Research on applying diagnosis method based on artificial neural networks to evaluate ICU patients´ health states

This paper gives an evaluation model based on artificial neural networks (ANNs) for ICU patients´ health states. This evaluation model uses Back-Propagation Neural Network (BPNN) algorithm to classify patients´ states. This paper builds BP network inputs of which are the same with the parameters of the Acute Physiology and Chronic Health Evaluation II (APACHE Π) scoring system, and outputs of which are the same with the scoring result. After training network by sample data from a number of ICU cases and using remaining sample data to test the network, the accuracy rate of the model is greater than 85% This model reduces the time of building the evaluation model and lessens the requirement of the quantity of data.