Linear Modeling and Prediction in Diabetes Physiology

In English Diabetes Mellitus is a chronic disease characterized by the inability
of the organism to autonomously regulate the blood glucose level due
to insulin deficiency or resistance, leading to serious health
damages. The therapy is essentially based on insulin injections and depends
strongly on patient daily decisions, being mainly based upon empirical
experience and rules of thumb. The development of a prediction engine
capable of personalized on-the-spot decision making concerning the most adequate
choice of insulin delivery, meal intake and exercise would therefore be a
valuable initiative towards an improved management of the desease.

This thesis presents work on data-driven glucose metabolism modeling and
short-term, that is, up to 120 minutes, blood-glucose prediction in Type 1 Diabetes Mellitus (T1DM)
subjects.

In order to address model-based control for blood glucose regulation, low-order, individualized, data-driven, stable, physiological relevant
models were identified from a population of 9 T1DM patients
data. Model structures include: autoregressive moving average with exogenous inputs
(ARMAX) models and state-space models.

ARMAX multi-step-ahead predictors were
estimated by means of least-squares estimation; next regularization of the
autoregressive coefficients was introduced. ARMAX-based predictors
and zero-order hold were computed to allow comparison.

Finally, preliminary results on subspace-based multi-step-ahead
multivariate predictors is presented.