重症监护病房紧急状况预警算法  

Early Diagnosis Algorithms for ICU Emergencies

作　　者：李开士[1,2] 樊建平[1] 周丰丰[1,2] 

机构地区：[1]中国科学院深圳先进技术研究院,深圳518055 [2]中国科学院健康信息学重点实验室,深圳518055

出　　处：《集成技术》2012年第2期13-19,共7页Journal of Integration Technology

基　　金：周丰丰研究员在中国科学院深圳先进技术研究院的启动资金(2011年8月开始)支持

摘　　要：重症监护病房中的病人身体状况通常很不稳定,常出现各种需要医护人员介入治疗的紧急状况。由于医疗资源有限,医护人员可能无法及时发现并处理这些紧急状况,给病人的存活率带来严重的负面影响。如果可以预测这些紧急状况的发生,并及时通知相关医护人员,将大大提高病人的存活率。常见重症监护病房紧急状况包括突然死亡、败血症、肺部感染、急性低血压、以及器官衰竭等。紧急状况预警建模主要采用病人的长时间生命体征监测数据,预测在一定时间之后发生某种紧急状况的可能性。预警模型所采用的监测数据分为静态数据、事件数据和时间序列数据等三类。静态数据具有容易采集、但预测准确性偏低的特点。事件数据或时间序列数据、以及多种类型数据的混合数据对于紧急状况预警模型的预测性能的提高有重要作用,将会获得更广泛的应用。The vital signs of patients in ICU are usually very unstable. This requires immediate assistance from medical personnel. Due to the limited resources, not all the emergencies were handled in time, leading to unexpected fatal outcomes. Most cases like these can be avoided, if they are predicted and the medical assistance is provided before the emergencies occur. Common emergencies include sudden death, septicemia, lung infection, acute hypotension, and organ failure. Current models based on the monitored physiological data can provide sensitive predictions for some emergency types. There are three types of commonly used data, i.e. static data, event data and time series data. The static data is easily obtained, but leads to less accurate predictions than the event data and time series data do. It is expected, that the interest for collecting and processing event data and time series data will grow in the near future.

关 键 词：重症监护病房(ICU) 病人监测 紧急状况 数据挖掘 异构数据 预警模型 

分 类 号：R459.7[医药卫生—急诊医学]