Uživatelské nástroje

Nástroje pro tento web


elearning:eatkins:elearning:elearning:atkins:cesnet:terena2011abstract

Title

Estimation of Physiological Model Parameter Using Desktop Grid System.

Authors

Tomáš Kulhánek tomaton@centrum.cz

  • CESNET z.s.p.o., Zikova 4, 160 00 Praha 6, Czech Republic
  • Institute of Patophysiology, First Faculty of Medicine, Charles University, U Nemocnice 5, 128 53 Praha 2, Czech Republic

Jiří Kofránek kofranek@gmail.com

  • Institute of Patophysiology, First Faculty of Medicine, Charles University, U Nemocnice 5, 128 53 Praha 2, Czech Republic

Jan Šilar jansilar@jansilar.cz

  • Institute of Patophysiology, First Faculty of Medicine, Charles University, U Nemocnice 5, 128 53 Praha 2, Czech Republic

Marek Mateják marek.matejak@gmail.com

  • Institute of Patophysiology, First Faculty of Medicine, Charles University, U Nemocnice 5, 128 53 Praha 2, Czech Republic

Pavol Privitzer pavol.privitzer@lf1.cuni.cz

  • Institute of Patophysiology, First Faculty of Medicine, Charles University, U Nemocnice 5, 121 08 Praha 2, Czech Republic

Martin Tribula martin.tribula@gmail.com

  • Institute of Patophysiology, First Faculty of Medicine, Charles University, U Nemocnice 5, 121 08 Praha 2, Czech Republic

Keywords

identification of physiological system, desktop grid computing, BOINC,

Abstract

Modeling gives an simplified view on the physiology of human body using mathematical equations and non-trivial relations could be seen. HumMod Golem Edition is our physiological model based on QHP and HumMod supported by NASA. We extend the original HumMod model with better Acid-Base and blood gases transport calculations. Today our model has about 150 state variables, 4000 parameters and 70000 auxiliary variables mostly with reasonable physiological meanings and physical units.[1]

Validation of a model on the specific patient can give exact estimation of quantities which are usually not measurable or measurable with extra high effort. The system for identification of physiological systems integrates optimization algorithms and numerical solver of a physiological model using distributed computation techniques to receive results in a reasonable time. The results of the system allows to estimate non measurable parameters and fits the model according to the live patient data [2].

We continue to develop a system which uses customized desktop grid system BOINC [3]. The server is in operation as an independent virtual machine and contains the web service operated by the identification algorithm. Identification algorithm uses optimization methods mainly based on genetic algorithm and is implemented in MATLAB environment. The web service delegates computation task generated by genetic algorithm to the BOINC system. The customized BOINC server part splits computation task into several parts which can be computed independently. The customized BOINC client part uses numerical solver of a model which gives an estimation of a model parameter in a given state. The BOINC system is used to distribute the computation in a standard way and any computer connected to the Internet can join and participate on the computation, we test a independent desktop computers or servers from laboratory in Institute of Patophysiology and shared computation resources provided by national grid initiative METACENTRUM (http://metavo.metacentrum.cz) provided by CESNET association and its members.

We plan to publish the preliminary results in the field of medicine and call other partners in medicine to enhance project computation area. We plan also to call on participation of public users to join their desktop computers or server to participate on computation of parameters of physiological models.

Acknowledgements

This work was supported by the CESNET z.s.p.o. and by project MSMT, CR Pr.No. 1M6138498401.

References

[1] Kofránek J., Mateják M., Privitzer P.: Web Simulator Creation Technology, in MEFANET report 03. (Daniel Schwarz, Martin Komenda,Ladislav Dušek, Stanislav Štípek, Vladimír Mihál Eds.),Masaryk University, Brno, 2010, ISSN 1804-2961, ISBN 978-80-210-5302-1

[2] Tomáš Kulhánek, Jan Šilar, Marek Mateják, Pavol Privitzer, Jiří Kofránek, Martin Tribula:Distributed computation and parameter estimation in identification of physiological systems, proceeding of VPH 2010, 30th Sep - 1st Oct 2010 Brussels, Belgium

[3] Anderson, D. P. 2004. BOINC: A System for Public-Resource Computing and Storage. In Proceedings of the 5th IEEE/ACM international Workshop on Grid Computing (November 08 - 08, 2004). International Conference on Grid Computing. IEEE Computer Society, Washington, DC, 4-10.

Author Biographies

Tomáš Kulhánek, holds MSc. (Computer Science) and is a PhD. student at 1st Faculty of Medicine, Charles University in Prague, Czech Republic. His research interests are medical application, grid computing and cooperates with research activity of Czech academic network operator - CESNET. He works also part-time as a consultant for Hewlett-Packard.

Marek Mateják, holds MSc. (Computer Science) and is a PhD. student at 1st Faculty of Medicine, Charles University in Prague, Czech Republic. He is interested in integrative mathematical human physiology modeling, especially Modelica implementation of huge physiological model HumMod and its integration with other related models.

Jan Šilar, holds Ing. (Applied mathematics) and is a PhD. student at 1st Faculty of Medicine, Charles University in Prague, Czech Republic. He is interested in numerical methods (mainly optimization and methods for solving both ODE and PDE) applied in medicine.

Jiří Kofránek, holds M.D. and PhD in Institute of Pathophysiology. He is head of Laboratory of Biocybernetics, 1st Faculty of Medicine, senior lecturer, Institute of Pathophysiology, 1st Faculty of Medicine Charles University. He is President of Board of Science of Czech Society of cybernetics and informatics and member of American Physiological Society, Society for Modeling and Simulation International.

Pavol Privitzer, holds M.D. and is a PhD. student at 1st Faculty of Medicine, Charles University in Prague, Czech Republic. His research interests are physiological simulators and advanced web technologies.

Martin Tribula, holds Ing. and is a PhD. student at 1st Faculty of Medicine, Charles University in Prague, Czech Republic. He is interested in advanced web technologies and computer graphics.

elearning/eatkins/elearning/elearning/atkins/cesnet/terena2011abstract.txt · Poslední úprava: 2019/08/26 10:33 autor: marek