基于影像组学急性百草枯中毒的预后研究

《基于影像组学急性百草枯中毒的预后研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、授予单位代码10089学号或申请号20153301中国图书分类号R445.3硕士学位论文专业学位基于影像组学急性百草枯中毒的预后研究研究生：卢姗导师：耿左军教授专业：影像医学与核医学二级学院：河北医科大学第二医院2018年3月目录中文摘要·············································································································1英文摘要······················

2、·······················································································3英文缩写·············································································································5研究论文基于影像组学：急性百草枯中毒的预后研究前言····························

3、················································································6材料与方法································································································7结果······································································

4、·····································11附图··········································································································13附表········································································································

5、··19讨论··········································································································21参考文献··································································································27综述百草枯中毒的预后研究现状·······················

6、··········································33致谢····················································································································46个人简历·······················································································

7、····················47中文摘要基于影像组学：急性百草枯中毒的预后研究摘要目的：建立和评估影像组学模型在预测急性百草枯中毒患者预后中的有效性。方法：数据来自2014年11月至2017年10月的80例明确诊断为急性百草枯中毒的患者的早中期胸部CT图像及相关临床资料，按7：3比例分层随机抽样分配为训练组及验证组。训练组（57例）用以建立预测模型，独立验证组（23例）用以模型验证。选择肺内病变进展高峰的CT图像，勾画全肺为ROI，提取影像组学特征，使用PCA及套索回归方法降维、选择关键特征并

8、建立影像组学标签。纳入影像组学标签及临床预后危险因子，采用多变量逻辑回归分析建立影像组学标签结合临床预后危险因子的综合预测模型，模型结果用列线图表示。并从区分度、校准度和临床有用性方面对列线图进行了评估。结果：7个关键特征组成的影像组学标签在训练数据集和验证数据集中生存组和死亡组之间具有显著统计学差异（P<0.001）。影像组学标签在训练数据集和验证数据集预测病人预后的AUC分别为0.942（95%CI0.886-0.997）