Predicting acute hypotensive episode by using hybrid features and a neuro-fuzzy network

Authors

MARZIEH ABBASINIA
FARDAD FAROKHI
SHAHRAM JAVADI

DOI

10.3906/elk-1403-117

Abstract

This paper presents an approach for acute hypotensive episode (AHE) time series forecasting based on hybrid feature space and a neuro-fuzzy network. Prediction was accomplished through a combination of time domain and wavelet features by using six vital time series of each patient, obtained from MIMIC-II and available in the context of the Physionet-Computers in Cardiology 2009 Challenge. At first, statistical time domain features were used and then the wavelet coefficient was utilized for extracting time scale features. Further UTA feature selection was applied and 30 effective features were determined and achieved to predict AHE with 96.30 accuracy 1.5 h before AHE onset.

Keywords

Acute hypotensive episode (AHE), prediction, neuro-fuzzy network (NF), wavelet transform, feature selection, mean arterial blood pressure (MAP)

First Page

3335

Last Page

3344

Recommended Citation

ABBASINIA, MARZIEH; FAROKHI, FARDAD; and JAVADI, SHAHRAM (2016) "Predicting acute hypotensive episode by using hybrid features and a neuro-fuzzy network," Turkish Journal of Electrical Engineering and Computer Sciences: Vol. 24: No. 5, Article 1. https://doi.org/10.3906/elk-1403-117
Available at: https://journals.tubitak.gov.tr/elektrik/vol24/iss5/1