Abnormal stresses are hypothesized to be a key driver in remodelling processes associated with heart failure
However, it is currently impossible to measure stresses safely in vivo in a human heart. This necessitates the use of computational models
in cardiac stress calculation.

A key step to making simulated stresses useful for clinical practice is patient specificity. This means that the simulated stresses should
come from a computational model that has been calibrated to behave in the same way as a patient's heart.
Doing this typically involves first creating a patient specific geometry, and then using available clinical data to personalize the mechanics of a computational model.

One source of mechanical data that is currently available is dynamic left ventricular strain, which can be
obtained cheaply and efficiently using 4D echocardiography methods. In our study, we combine such strain data
with left ventricular pressure and volume measurements in order to match simulated bi-ventricular mechanics to those observed in the ventricles of a patient.
We formulate this matching as a mathematical optimization problem in which the least squares difference between
simulated and measured strains and volumes is minimized. This minimization is carried out using a gradient based optimization algorithm and an automatically
derived adjoint equation. As a result, we obtain patient specific stress maps that can be used
to improve the treatment of cardiac conditions in which stress plays a significant role.