BME Seminar: “Real-time Hemodynamic Forecasting & Clinical Risk Assessment via Hybrid Mechanistic Modeling…”

Details

T.J. Sego, Ph.D.
Assistant Professor
Laboratory for Systems Medicine, Department of Medicine
University of Florida

“Real-time Hemodynamic Forecasting and Clinical Risk Assessment via Hybrid Mechanistic Modeling: A Medical Digital Twin Approach for Improved Septic Shock Management”

Abstract: Septic shock is the most dangerous systemic manifestation of infection resulting in compromised end-organ perfusion. Current management is largely reactive, since it is currently difficult to predict decline in blood pressure in the intensive care setting. In this work, we introduce a framework for iterative Bayesian inference upon a mechanistic model of the cardiovascular system for forecasting tachycardia and hypotension in septic shock patients. We apply our framework to forecast heart rate and blood pressure from patients identified to have been diagnosed with septic shock in retrospect. We generated a computational “digital twin” personalized to each subject’s real-time heart rate and blood pressure data via a combined mechanistic modeling and Bayesian inference approach. Our digital twin is a dynamic virtual model of an individual’s relevant physiology that allows for continuous forecasting of heart-rate and blood pressure. This digital twin framework achieves >80% area under the receiver operator characteristic curve at forecasting hypotension and tachycardia. In a comparison with an existing clinical forecasting technology, our approach delivers superior accuracy to predict hypotension over the same forecasting periods and similar performance capability but over 24x greater forecasting periods. Our approach represents a step towards personalized, anticipatory therapy for the clinical management of septic shock.

Bio: My career focus is to build a research program that produces computational approaches and technologies that improve human health. Much of my research program operates around the concept of the medical digital twin. I believe in a future where computers explain the underlying mechanisms responsible for observed human health, in the same way that complex systems are modeled and analyzed in modern engineering practice using theory and computer simulation. I have developed numerical and theoretical models and simulations of cellular and tissue dynamics with a focus on lung viral infection and immune response. More recently, my work has focused on real-time patient health forecasting and personalized medicine in critical care. I also have a strong interest in promoting findable, accessible, interoperable, and reusable (FAIR) biological models.