The HTP tool: Monitoring, detecting and predicting hypotensive episodes in critical care

摘要

The sudden fall of blood pressure (hypotension — HT) is a common complication in medical care. In critical care patients, HT may cause serious neurological, heart, or endocrine disorders, inducing severe or even lethal events. Recent studies report an increase of mortality in HT prone hemodialysis patients in need of critical care. Predicting HT episodes in advance is crucial to enable medical staff to minimize its effects or even avoid its occurrence. Most medical systems have focused on monitoring and detecting current patient status, rather than determining biosignal trends or predicting the patient's future status. Therefore, predicting HT episodes in advance remains a challenge. In this paper, we present a solution for continuous monitoring and efficient prediction of HT episodes. We propose an architecture for a HT Predictor (HTP) Tool, capable of continuously storing and real-time monitoring all patient's heart rate and blood pressure biosignal data, alerting probable occurrences of each patient's HT episodes for the following 60 minutes, based on non-invasive hemodynamic variables. Our system also promotes medical staff mobility, taking advantage of using mobile personal devices such as cell phones and PDA's. An experimental evaluation on real-life data from the well-known Physionet database shows the tool's efficiency, outperforming the winning proposal of the Physionet 2009 Challenge.