摘要： 目的 探讨基于磁共振成像(MRI)T2加权像(T2WI)的海马和杏仁核影像组学方法诊断脑卒中后创伤后应激障碍(PTSD)的可行性。方法 回顾性分析146例脑卒中患者的临床资料、相关量表评分和MRI影像资料。在显示海马、杏仁核最大层面的T2WI图像上手动放置感兴趣区,分别提取274个纹理特征参数,采用LASSO回归降维并构建诊断脑卒中后PTSD与非PTSD的海马和杏仁核影像组学模型,采用受试者操作特征曲线评估海马、杏仁核影像组学模型在训练组和测试组中的诊断效能。结果 训练组和测试组中,脑卒中后PTSD与非PTSD患者比较,人口学和临床特征的差异无统计学意义(P>0.05),脑卒中后PTSD患者的改良年龄相关白质改变量表总评分、PTSD症状清单平民版评分均高于非PTSD患者(P<0.05)。LASSO回归分别筛选出最具鉴别意义的9个海马和14个杏仁核纹理特征参数,构建诊断脑卒中后PTSD和非PTSD的影像组学模型。训练组和测试组中,海马T2WI影像组学模型曲线下面积(AUC)分别为0.8920(95%CI:0.8332~0.9507)和0.7025(95%CI:0.5366~0.8683),灵敏度和特异度分别为82.35%和80.39%、68.18%和77.27%;杏仁核T2WI影像组学模型AUC分别为0.9214(95%CI:0.8736~0.9691)和0.7975(95%CI:0.6676~0.9275),灵敏度和特异度分别为90.20%和72.55%、63.64%和86.37%。结论 海马和杏仁核T2WI影像组学模型对脑卒中后PTSD具有较高的诊断效能,可作为一种新的影像学标志物用于脑卒中后PTSD的临床诊断和评估。

磁共振影像组学诊断脑卒中后创伤后应激障碍

赖树盛¹, 王晗², 褚建祎³, 张帆⁴, 郑石磊³

1. 中国医科大学附属第一医院放射科, 沈阳 110001;
2. 泰安市中心医院医学影像中心, 山东 泰安 271000;
3. 锦州医科大学附属第一医院放射科, 辽宁 锦州 121001;
4. 锦州医科大学附属第一医院神经内科, 辽宁 锦州 121001

收稿日期:2022-03-07出版日期:2022-07-30发布日期:2022-06-27
通讯作者:郑石磊E-mail:114305970@qq.com
作者简介:赖树盛(1984-),男,技师,大专.
基金资助:辽宁省教育厅青年科技项目(JYTQN2020013)


关键词: 磁共振成像, 影像组学, 脑卒中, 创伤后应激障碍, 精神疾病合并症
Abstract: Objective To investigate the feasibility of using a magnetic resonance imaging (MRI) radiomics to diagnose post-stroke post-traumatic stress disorder (PTSD) based on T2-weighted imaging (T2WI) of the hippocampus and amygdala. Methods The data of 146 stroke patients were retrospectively analyzed. Regions of interest (ROIs) were manually placed at the hippocampus and amygdala,and 274 texture features in each ROI were extracted. The feature selection and radiomics signature building to identify the hippocampus and amygdala radiomics signatures consisting of optimal features were performed using LASSO regression. Receiver operating characteristic curve was used to assess the diagnostic performance of the model. Results Compared with post-stroke non-PTSD patients,the age-related white matter changes scale and PTSD checklist-civilian version scores were higher in post-stroke PTSD patients (P< 0.05). The hippocampus and amygdala radiomics signatures were built using 9 and 14 optimal features,respectively. In the training and test groups, areas under the curve (AUCs) of the hippocampus radiomics signatures were 0.892 0 (95%CI:0.833 2-0.950 7) and 0.702 5 (95%CI:0.536 6-0.868 3),and sensitivity and specificity were 82.35% and 80.39% and 68.18% and 77.27%,respectively; AUCs of the amygdala radiomics signatures were 0.921 4 (95%CI:0.873 6-0.969 1) and 0.797 5 (95%CI:0.667 6-0.927 5),and sensitivity and specificity were 90.20% and 72.55% and 63.64% and 86.37%,respectively. Conclusion Hippocampus and amygdala T2WI radiomic signatures can aid in diagnosing post-stroke PTSD and could thus serve as a novel neuroimaging marker for the clinical diagnosis of post-stroke PTSD.
Key words: magnetic resonance imaging, radiomics, stroke, post-traumatic stress disorder, psychiatric co-morbidity
PDF全文下载地址:

https://journal.cmu.edu.cn/CN/article/downloadArticleFile.do?attachType=PDF&id=3027