ICE-NODE: Integration of Clinical Embeddings with Neural Ordinary Differential Equations
ICE-NODE: Integration of Clinical Embeddings with Neural Ordinary Differential Equations
复制标题
DOI：
10.48550/arxiv.2207.01873
复制
发表时间：
2022-07
期刊：
影响因子：
--
通讯作者：
Asem Alaa;Erik Mayer;Mauricio Barahona
中科院分区：
文献类型：
--
作者：
Asem Alaa;Erik Mayer;Mauricio Barahona
疾病的早期诊断可以改善健康结果，包括提高存活率和降低治疗费用。随着电子健康记录（EHRs）中可用的大量信息，使用机器学习（ML）方法模拟疾病进展的潜力很大，旨在早期预测疾病发作和其他结果。在这项工作中，我们采用了神经ode的最新创新，结合临床代码的丰富语义嵌入来利用电子病历的完整时间信息。我们提出了ICE-NODE（集成临床嵌入与神经常微分方程），这是一个暂时集成临床代码和神经ode嵌入的架构，以学习和预测电子病历中的患者轨迹。我们将我们的方法应用于公开可用的MIMIC-III和MIMIC-IV数据集，我们发现与最先进的方法相比，预测结果有所改善，特别是对于在电子病历中不经常观察到的临床代码。我们还表明，ICE-NODE在预测某些医疗状况方面更有能力，如急性肾功能衰竭、肺源性心脏病和出生相关问题，在这些方面，完整的时间信息可以提供重要的信息。此外，ICE-NODE还能够产生随时间推移的患者风险轨迹，可用于进一步详细预测疾病演变。
Early diagnosis of disease can lead to improved health outcomes, including higher survival rates and lower treatment costs. With the massive amount of information available in electronic health records (EHRs), there is great potential to use machine learning (ML) methods to model disease progression aimed at early prediction of disease onset and other outcomes. In this work, we employ recent innovations in neural ODEs combined with rich semantic embeddings of clinical codes to harness the full temporal information of EHRs. We propose ICE-NODE (Integration of Clinical Embeddings with Neural Ordinary Differential Equations), an architecture that temporally integrates embeddings of clinical codes and neural ODEs to learn and predict patient trajectories in EHRs. We apply our method to the publicly available MIMIC-III and MIMIC-IV datasets, and we find improved prediction results compared to state-of-the-art methods, specifically for clinical codes that are not frequently observed in EHRs. We also show that ICE-NODE is more competent at predicting certain medical conditions, like acute renal failure, pulmonary heart disease and birth-related problems, where the full temporal information could provide important information. Furthermore, ICE-NODE is also able to produce patient risk trajectories over time that can be exploited for further detailed predictions of disease evolution.