Data Assimilation of Glucose Dynamics for the Neuro-ICU

Open Access
Hall, Kenneth Timothy
Area of Honors:
Engineering Science
Bachelor of Science
Document Type:
Thesis Supervisors:
  • Bruce Gluckman, Thesis Supervisor
  • Clifford Jesse Lissenden III, Honors Advisor
  • Data Assimilation
  • Glucose Dynamics
  • Kalman Filter
  • Mathematical Physiology
Measurements of plasma glucose are used to make decisions about insulin delivery in diabetic patients, make judgments about the effectiveness of diabetes treatment, and set feeding levels for incapacitated patients in the Intensive Care Unit. Glucose/insulin dynamics are highly nonlinear and oscillatory even at constant input, which complicates therapy decisions by necessitating an understanding of both current and future system behavior. This paper presents an application of data assimilation to the study of glucose/insulin dynamics. A state estimate is propagated forward in time by using an Unscented Kalman Filter to fit a mathematical representation of glucose and insulin dynamics to serial glucose and feeding data. This approach enables a prediction of near-future dynamics, as well as the reconstruction of unmeasured variables including interstitial insulin, plasma insulin, or exogenous glucose delivery.