基于RWS的晕厥临床不良事件风险评估与智能预警关键技术研究

Epidemiology shows that there are more than 40 million patients with severe syncope in China, and the prevalence of syncope in the emergency department is as high as 41%. Clinical adverse events such as falls in the elderly that causes severe health injury and rehabilitation problems. As the aging society coming in China, the social-economic problems caused by syncope are becoming more and more prominent. Thus, the risk assessment and early warning of syncope adverse events have significant scientific and clinical value. With the rapid development of medical informatics, more and more attention has been paid to Real World Study (RWS). Compared to the cohort study, RWS has no strict criteria of test-object inclusion, and the data is plentiful and reflects the real clinical environment. Right now, RWS is labeled as a new medical research method and knowledge discovery direction. This project focuses on the risk assessment and intelligent early warning technology of clinical adverse events caused by syncope. In this project, we first establish a syncope clinical medical knowledge base. Secondly, we structure the electronic medical records through natural language processing (NLP), and then construct a syncope risk assessment model based on multi-source multi-modal medical data sources. Finally, our research results are expected to help pathogeny-unclear syncope patient in-hospital or out-of-hospital with risk early-warning by finding abnormal ECG signals intelligently.

流行病学显示我国严重晕厥患者超过4000万，在急诊中晕厥的患病率高达41%。晕厥的临床不良事件如老年人摔倒等，给患者带来严重的健康伤害与康复问题。随着我国进入人口老龄化时代，晕厥引起的社会民生问题日益凸显，对晕厥不良事件进行风险评估和预警具有重大科学和临床价值。.随着医疗信息化的快速发展，基于真实世界研究（Real World Study,RWS）引起越来越多的关注。RWS相对于队列研究没有严格的入组标准，数据反应真实临床环境；数据丰富，可用患者信息多，代表了一种新的医学研究发展方向。.本课题围绕着晕厥引起临床不良事件的风险评估及智能预警技术进行深入探讨。拟建立基于临床指南的晕厥医学知识库，通过自然语言处理（NLP）对电子病历进行结构化处理，之后基于RWS的多源多模态医疗数据构建晕厥风险评估模型，进而对于不明原因晕厥患者通过异常心电信号（ECG）智能判读，实现院内或院外预警。

随着大数据、人工智能等新兴技术群的兴起，推动了过去基于队列等的循证医学研究向数据驱动的真实世界研究（Real-World Study（RWS）的转变。RWS相对于队列研究没有严格的入组标准，数据反应真实临床环境；数据丰富，可用患者信息多，代表了一种新的医学研究发展方向。本课题围绕着临床不良事件的风险评估及智能预警技术进行深入探讨，主要开展以下研究：大型、多中心的多模态医疗数据集整合和标注构建，与国内三甲医院合作，汇聚和清洗了数十万份急诊病历、心电数据等基于高效医学知识图谱构建及标准化；研究基于语义推断的医学信息抽取框架，对电子病历进行结构化处理；对于不明原因晕厥等心律失常患者通过异常心电信号（ECG）智能判读，实现对于院内或院外预警，构建基于RWS的多源多模态医疗数据构建晕厥风险评估模型及临床验证体系。通过本项目研究，为探索基于人工智能技术的疾病早筛早治和动态病情管理提供经验与技术积累。