Post doctoral subject : Sensitivity analysis and uncertainty modelling : application to electromyographical models
Post doc Advisors
Sofiane Boudaoud, MCF 61, BMBI Laboratory, NSE team
Tel : +33 (0)3 44 23 79 29, sofiane.boudaoud@utc.fr
Sébastien Destercke, CR CNRS 7, HEUDIASYC Laboratory, DI team
Tel : +33 (0)3 44 23 79 85, sebastien.destercke@utc.fr
Catherine Marque, PR 61, BMBI Laboratory, NSE team
Tel : +33 (0)3 44 23 45 71, catherine.marque@utc.fr
Context of the study
The thesis is part of the project activities of the Laboratory of Excellence (LABEX) at the Université de Technologie de Compiègne in France on the Control of Technological Systems of Systems (MS2T) (www.utc.fr/labexms2t).
Life systems are particularly complex systems, composed of sub-systems interacting with each others. To be sufficiently realistic, models of such systems have to be multi-scales (cells, organs, organisms) and multi-physics (chemical, electrical, mechanical, etc.)
Such systems are often associated with uncertainty, due to the important variability (between individuals and measurements) and to the poor knowledge of some parameters or mechanisms. Classical treatments of uncertainty in such systems are often very rough : for instance sensitivity analysis, that aims to identify important parameters, is often reduced to a simple and identical variation of each parameter (e.g., +/- 30%) that does not take into account parameter specificity. At most, such studies are performed for couples of parameters [1].
Such methods are therefore based on unrealistic hypothesis and should be refined. In particular, using more realistic uncertainty models would allow for a better identification of the role and importance of various model parameters. Uncertainty analysis results could then be used to better understand the model and to simplify it for parameter identification task.
Post doc description
The post-doctorate will start by formalizing uncertainties [2], both random (variability in a population) or epistemic (lack of information). This will be done through data collection and bibliographical search about existing models developed in the NSE team [3, 4].
Sensitivity analysis based on this formalization will then be performed. In addition to recent analysis methods, based on classical probabilistic formalism, analyses based on imprecise
probabilities and belief functions [5, 6] will be performed, to deal with severe uncertainties. Such analyses will require the definition of sensible design experiments, to cover the whole parameter space with a limited number of computations [7]. Such analyses will be used to identify interaction between variables as well as their links with measured quantities (mean frequency of signals, action potential shapes,…).
Analyses will be first applied to reduced systems such as single cell for electromyogram (EMG) of the uterine muscle and one motor unit for the EMG of the striated muscle before being extended to global multi-scales systems (e.g., entire muscles). Results will be used to characterize links between measured variables and model parameters, these links will therefore permit to estimate the relevance of measured variables or to propose new ones.
Finally, one possible perspective of the current work is to adapt the model and its parameters to measured variables in order to identify the most influent parameter set for the studied electromygraphical models.
Candidate’s profile
Candidate should possess a PhD thesis and skills in the following ordered (of importance) fields :
Modelling of electrophysiological systems
Biomedical signal processing
Sensitivity analysis
Python programming
Skills in uncertainty modelling and parallel programming would be appreciated.
Documents required to apply
Send to : sofiane.boudaoud@utc.fr
Curriculum vitae
Motivation letter
At least two references and/or recommendation letters
At least two significative papers (peer reviewed journal)
Location
Laboratory Heudiasyc UMR CNRS 7253 and
Laboratory BMBI UMR CNRS 7338
Université de Technologie de Compiègne
Centre de recherche de Royallieu
BP 20529 Rue Personne de Roberval
60205 Compiègne cedex - France
www.utc.fr/bmbi
www.hds.utc.fr
References
[1] Jose F. Rodriguez, Jesus Carro Fernandez, Esther Pueyo, Kevin Burrage, and Blanca Rodriguez. Impact of multiple ionic changes in arrhythmic risk biomarkers in human ventricular electrophysiology. Biophys J, 102(3) :543a–543a, January 2012
[2] M. Hanss and S. Turrin. A fuzzy-based approach to comprehensive modeling and analysis of systems with epistemic uncertainties. Structural Safety, 32(6) :433 – 441, 2010.
[3] Jeremy Laforet, Chiara Rabotti, Jeremy Terrien, Massimo Mischi, and Catherine Marque. Toward a multiscale model of the uterine electrical activity. IEEE Transactions on Bio-Medical Engineering, 58(12) :3487–3490, December 2011.
[4] F. Ayachi, S. Boudaoud, J.F. Grosset, and C. Marque. Study of the muscular force/hos parameters relationship from the surface electromyogram. In 15th NBC on Biomedical Engineering & Medical Physics, volume 34, pages 187–190, 2011.
[5] Diego A. Alvarez. Reduction of uncertainty using sensitivity analysis methods for infinite random sets of indexable type. International Journal of Approximate Reasoning, 50(5) :750 – 762, 2009.
[6] Michael Oberguggenberger, Julian King, and Bernhard Schmelzer. Classical and imprecise probability methods for sensitivity analysis in engineering : A case study. Int. J. Approx. Reasoning, 50(4) :680–693, April 2009.
[7] Jack P. C. Kleijnen, Susan M. Sanchez, and Thomas W. Lucas. State-of-the-art review : A user’s guide to the brave new world of designing simulation experiments, 2005.
|