对于医生而言，跟踪病人的化验结果、各项图表以及其它各项指标是一件非常耗时，但又必须要做的事：
For doctors, tracking patients' results, graphs and other indicators is a very time-consuming thing to do, but it has to be done:
.
设想一下，一名普通医生每天要面对多个病人，要把他们的各项数据查找出来，在自己的大脑中整合起来，进而决定采取哪一种治疗方案。当病人的数据并不是特别完整，比如此前在另一家医院做的检查治疗，病人手头并没有全部检查单据，对于医生来说会有大量的时间浪费（比如联系之前的医院或重新检查）。
Imagine that a general doctor has to deal with multiple patients every day, find out their data, integrate them in his own brain, and decide which treatment to take. When the patient’s data are not particularly complete, for example, when a previous examination is done in another hospital, the patient does not have all the examination documents at hand, and there is a lot of time wasted for the doctor (such as contacting a pre-hospital or re-examination).
雷锋网消息，在最新的一组论文中，MIT 计算机科学与人工智能实验室 CSAIL 的研究员，提出了两套帮助医生更好做治疗方案决策的系统。
Thundernet news, in the latest set of papers, the researcher at the MIT Computer Science and Artificial Intelligence Laboratory CSAIL proposed two systems to help doctors make better decisions about treatment programmes.
一支团队开发了一套名为 “ICU Intervene” ，即“重症监护室干预”的机器学习系统。大量重症监护室（ICU）的数据，从病人关键生命体征、之前医生的治疗备注，到人口统计学信息，都会被整合到一起，以帮助医生做出决策——哪些治疗方案最适合当前病人的症状。该系统使用深度学习来做出实时预测，从过去的 ICU 案例中学习，对当前情况严重的病例（病危护理）做出推荐，并能对其背后的原因与逻辑做出解释。
A team developed a machine learning system called ICU Intervene 论文的第一作者、MIT 博士生 Harini Suresh 表示： The first author of the ICU Intervene dissertation, MIT PhD & nbsp; Harini Suresh says: “这套系统有潜力成为 ICU 值班医生的助手，这些医生的工作环境有巨大压力以及极高要求。这项研究的目标是充分利用医疗记录的数据，提高医疗水平，并对必要的干预提前做出预测。” “The system has the potential to be an assistant to ICU duty doctors, who work in stressful and highly demanding conditions. The objective of the study is to make full use of data from medical records, raise the level of care and anticipate the necessary intervention.” 另一支团队开发的系统被称为“EHR Model Transfer” ，即“EHR 模型迁移”。它能推动跨电子医疗档案系统（EHR）预测模型的落地。也就是说，用一套 EHR 的数据训练出来的预测模型，能够迁移到另一套 EHR 系统上进行应用，做出有效预测。该团队发现，“EHR 模型迁移”能对病人的死亡率、住院延长时间做有效预测。 Another system developed by the team is known as “EHR Model Transfer” or “EHR Model Migration.” It facilitates the landing of a cross-electronic medical file system (EHR) predictive model. That is, a predictive model, trained with data from one EHR, can be migrated to another EHR system for effective prediction. The team found that “EHR Model Migration” can be effective in predicting the patient's mortality and length of hospitalization. 两套系统都使用病危护理数据库 MIMIC 进行训练，后者包含四万个病危病例的脱敏数据，由 MIT 生理计算实验室（MIT Lab for Computational Physiology）开发。 Both systems are trained using the Crisis Care Database, MIMIC, which contains desensitization data for 40,000 cases at risk, developed by the MIT Lab for Physical Calculation Laboratory. via mit 雷峰网版权文章，未经授权禁止转载。详情见转载须知。 Copyright article by Le Peaksnet, banned without authorization. For further information, see
注册有任何问题请添加 微信：MVIP619 拉你进入群
打开微信扫一扫
添加客服
进入交流群
发表评论