This title appears in the Scientific Report :
1999
Please use the identifier:
http://hdl.handle.net/2128/20556 in citations.
Stochastische Algorithmen zur Parameterschätzung in der Kompaktmentanalyse von dynamischen Positronen-Emissions-Tomographischen Daten
Stochastische Algorithmen zur Parameterschätzung in der Kompaktmentanalyse von dynamischen Positronen-Emissions-Tomographischen Daten
Positron-Emission-Tomography (PET) is a modern medicine technology to gain information about the distribution of biochemical substances in the living human. The substances used for the examination are called tracers and show the metabolic behavior of the substance . The measurement is quantitative s...
Saved in:
Personal Name(s): | Schuth, G. (Corresponding author) |
---|---|
Contributing Institute: |
Institut für Medizin; IME |
Imprint: |
Jülich
Forschungszentrum, Zentralbibliothek
1999
|
Dissertation Note: |
Aachen, Techn. Hochsch., Diss., 1999 |
Document Type: |
Book Dissertation / PhD Thesis |
Research Program: |
ohne FE |
Series Title: |
Berichte des Forschungszentrums Jülich
3720 |
Subject (ZB): | |
Link: |
OpenAccess OpenAccess |
Publikationsportal JuSER |
Positron-Emission-Tomography (PET) is a modern medicine technology to gain information about the distribution of biochemical substances in the living human. The substances used for the examination are called tracers and show the metabolic behavior of the substance . The measurement is quantitative so that the dynamics of the metabolization can be calculated by recording a series of pictures. Because of the complex structure and limited spatial and temporal resolution of PET, simplifications have to be introduced to analyze the data . Biomathematical models are used to perform this task . The defmition of appropriate models and the estimation of related parameters is called modelling . Its goal is to extract relevant metabolic parameters showing differences in normal or diseased tissue. Models for PET data, especially the mainly used compartment models, are discussed in this paper. Different methods for parameter extraction, i .e. graphical and algebraic nonlinear algorithms, are described in detail. A common problem of the nonlinear parameter estimation is its dependence on starting values. Another problem is the uniqueness of the estimated parameters. These problems are addressed by two new algorithms . They are based on the stochastic approach of simulated annealing. Comparisons to conventional algorithms known from literature are given . The new algorithms show independence of starting values and a better numerical stability . Furthermore, they help with the identification of the parameters . In this way the proposed procedures are regarded as important for the development of new tracers and models. |