Uncertainty Analysis of Healthcare Data via Statistical Modeling and Optimization
MetadataShow full item record
Healthcare data have been useful resources for improving patient care in the health care system. Many efforts have been made to analyze healthcare data of different structures to draw actionable insights into the appropriate application of interventions, and quality and efficiency of care. There is often considerable uncertainty surrounding health data, e.g., measurement noises, motion artifacts, which has attracted increasing attention, given the growing emphasis on data-driven decision-making, evidence-based medicine, and personalized patient care. It is crucial to unravel these uncertainties and develop effective strategies to cope with them in an effective way. This research investigates four tasks to address the uncertainty challenges in different types of healthcare data. The first task deals with uncertainties in clinical data to support treatment design for Atrial Fibrillation (AF). This research developed a robust probabilistic approach that accounts for uncertainties in Intracardiac Electrograms for abnormal electrical impulse identification. The second task concentrates on removing motion artifacts and measurement noises in photoplethysmography signals measured through wearable devices for real time Heart Rate (HR) monitoring during physical exercises. The robust and accurate estimation of HR, an indicator of cardiovascular condition and physiological adaption, can benefit both healthy persons who seek to meet fitness goals and cardiac patients in rehabilitation program to strength their heart health. The third task explores the combination of nonlinear mapping of cardiac related risk factors and a well-known adaptive learning algorithm to predict the risk of heart disease. This research provides a straight-forward Machine Learning tool to aid cardiac disease diagnosis and risk prediction.