摘要
目的:基于弥散加权成像(diffusion weighted imaging,DWI)探讨急性缺血性脑卒中患者影像组学特征与预后的关系。方法:选取2018年1月—2020年12月在云南省滇南中心医院就诊的232例急性脑梗死患者,收集并记录入院时患者的ASPECTS评分、24 h内磁共振DWI序列图像、经机械取栓或溶栓治疗后90d病人的改良Rankin量表(modified rankin scale,mRS)评分。使用3D slicer软件勾画梗死灶所有层面的ROI,使用SlicerRadiomics获得影像组学特征。经过独立样本t检验、单因素Logistics回归分析、LASSO回归和5折交叉验证选择预测效能最高的影像特征。使用支持向量机算法(support vector machine,SVM)建立、优化预测模型。结果:107个影像组学特征经降维筛选后选出11个预测效能最高的特征参数。基于DWI预测模型训练集曲线下面积(area under curve,AUC)、特异度分别为0.901、0.914。基于DWI预测模型测试集AUC、特异度分别为0.854、0.939。结论:基于DWI的影像组学分析可以提供多个参数用于预测90天后患者神经功能恢复状况。影像组学预测模型具有较好的预测效能。
Objective To investigate the correlation between imaging omics characteristics and outcome in patients with acute ischemic stroke based on diffusion weighted imaging(DWI).Methods 232 patients with acute cerebral infarction treated in Southern Central Hospital of Yunnan Province from January 2018 to December 2020 were selected.The aspects scores of patients at admission DWI images of magnetic resonance imaging within 24 hours,and the modified Rankin Scale(MRS) scores of patients 90 days after mechanical thrombectomy or thrombolysis were collected.ROI at all levels of infarctions was delineated using 3D slicer software,and imaging omics characteristics were obtained using SlicerRadiomics.The imaging omics characteristics with the highest predictive efficacy were selected by independent sample t-test,univanate logistics regression analysis,LASSO regression and 5-fold cross-validation.Support vector machine was used to establish and optimize the models.Results The A UC,specificity of the model in predicting outcome based on training set were 0.901,0.914.The A UC,specificity of the model in predicting outcome based on test set were 0.854,0.939.Conclusion DWI-based imaging omics analysis can provide multiple parameters to predict patient neurological recovery after 90 days.The imaging omics predictive model can effectively predict outcome.
作者
唐太松
吴丽霞
熊巧玲
TANG Taisong;WU Lixia;XIONG qiaoling(Department of Medical Imaging,Southern Central Hospital of Yunnan Province<The First People's Hospital of Honghe State>,Mengzi,Yunnan 661199,China)
出处
《影像研究与医学应用》
2022年第19期37-39,43,共4页
Journal of Imaging Research and Medical Applications
关键词
急性缺血性脑卒中
弥散加权成像
影像组学
机器学习
预后
Acute ischemic stroke
Diffusion weighted imaging
Imaging omics
Machine learning
Prognosis
