A new model for tracking and detection of deterioration of vital signs based on artificial neural network

Abstract

Tracking and detection of the deterioration of vital signs has always been a challenging issue since it always happens suddenly and is associated firmly with serious problems such as recurrent readmissions of patients, increase the mortalities, and very little time window left for the clinician to take prompt medical action to treat the patient upon the detection. Many research have proposed various methods to predict and detect the deterioration of vital signs, but each of them has some strength and limitation, in terms of algorithm complexity and detection accuracy. This paper evaluates the capability of various Artificial Neural Network (ANN) models based on machine learning method to detect the deterioration of vital signs which consists of heart rate, blood pressure, body temperature and the saturation of oxygen in the blood. To evaluate and benchmark the detection accuracy of vital signs deterioration, various ANN models were constructed with the specific characteristics of each vital sign as input variables. Results show that the Levenberg-Marquardt ANN model yields the highest detection accuracy of 95%, hence it is reliable in detecting the deterioration of vital signs.