急性胰腺炎状态下早期肝损伤的CT影像组学研究  

CT-Based Radiomics Research of Early Liver Injury in Acute Pancreatitis

作　　者：方杰[1] 黄小华[1] 明兵[2] 唐玲玲 胡云涛 刘念[1] FANG Jie;HUANG Xiaohua;MING Bing;TANG Lingling;HU Yuntao;LIU Nian(Department of Radiology,Affliated Hospital of North Sichuan Medical College,Nanchong637000,China;Department of Radiology,Deyang People's Hospital)

机构地区：[1]川北医学院附属医院放射科,南充637000 [2]四川省德阳市人民医院放射科

出　　处：《中国医学计算机成像杂志》2022年第5期491-496,共6页Chinese Computed Medical Imaging

基　　金：四川省卫生健康科研课题(19PJ203);南充市市校科技合作战略合作项目(19SXHZ0429);南充市市校科技战略合作(20SXHZ0303)。

摘　　要：目的:探讨急性胰腺炎(AP)状态下基于肝脏灌注CT纹理分析的影像组学模型评估早期肝损伤的临床价值。方法:回顾性分析川北医学院附属医院与德阳市人民医院共144例经临床诊断为AP的患者的灌注CT图像,其中肝损伤患者77例,非肝损伤患者67例。按照7∶3的比例划分为训练组(99例)及验证组(45例)。使用分割软件在灌注CT图像上进行特征提取,使用最小绝对收缩与选择算子(LASSO)最小方差法进行特征降维,保留最佳影像组学特征;比较肝损伤组与非肝损伤组的临床特征,遴选2组间有统计学差异的指标。采用logistic回归分析构建影像组学模型及其与临床特征相结合的联合模型,采用受试者工作特征(ROC)曲线下面积(AUC)评价模型的实用性与应用价值。结果:特征降维后生成包含11个影像组学特征的特征集,所建立的影像组学模型,其训练组AUC值为0.863 (特异度0.935、灵敏度0.698),验证组AUC值为0.732 (特异度0.714、灵敏度0.75)。将肝损伤组与非肝损伤组间有统计学差异的血清淀粉酶指标与影像组学特征结合所构建的联合模型,其训练组AUC值为0.861 (特异度0.848、灵敏度0.774),验证组AUC值为0.764 (特异度0.714、灵敏度0.792)。结论:AP状态下基于肝脏灌注CT纹理分析的影像组学模型在评估早期肝损伤方面具有潜在价值,能为临床评估AP患者预后及指导其个体化治疗提供新的方法。Purpose:To investigate the clinical value of the radiomics model based on CT texture analysis of liver perfusion in evaluating early liver injury in acute pancreatitis(AP).Methods:The perfusion CT images of 144patients clinically diagnosed as AP in the Affiliated Hospital of North Sichuan Medical College and Deyang People’s Hospital were analyzed retrospectively,including 77 patients with liver injury and 67 patients without liver injury.According to the ratio of 7∶3,they were divided into training group(99 cases)and testing group(45 cases).The segmentation software was used to extract features on perfusion CT images,and the least absolute shrinkage and selection operator(LASSO)minimum variance method was used to reduce the dimension of features,so as to retain the best radiomics features.The clinical characteristics of liver injury group and non-liver injury group were compared,and the indexes with statistical difference between the two groups were selected.Logistic regression analysis was used to construct the radiomics model and combined model with clinical characteristics,and the area under the ROC curve(AUC)was used to evaluate the practicality and application value of the model.Results:After feature dimensionality reduction,a feature set containing 11 radiomics features was generated.The established radiomics model had an AUC value of 0.863(specificity 0.935,sensitivity 0.698)in the training group and 0.732(specificity 0.714,sensitivity 0.75)in the validation group.The combined model was constructed by combining the serum amylase which were statistically different between the liver injury group and the non-liver injury group with radiomics characteristics,and had an AUC value of 0.861(specificity 0.848,sensitivity 0.774)in the training group and 0.764(specificity 0.714,sensitivity 0.792)in the validation group.Conclusions:The radiomics model based on CT texture analysis of liver perfusion in AP state has potential value in evaluating early liver injury,and can provide a new method for clinical evaluatio

关 键 词：影像组学 急性胰腺炎 肝损伤 灌注CT 

分 类 号：R445.3[医药卫生—影像医学与核医学]