A. Pironet
1, P.C. Dauby
1, T. Desaive
11
University of Liège (ULg), GIGA-‐Cardiovascular Sciences, Liège, Belgium
DIRECT PARAMETER IDENTIFICATION
IN A MODEL OF THE CARDIOVASCULAR SYSTEM
IntroducAon
Methods
• Parameter idenAficaAon is crucial if models are to be used to track an individual paAent’s condiAon.
• In cardiovascular system modeling, model-‐dependent iden,fica,on methods can find the best set of parameters, whereas in general nonlinear regression methods cannot*.
• In this work, a model-‐dependent method is developed which allows compuAng model parameters without requiring any model simulaAon.
Ppu Plv, Vlv LV Elv Rmt Pao, Vao Aorta Eao Rav Pvc Rsys Model outputs:
• leO ventricle pressure Plv(t), • aorAc pressure Pao(t), • stroke volume SV,
Lumped model of the leK ventricle (LV) and systemic circulaNon* Model inputs: LV driver funcAon e(t), heart rate HR
EquaNons for direct parameter computaNon:
• opening Ame of the mitral valve tMVO, • closing Ame of the mitral valve tMVC, • opening Ame of the aorAc valve tAVO, • closing Ame of the aorAc valve tAVC.
Results
• The equaAons permit us to directly and exactly retrieve model parameters without any model simulaAon. • This work proves that the model is structurally globally idenNfiable from the selected model outputs. • Finding bounds on the equaAons provides physiological limits on the parameter values.
Reference
*Hann, C. E. et al. Unique parameter idenAficaAon for cardiac diagnosis in criAcal care using minimal data sets.
Comput Methods Programs Biomed,
99(1), 75-‐87, 2010. • This method is not clinically applicable as it relies on ventricular
pressure, which is not available in a clinical se]ng.
• This invasive measurement can easily be replaced by the result of model simulaAons, implying iteraAons of the method*.
This work was financially supported by the FederaAon Wallonia-‐ Brussels (AcAons de Recherches Concertées).
Conclusion
Acknowledgments
A.Pironet@ulg.ac.be
Contact
Pulmonary
vein Mitral valve AorAc valve resistance Systemic Vena cava Ppu=Plv(tMVO) =Plv(tMVC) Rmt= RtMVC tMVOPpu Plv(t)dt SV Rav= RtAVC tAVOPlv(t) Pao(t)dt SV Rsys= ¯ Pao Pvc HR·SV Eao= Rsys tAVO tAVCln ✓ Pmaxao Pvc Pmin ao Pvc ◆ Elv= 1 SV ✓ Pao(tAVO) e(tAVO) Pao(tAVC) e(tAVC) ◆ 1