System identification finds nowadays application in various areas of biomedical research as a tool of empiric mathematical modeling and model individualization. Hormone regulation is a classical example of biological feedback where control theories, in general, and system identification, in particular, are indispensable in unraveling the regulation mechanisms and explicating the complex dynamical phenomena arising in endocrine systems.
The main function of endocrine feedback regulation is to maintain the hormone levels within a particular physiological range as well as to sustain an appropriate hormone secretion pattern. Therefore, a natural operating mode of a closed-loop endocrine system is a stable periodic cycle. This property significantly reduces the repertoire of readily available identification techniques, not least due to the fact that the regulation (input) signal is immeasurable in many practical cases.
There are two approaches to blind identification of hormone dynamics presented in this thesis. The first one is based on constrained nonlinear least-squares method. Weighting functions play an important role in satisfying the biological conditions on the identified model. The second approach is derived from a novel time-delay system identification method in Laguerre domain. In the latter, the time delay appears due to a specific input signal model and is estimated along with the finite-dimensional dynamics of hormone kinetics.
Available as PDF (1.05 MB)
Download BibTeX entry.