Identification for control of biomedical systems using a very short experiment

This paper presents a combined experiment and identification procedure, well suited to obtain low-order dynamic models of a patients' response to continuous drug administration. The experiment requires no a priori information and is of very short duration. The identification method provides both a parametric low-order model, and an estimate of the parameter error covariance. It has been demonstrated to work well with very noisy measurements, as typically encountered in drug dosing applications.