCT features of internal hernia,and to construct and validate a decision tree model for the diagnosis.Methods:42 patients with surgically confirmed internal hernia(the internal hernia group)and 76 patients with other intestinal diseases(the control group)were retrospectively collected,and randomly divided into a training set(n=82)and a verification set(n=36)according to a ratio of 7∶3 for the construction of decision tree model.Two radiologists analyzed the abdominal CT images of each patient and recorded six CT signs,the presence or absence of ectopic aggregation of clustered loops,beak sign,mushroom sign,swirl sign,target sign and intestinal obstruction.The frequencies of each sign were calculated,and each sign was included in different nodes of the decision tree model,then ROC curve and the AUC were used to assess the diagnostic performance of each sign.Results:The frequency of CT signs in 42 patients with intra hernias was intestinal obstruction(85.7%),beak sign(76.2%),clustered loops(64.3%),swirl sign(52.4%),mushroom sign(33.3%)and target sign(26.2%).In the decision tree model,the combined sign of vascular aggregation sign and mushroom sign had the highest diagnostic value(validation set,AUC of 0.74,95%CI 0.57~0.91).Conclusions:The decision tree model has an important application value in the differential diagnosis of internal hernia and other diseases,and can provide a reliable basis for clinical decision-making.

关 键 词：腹内疝 决策模型 体层摄影术 X线计算机 

分 类 号：R656.2[医药卫生—外科学] R816.5[医药卫生—临床医学]