Simulation, Identification and Statistical Variation in Cardiovascular Analysis (SISCA) - a Software Framework for Multi-compartment Lumped Modeling

Abstract

Modeling approaches that are suitable to cover a wide range of real world scenarios in cardiovascular physiology could not be obtained so far, because many of the system parameters are uncertain or even unknown at all. The natural variability and statistical variation of cardiovascular system parameters in healthy and diseased conditions is assumed to be one of the characteristic features to understand cardiovascular diseases in more detail. Within this paper, a novel software framework for statistical variation, system identification, patient-specific simulation in cardiovascular system modeling is described by a multi-model statistical ensemble approach using dimension reduced multi-compartment models. The data driven approach is applied to model pre- and post-surgery clinical data sets from a patient (13 years old, 148cm, female) diagnosed with coarctation of aorta (CoA). A patient specific structure was obtained by the remapping of geometric parameters according to available MRI and metadata. Characteristic periodic boundary conditions were generated from pressure measurements, the systemic resistances and mean flow conditions at the outlets were adjusted subject to a weighting scheme, while systemic compliance and viscous parameters were fitted by reweighting the elastic modulus to the peak flows and pressure levels of all compartments. Archived model parametrization reflects the post-treatment data set and was used as a prior for modeling the pathologic pre-treatment state. According to geometry data, the stenosis was implemented for CoA modeling, such that the pressure drop was reproduced. In both scenarios, the simulated flows and pressure amplitudes are correctly reproduced by the simulation and stenosis and stent treatment were adequately reflected. Furthermore, the pre-treatment cross stenosis phase shift of the pulse wave is fairly well reproduced by the simulation. However, significant decrease of unrealistic phase shifts within the measurements was observed in the post-treatment after stent implementation, which is assumed to raise from specific systemic or measurement conditions that occurred during surgery. While patient specific modeling heavily depends on an effective and highly versatile modeling process, the methods and results presented in this paper suggest that the conditioning and uncertainty management of routine clinical data sets have to be further improved to obtain reasonable results in patient-specific cardiovascular modeling.