Icahn School of Medicine at Mount Sinai, Department of Orthopaedics, New York, NY, United States of America.
Icahn School of Medicine at Mount Sinai, Department of Neurological Surgery, New York, NY, United States of America.
PLoS One. 2019 Feb 13;14(2):e0211057. doi: 10.1371/journal.pone.0211057. eCollection 2019.
This study trained long short-term memory (LSTM) recurrent neural networks (RNNs) incorporating an attention mechanism to predict daily sepsis, myocardial infarction (MI), and vancomycin antibiotic administration over two week patient ICU courses in the MIMIC-III dataset. These models achieved next-day predictive AUC of 0.876 for sepsis, 0.823 for MI, and 0.833 for vancomycin administration. Attention maps built from these models highlighted those times when input variables most influenced predictions and could provide a degree of interpretability to clinicians. These models appeared to attend to variables that were proxies for clinician decision-making, demonstrating a challenge of using flexible deep learning approaches trained with EHR data to build clinical decision support. While continued development and refinement is needed, we believe that such models could one day prove useful in reducing information overload for ICU physicians by providing needed clinical decision support for a variety of clinically important tasks.
本研究在 MIMIC-III 数据集上，训练了带有注意力机制的长短期记忆（LSTM）循环神经网络（RNN），以预测 ICU 患者两周内的日常败血症、心肌梗死（MI）和万古霉素抗生素治疗情况。这些模型对败血症、MI 和万古霉素治疗的次日预测 AUC 分别达到 0.876、0.823 和 0.833。从这些模型构建的注意力图突出了输入变量最能影响预测的时刻，并可以为临床医生提供一定程度的可解释性。这些模型似乎关注的是代表临床医生决策的变量，这表明使用基于 EHR 数据训练的灵活深度学习方法来构建临床决策支持存在挑战。虽然需要进一步开发和完善，但我们相信，随着时间的推移，此类模型最终可能会通过为各种重要临床任务提供所需的临床决策支持，从而有助于减少 ICU 医生的信息过载。
