基于多模态影像学的缺血性脑血管病疗效预测模型研究

Ischemic stroke has a high morbidity and high frequency of disability. It is crucial to prevent occurrence of ischemic stroke by variable effective treatment strategies based on variable mechanisms. It remains the most important challenge in this field. Our previous works revealed that cerebral perfusion imaging, vessel wall imaging, and other imaging modalities were able to assess cerebral hemodynamics, intrcranial arterial atherosclerotic plaque, collateral circulation and other characteristics closely associated with ischemic stroke. In this study, we prospectively enrolled patients with artery atherosclerosis and conducted multi-model MRI which especially equipped three dimensional whole brain pseudo-continuous arterial spin labeling (3D pCASL), territorial arterial spin labeling (TASL), vessel wall high resolution MRI (VW-HRMRI), and susceptibility weighted imaging (SWI). Based on the clinical data and imaging features, by using the data mining techniques, we tried to extract the predictive imaging markers from the imaging cloud data, and then quantify the predictive weight of these imaging markers in variable stages of stroke. This study aim to solve the scientific challenges in ischemic stroke field including the ischemic penumbra and collateral flow in dynamic evolution of ischemic stroke by using multimodel imaging technology combined with novel Radiomics approaches, and then to establish the predictive model to make treatment decision under precise treatment strategy, as well as to build the optimal protocol of secondary prevention for ischemic stroke.

缺血性脑血管病发病率和致残率高，针对不同致病机制进行有效治疗和预防复发是关键，也是当前面临的挑战。我们的前期研究表明，脑灌注和血管管壁磁共振成像等影像技术能够评估脑血流、动脉粥样硬化斑块、侧支循环等与卒中密切相关的指标。本研究以动脉粥样硬化所致缺血性脑血管病为研究对象，拟采用三维动脉自旋标记全脑灌注（3D pCASL）及分区灌注成像（TASL）、高分辨管壁成像（VW-HRMRI）和磁敏感加权成像（SWI）等多模态磁共振成像方法，应用大数据挖掘等技术，从影像中提取并量化海量特征数据，结合临床信息及影像学宏观特征，研究卒中发展过程中不同模态的影像学表征。本课题旨在通过多模态影像学结合影像组学（Radiomics）的研究新模式，深入探讨缺血半暗带、侧支循环建立及其动态演变的特征等科学问题，建立全面准确的疗效预测模型，为制定精准化的治疗策略提供依据，促进卒中二级预防的完善。

缺血性脑血管病发病率和致残率高，针对不同致病机制进行有效治疗和预防复发是关键，也是当前面临的挑战。课题组按照研究计划，采用高分辨管壁成像、动脉自旋标记全脑灌注成像及磁敏感加权成像等多模态磁共振成像方法，结合机器学习技术，围绕侧支、血流、斑块等与卒中密切相关的指标，开展了两项基于多模态影像学评估缺血性卒中的多中心研究，涵盖了全国23家合作单位，超额完成项目计划的各项指标。建立了缺血性脑血管病影像和临床数据库，完善了侧支-血流-斑块评价标准，提出了多模态磁共振成像最佳扫描策略；首次观察到颅内动脉栓塞急性期及再通后侧支循环动态变化规律，提出了量化低灌注新模式评价第三级侧支；提出无对比剂无辐射评价缺血半暗带最优成像方法，有望替代金标准CT灌注成像筛选卒中治疗获益患者；发展了临床适用快速成像新技术，在保证图像精度的前提下，将探测脑出血的磁敏感加权成像时间从5分钟缩短到1分钟之内，加快了院内检查流程；提出FVH-ASPECT、侧支循环/梗死核心比率、低灌注容积比等系列预测预后影像新指标，构建了疗效和预后预测模型；构建了临床适用智能诊断新模型，用于脑出血、卒中等常见疾病诊断。本项目共发表基金标注论著48篇，代表性论著包括Lancet digital health、Stroke等。国际国内大会交流31次，培养研究生12名。以第一发明人授权申请国家发明专利7项，软件著作权2项，研发了一套具有自主知识产权软件系统并在全国百余家医院临床试用。课题组受国家卫健委邀请撰写了《脑血管病影像指导手册》、《中国脑卒中防治指导规范》、《脑血管病影像规范化应用中国指南》，用于规范我国脑血管病影像流程，明确各级卒中防治单位职责；出版了人民卫生出版社教材《脑卒中影像学评估》。本项目研究成果为制定脑卒中精准化治疗策略提供了依据，促进卒中二级预防的完善，推动了我国脑血管病影像同质化建设。