Abstract
Accurate estimation of myocardial stiness is an essential step in the developmentof realistic patient specic models of the heart, and can provide a reliable biomarker
for disease stratication. Current methods for the estimation of myocardial material
parameters are limited by the low identiability with clinically available data. This
PhD work addresses the general problem of the compromise that needs to be made
between model delity and parameter identiability.
Parameter estimation is aected by the quality of clinical data used, the level of
realism (or complexity) in the mathematical model of the ventricular mechanics, and
the strategy adopted for comparing the data derived observations to the outputs of
the model. This work focuses on the last aspect aecting parameter estimation, and
proposes a novel cost function (CF), a novel criteria to compare data and model that
improves the identiability of parameters without any loss of model delity.
Using a popular transversely isotropic law for modelling the myocardium, we focus
on the reported coupling between the two main parameters, the coecients of the linear
and exponential part of the strain energy. The hypothesis investigated is that these
two coecients can be decoupled by a novel CF that is based on the analysis of the
strain energy stored in the myocardium and the external work performed while lling
the ventricle.
We then investigate the ability of the novel and existing CFs to dissociate the
coupled material parameters. In this context, four geometry based CFs, that can be
readily obtained from clinically available imaging data, were examined together with the
novel CF. Finite element (FE) models of the myocardium in diastole were constructed
from both synthetic and clinically derived datasets and parameter sweep simulations
over the two main parameters of the material law were conducted. Our results showed
that all geometry based CFs conducted to the same coupling of the parameters, both in
silico and in the clinical data derived models. In both these models however, the energy
based CF managed to isolate one of the parameters, and therefore in conjunction with
one of the other geometry based CFs can uniquely identify the parameter set.
In conclusion, we introduce a novel pipeline of a combination of CF that can be
used for unique estimation of material parameters of passive myocardium within a
standard FE framework. We demonstrate the method's accuracy in silico and also its
applicability to clinical datasets.
| Date of Award | 2016 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Steven Niederer (Supervisor), Pablo Lamata de la Orden (Supervisor) & Nicolas Smith (Supervisor) |