ERI: Precision Dosing in Critical Care: An Automated Modeling and Control Approach

Kent State University, Kent OH

Investigators

Abstract

This Engineering Research Initiation (ERI) grant will fund research that enables patient-specific fluid resuscitation therapies in critical care of hemorrhage, a leading cause of deaths from traumatic injuries, thereby promoting the progress of science and advancing the national health and welfare. Death caused by hemorrhagic injuries may be preventable if critical care interventions are urgently and effectively deployed to restore lost blood volume. Fluid infusion dosages, including both rate and timing of administration, notably impact resuscitation outcomes. Under-dosing strategies are inefficient in restoring cardiac functions, whereas overly aggressive dosing regimens may lead to serious adverse events such as soft clots, thereby increasing mortality rates. Most existing dosing techniques are based on one-size-fits-all models or a small number of dose-response profiles. Such approaches sacrifice the control performance at the expense of robustness against inter-patient variability. A framework for identifying the correct dosage for each individual patient is currently lacking. This project will address this deficiency by developing a data-driven, integrated modeling and control framework for precision dosing that is able to resolve individual differences in dose response and predict optimal patient-specific dosing strategies, even with limited and noisy measurements. The knowledge advanced by this project may not only improve patient outcomes, but also significantly reduce costs associated with critical care services. Efforts aiming to attract and train students in STEM will include K-12 outreach using a fluid resuscitation testbed, as well as research experiences for undergraduate students. This research aims to make fundamental contributions to the science of physiological closed-loop control systems: complex cyber-physical systems that involve interactions between patient monitors, therapeutic devices, complex patient physiology, and clinical users. It will achieve this outcome by first developing and validating a new system identification framework that uses a statistical bound on the prediction error to constrain a robust nonlinear state space model of a patient-specific dose response. Next, it will formulate an automated, computationally efficient control strategy for achieving the desired outcome subject to the predicted dose-response dynamics, initial and boundary conditions, and mixed state-control path constraints. Finally, feasibility and performance will be assessed against real-world clinical datasets of human subject dose-response measurements. Evaluation will be conducted both using computer simulations and by implementing computational models of physiologic variables and computer-based fluid resuscitation controllers in a hardware-in-the-loop testing platform. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.