Prediction of acute hypotensive episodes by means of neural network multi-models



This work proposes the application of neural network multi-models to the prediction of adverse acute hypotensive episodes (AHE) occurring in intensive care units (ICU). Contrasting with classical auto regressive model structures, multi-models do not recursively use model outputs as inputs for step ahead predictions. As result, prediction errors are not propagated over the forecast horizon and long-term predictions can be more accurately estimated. Among regression models, neural network structures present considerable capabilities to learn and generalize from non-linear environments. Additionally, given their ability to capture system dynamics, neural networks are well-suited to be incorporated into multi-model schemes, namely for prediction purposes.
The effectiveness of the neural network multi-models approach is validated was the context of the 10th PhysioNet/Computers in Cardiology Challenge - Predicting Acute Hypotensive Episodes. Applied to the available vital signals time-series, collected from MIMIC-II database, the proposed strategy enables to adequately capture arterial blood pressure dynamics and, consequently, to predict the onset of an AHE within a pre-defined forecast window. A correct prediction of 10 out of 10 AHE for event 1 and of 37 out of 40 AHE for event 2 was achieved, corresponding to the best results of all entries in the two events of the challenge


Hypotension episodes, Blood pressureprediction, Neural network multi-models, MIMIC-II database


Biosignals prediction and classification


Computers inBiologyandMedicine, Vol. 41, pp. 881-890, Elsevier, October 2011

PDF File

Cited by

No citations found