
Speciale  modelbyggervariant, 2. modul, Sommer 2011, id:419  
Vejleder:  Johnny T. Ottesen 
Findes på RUb:  Ja 
English abstract
In this thesis a model of the baroreflex as presented in Ottesen and Olufsen 2010 is modified and later optimized against experimental data. It is investigated if the model can describe the experimental data, the answer is inconclusive. The model takes the change in pressure as an input and provides a heart rate as output. In the thesis the model is made dimensionless and the original models discrete time delay funktion is replaced by a distributed delay function. The experimental data originates from a series of HUTexperiment sessions on a group of test persons, young healthy men in their mid twenties. The HUTsessions exposed the test subjects to a consecutive series of five HUT's. The first of the five HUT's took place after an injection with saltwater (placebo), and the following four HUT's each took place after an atropine injection. The mathematical model is optimized on data from a series of HUTexperiments on the test subject referred to as Jens. The model is solved using the CVODES solver, which applies a variable order and variable step size implementation of the backwards Difference method. The sensitivity analysis performed to lock parameters of the model is a direct method that runs parallel with the BDF method. The parameter estimation method is a Metropolis Markov Chain Monte Carlo method (MCMC). The model fits reasonably well on data from Jens held in supine position after the saltwater injection, and the parameters found by MCMC are not physiologically unreasonable. The parameter values are used as an initial guess for a model describing Jens during a tilt. This procedure of sensitivity analysis followed by MCMC optimization and then applying the most likely parameters as initial guess for the following experiment are repeated for all 5 tilts. None of the heart rate data in the tilts are captured well by the optimized parameters which attain physiologically unrealistic values. Another model optimization is done on a HUTexperiment on the test person Danny. The optimization attempts to fit parameters to a pre tilt supine phase, a tilt up phase and an upright phase. The optimization produces poor fits. Due to physiologically confusing behavior in pulse data of the test persons compared to measured pressure data. And the fact that the model at several of these points produces qualitatively more meaningful ouputt, it is concluded that more investigations on other HUT data must be undertaken in order to give a conclusive answer to the problem statement.