Semi-synthetic Trauma Resuscitation Process Data Generator.
Semi-synthetic Trauma Resuscitation Process Data Generator.
复制标题
半合成创伤复苏过程数据生成器。
DOI：
10.1109/ichi.2017.67
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
Burd,RandallS
中科院分区：
文献类型：
--
作者：
Yang,Sen;Zhou,Yichen;Guo,Yifeng;Farneth,RichardA;Marsic,Ivan;Burd,RandallS
过程挖掘技术已被应用于医疗过程的可视化、解释和分析。然而，这些分析所需的过程数据只有非常有限的数量是公开的，特别是在医疗领域，因为患者的隐私。这限制了新的医疗过程研究使用不够大或随机生成的合成数据集。我们在这项研究中的目标是训练一个模型（使用有限数量的观察到的过程数据），可以生成大量的半合成过程数据。该生成的数据具有与真实世界的过程数据相似的特征，并且可能在现实中观察到。本研究使用的医学数据得到了华盛顿儿童国家医学中心机构审查委员会的批准。从创伤复苏的监控视频中编码了122例创伤复苏病例，总计4347次活动。在我们的方法中，我们试图从观察到的过程轨迹Ob中学习描述性HMM λ，并使用λ来产生半合成过程数据。我们的方法可以总结为两个步骤（Alg. 1）、
Process mining techniques have been applied to the visualization, interpretation, and analysis of medical processes. However, only a very limited amount of process data necessary for these analyses is publicly available, especially in the medical field because of patients’ privacy. This limits novel medical process research to using insufficiently large or randomly-generated synthetic datasets. Our goal in this study is to train a model (using a limited amount of observed process data) that can generate large amounts of semi-synthetic process data. This generated data has characteristics similar to those of real-world process data, and could potentially be observed in reality.The use of medical data for this study was approved by the Institutional Review Board at the Children’s National Medical Center in Washington, DC. One hundred and twenty-two trauma resuscitation cases totaling 4347 activities were coded from surveillance videos of trauma resuscitations. In our approach, we tried to learn a descriptive HMM λ from observed process traces Ob and use λ to produce semi-synthetic process data. Our method can be summarized in two steps (Alg. 1):