Data-driven patient-specific fluid resuscitation

Abstract

Automated fluid resuscitation systems determine fluid infusion rates for patients recovering from, e.g., hemorrhage or dehydration. Instead of relying on models that require patient-specific parameter identification, we present a data-driven patient-specific alternative that bypasses modeling and identification. Specifically, we implement a predictive controller that only uses blood pressure data and manages to regulate the patient’s blood pressure to a specified setpoint. We demonstrate the results in a simulation on 100 virtual patients created using a well-established model of cardiovascular hemodynamics.