Prediction of Acute Hypotensive Episode in ICU Using Chebyshev Neural Network

Abstract

An Acute Hypotensive Episode (AHE) is a serious threat to the lives of Intensive Care Unit (ICU) patients. The proposed method of accurately predicting an AHE will allow doctors to make a timely and effective intervention; therefore, it has a high clinical value. In recent years, the Chebyshev neural network model has been favorably applied in other fields. In this study, we built a Chebyshev neural network based on pattern recognition to predict AHE in ICU patients. We preprocessed the arterial blood pressure (ABP) data of the ICU patients, and then, extracted time-domain signal features from the data to construct the feature vector. We trained the neural network using a classified predictive model to predict AHEs. We used the classic BP neural network and its improved versions for comparison. Our experimental results show that our Chebyshev neural network model performs better than other solutions in predicting AHEs; therefore, it can provide a reference for clinical applications.