Name
Title | ||
|---|---|---|
Mortality prediction with self normalizing neural networks in intensive care unit patients |
Abstract | ||
|---|---|---|
Mortality prediction of intensive care unit (ICU) patients is challenging and important in clinical decision making. Traditionally, severity of illness (SOI) scores are used for predicting mortality. While many SOI scores have been proposed, they tend to underperform on validation. In this work, we investigate Deep Learning (DL) methods focusing on the self normalizing neural network (SNN) for predicting mortality in ICU patients. We evaluate the prediction model on approximately 17150 patients from the MIMIC II dataset. The primary outcomes were 30 days and hospital mortality. Compared to the existing methods in the literature, DL models resulted in superior or comparable predictive performance. The final calibrated SNN resulted in an AUC of 0.8445 (±0.08) for 30 days mortality and 0.86 (±0.12) for hospital mortality. This study warrants further application of DL to prediction problems in the ICU. |
Year | DOI | Venue |
|---|---|---|
2018 | 10.1109/BHI.2018.8333410 | 2018 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI) |
Keywords | Field | DocType |
clinical decision making,illness scores,SOI scores,ICU patients,prediction model,hospital mortality,superior performance,prediction problems,mortality prediction,intensive care unit patients,predictive performance,SNN,self normalizing neural networks,Deep Learning methods,DL methods,MIMIC II dataset,time 30.0 d | Intensive care unit,Severity of illness,Clinical decision making,Emergency medicine,Artificial intelligence,Deep learning,Artificial neural network,Medicine | Conference |
ISBN | Citations | PageRank |
978-1-5386-2406-7 | 0 | 0.34 |
References | Authors | |
0 | 2 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| M. A. H. Zahid | 1 | 0 | 0.34 |
| Joon Lee | 2 | 29 | 5.54 |