Predictive Analytics

Data-driven and Domain Knowledge Based Forecasting of Acute Myocardial Infarction Events

Forecasting the onset of the acute events has become critical for the development of smart screening cyber-physical system in healthcare. Investigations on the triggering mechanism of various acute disorders onset in patients are still not well established, despite the urgent demanding of the onset forecasting. The utilization the morphological-temporal features of the surface electrocardiogram (ECG) or heart rate variability (HRV) features significantly requires the explanation from clinical perspectives, especially about the trigger modes of onset and the disease pathologies. Therefore, this work aims to construct data-driven analytic forecasting method, gain a profound understanding of the physiological dynamics of the cardiovascular system for the forecast prior (i.e. 5 mins to 60 mins in advance) the onset of proximal atrial fibrillation and myocardial infarction attacks before the clinical symptoms actually happen. Such research will provide a reliable clinical decision-making support for physicians.


Nonlinear Dynamics Forecasting of Obstructive Sleep Apnea Onsets

Advancements in wearable sensors, flexible electronics, and “big data” predictive analytics have been encouraging the use of point-of-care (POC) technologies for obstructive sleep apnea (OSA) treatment. These include multiple variants of automatic positive airway pressure (APAP) machines, position-adjusting beds, and nerve stimulation devices. However, these adjustments tend to be reactive in that a control or intervention is initiated upon the detection of an OSA episode. We present an approach to predicting the time to OSA onset using nonlinear dynamics with the assumption that the underlying OSA-driven system equations are unknown and only the time series reflecting the evolution of the dynamics of the system are available. The timeliness and effectiveness of the intervention can be enhanced substantially if an impending OSA episode can be predicted before the clinical symptoms become evident. Previous attempts at OSA prediction are limited to capturing the nonlinear, nonstationary dynamics of the underlying physiological processes. This project reports an investigation into heart rate dynamics aiming to predict in real time the onsets of OSA episode before the clinical symptoms appear.