From Prediction of Life Threatening Events to Optimization of Treatment Strategies: An Overture to a Continuously Learning Healthcare System

This meeting is part of the annual Signal Processing Seminar Series.

Abstract

Clinical medical decision-making is an imperfect science influenced by numerous factors including: incomplete knowledge of critical care physiology and varied patient response to standard of care. Most patient care protocols are derived from population-based studies that often rely on low-dimensional and static patient phenotypes. As an example, we consider the problem of early prediction and treatment of Sepsis; the leading causes of morbidity and mortality in critically ill patients and the most expensive condition by healthcare spending. The major tenet of sepsis care is prompt recognition and initiation of treatment, however, no clinically validated system exists for accurate, real-time prediction of sepsis onset, and considerable controversies remain concerning the effectiveness of various treatment options for septic patients. In this talk, I will discuss how modern signal processing and machine learning tools in association with clinical big data can be adopted to improve early prediction of sepsis and design effective treatment strategies.