Transfer Learning for Clinical Time Series Analysis Using Deep Neural Networks
Transfer Learning for Clinical Time Series Analysis Using Deep Neural Networks
复制标题
DOI：
10.1007/s41666-019-00062-3
复制
发表时间：
2020-06-01
影响因子：
5.9
通讯作者：
Shroff, Gautam
中科院分区：
文献类型：
--
作者：
Gupta, Priyanka;Malhotra, Pankaj;Shroff, Gautam
由患者病史组成的电子健康记录（EHR）在各种临床应用中非常有用，例如诊断和推荐药物[22]。传统的机器学习方法通常需要仔细的特定于领域的特征工程来实现良好的预测性能。另一方面，深度学习方法可以实现端到端学习，而无需手工制作和特定领域的功能，并且最近已经为各种临床预测任务产生了有希望的结果[17，22，29]。因此，深度学习在电子健康记录的各种临床预测任务中的应用迅速增长，例如用于医疗诊断的Doctor AI [6]，用于预测患者未来疾病的Deep Patient [21]，以及用于预测出院后计划外再入院的DeepR [23]。随着EHR数据库中在一段时间内记录各种医疗参数，递归神经网络（RNN）可以成为对EHR数据的顺序方面进行建模的有效方法，从而实现诊断[3，6，17]，死亡率预测和估计住院时间[9，27，28]的应用。
Electronic health records (EHR) consisting of the medical history of patients are useful in various clinical applications such as diagnosis and recommending medicine [22]. Traditional machine learning approaches often require careful domain-specific feature engineering to achieve good prediction performance. On the other hand, deep learning approaches enable end-to-end learning without the need of hand-crafted and domain-specific features, and have recently produced promising results for various clinical prediction tasks [17, 22, 29]. As a result, there has been a rapid growth in the applications of deep learning to various clinical prediction tasks from electronic health records, eg, Doctor AI [6] for medical diagnosis, Deep Patient [21] to predict future diseases in patients, and DeepR [23] to predict unplanned readmission after discharge. With various medical parameters being recorded over a period of time in EHR databases, recurrent neural networks (RNNs) can be an effective way to model the sequential aspects of EHR data and, in turn, enable applications in diagnoses [3, 6, 17], mortality prediction, and estimating length of stay [9, 27, 28].