[Back]

H. Kerschner, M. Kozek:
"Identification of Combined Patient Pharmacokinetics and Pharmacodynamics Using Neural Networks";
Talk: 4th International EUROSIM 2001 Congress, Delft, Netherlands; 2001-06-26 - 2001-06-29; in: "Proceedings of the 4th International Eurosim 2001 Congress", (2001), ISBN: 90-806441-1-0.

For the control of depth of anaesthesia (DOA) several models have been developed. With the bispectral
index (BIS) a single parameter is used as input which represents a non-invasive measurement of
the DOA. The standard model consists of a linear dynamic pharmacokinetic part and a nonlinear static
pharmacodynamic part. On-line identification of the model parameters based on input-ouput measurements
is desireable and may be achieved by nonlinear identification schemes. In the present paper the
performance of artificial neural networks with a feedforward structure in the identification of both static
nonlinearities and nonlinear dynamic physiological systems is investigated. From these realizations it
is evident that the neural networks are able to reproduce the system dynamics even in the presence of
measurement noise.

Identification, Artificial Neural Network, Radial Basis Function, Pharmacokinetic-Pharmacodynamic Model

http://publik.tuwien.ac.at/files/pub-mb_5354.pdf