Based On EMD And GEP Combination Forecast Method Applied To Acute Hypotension Episodes Forecasting

In the process of ICU guardianship, patients with acute hypotension often leads toirreversible organ damage, seriously even endanger the patient’s life. Clinically, after acutehypotension occurs mainly rely on doctor’s experience for disposal. If acute hypotensionoccurs can be forecast, then give professionals enough time to select a more effectivetreatment, it is possible to improve the clinical decision concerning which intervention ismore appropriated for each specific condition (such as blood volume reduction, cardiacoutput, or the specific reasons for vasodilatory shock, such as sepsis, myocardial infarction,arrhythmia, pulmonary embolism, bleeding, dehydration). Early detection of AHE will giveprofessionals enough time to select a more effective treatment, without exposing the patientto additional risks of delaying therapy. Therefore, the development of methodologies able todetect not only the presence of this condition but also to forecast its occurrence, is ofextreme importance.In ICU there have various of clinical monitoring data of patients, including patientbasic information, history, medical records, medical records, medical records, patients withall kinds of data real-time physical signs (such as blood pressure, pulse, breathing).Experiments show that there are certain connections between some specific monitoring dataand its changes and the acute hypotension occurs of patient, in a specific period of time,according to the monitoring data to forecast acute hypotension occurred in the next periodof time is clinically feasible.In this paper, we introduce the modeling methodology based on EMD and GEP, using thereal ABP records of ICU patients from American PhysioNet MIMIC II database, to establishthe ability to forecast the patients’ a period change rule of Mean Artery Pressure in thefuture. This methodology is divided into three stages:1. Change the ABP into mean arterialblood pressure(MAP) based on the relevant medical knowledge, and then preprocesses datausing the signal processing methods(including resampling, denoising, etc);2. By the EMDdecomposition algorithm, the data is decomposed into several Intrinsic ModeFunctions(IMFs), next, observation and establish the mathematical models of each IntrinsicMode Function with GEP evolution algorithm;3. The test records for new patients, throughcorrelation calculation and to match the best mathematical models of the historical records,with the correlation models to forecast new patients’ Mean Artery Pressure values of thefuture, through the weighted. Through experiments, this methodology can accurately forecast future1hour blood pressure value of the patients with accuracy more than75%. Itshows that this methodology can be used as an effective method to forecast the acutehypotension episodes.