Skip to main content
Fig. 1 | BioData Mining

Fig. 1

From: Interpretable recurrent neural network models for dynamic prediction of the extubation failure risk in patients with invasive mechanical ventilation in the intensive care unit

Fig. 1

Study overview. A Raw data was collected from the MIMIC-IV, and preprocessed. Dynamic features varied over time, whereas static features kept constant. The patients had various lengths of time sequence. 80% of the included patients constituted the training set, and fivefold cross validation was used to optimize model hyperparameters and yield 5 trained models. The rest 20% constituted the test set for model validation. B The architecture of RNN contained one layer of LSTM/GRU neurons and three linear layers. RNN received a time sequence of feature vectors, and output a corresponding sequence of predicted EF risk. Abbreviations: EF extubation failure, GRU gated recurrent unit, LSTM long short-term memory, RNN recurrent neural network

Back to article page

BioData Mining

ISSN: 1756-0381