脾脏影像纹理特征预测肝硬化患者食管静脉曲张严重程度的研究  

Spleen texture features on CT image in predicting the degree of esophageal varices in patients with liver cirrhosis

作　　者：张华妮 李嘉颖 周敏辉 陈海波[2] 孙志超[2] ZHANG Huani;LI Jiaying;ZHOU Minghui;CHEN Haibo;SUN Zhichao(the First School of Clinical Medicine of Zhejiang Chinese Medical University,Hangzhou 310053,China;不详)

机构地区：[1]浙江中医药大学第一临床医学院,杭州310053 [2]浙江中医药大学附属第一医院(浙江省中医院)医学影像科

出　　处：《浙江医学》2023年第5期470-474,I0004,共6页Zhejiang Medical Journal

基　　金：浙江省医药卫生科技计划项目(2020KY671)。

摘　　要：目的 探讨脾脏影像纹理特征预测肝硬化食管静脉曲张严重程度的临床应用价值。方法 选取2015年1月至2021年5月浙江中医药大学附属第一医院收治的肝硬化食管静脉曲张患者73例,以内镜检查结果为金标准,将患者分为低危食管静脉曲张组23例和高危食管静脉曲张组50例。在脾脏CT平扫图像上手动勾画感兴趣区并提取纹理特征,筛选出与肝硬化食管静脉曲张严重程度相关的脾脏影像纹理特征参数,并分析其构建的模型对肝硬化食管静脉曲张严重程度的诊断效能。结果最终筛选获得8个脾脏纹理特征参数与食管静脉曲张严重程度有相关性(r=0.282~0.437,均P<0.05)。基于筛选的纹理特征参数构建的LinearSVC影像组学预测模型,训练组AUC、灵敏度、特异度分别为0.870(95%CI:0.773~0.967)、0.823、0.765,验证组分别为0.876(95%CI:0.723~1.000)、0.867、0.714,有较高的预测价值。结论 脾脏影像纹理特征模型对肝硬化食管静脉曲张严重程度有一定的预测价值,可以辅助临床医师为肝硬化食管静脉曲张患者早期精准制定个体化诊疗方案。Objective To investigate the clinical value of spleen texture features on CT image in predicting the degree of esophageal varices in patients with liver cirrhosis.Methods Clinical data and CT imaging finding of 73 patients with esophageal varices diagnosed by gastroscopy from January 2015 to May 2021 were retrospectively analyzed.According to results of gastroscopy,patients were divided into low-risk esophageal varices group(n=23)and high-risk esophageal varices group(n=50).The ROI was manually sketched and the texture features were extracted from non-contrast enhanced CT.The spleen image texture feature parameters related to the degree of esophageal varices in liver cirrhosis were screened,and the diagnostic performance of the model constructed by these parameters for the degree of esophageal varices was evaluated.Results Eight spleen texture features parameters were correlated with the degree of esophageal varices(r=0.282-0.437,P<0.05).The LinearSVC radiomics prediction model was constructed based on the texture features parameters.The area under the ROC curve(AUC),sensitivity and specificity of the established LinearSVC model in the training group were 0.870(95%CI:0.773-0.967),0.823 and 0.765,respectively;in the validation group were 0.876(95%CI:0.723-1.000),0.867,and 0.714,respectively.Conclusion The LinearSVC model based on spleen CT images has a high predictive value for the degree of esophageal varices in patients with liver cirrhosis,which is helpful for clinicians to formulate early and accurate individualized treatment plan for cirrhotic patients with esophageal varices.

关 键 词：肝硬化 食管静脉曲张 影像组学 脾脏 计算机断层扫描 

分 类 号：R575.2[医药卫生—消化系统] R571[医药卫生—内科学]