|Authors||H. Finsberg, J. Aalen, C. Larsen, E. Remme, J. Sundnes, O. A. Smiseth and S. Wall|
|Title||Assessment of regional myocardial work through adjoint-based data assimilation|
|Project(s)||inHeart: In Silico Heart Failure - Tools for Accelerating Biomedical Research|
|Publication Type||Talks, contributed|
|Year of Publication||2017|
|Location of Talk||Oslo, Norway|
|Type of Talk||International Conference on Computational Science and Engineering, In memory of Hans Petter Langtangen|
Assessment of regional myocardial work through adjoint-based data assimilation
To achieve efficient pumping of blood to the body, the healthy heart contracts in a synchronous manner. However, heart disease can alter how the heart is activated during a beat, and dyssynchronous contraction can occur, reducing the overall pumping efficiency. Advanced treatments exist for such cases, but selecting patients likely to respond can be challenging. The existing selection criteria, based on organ level measures of activation and contraction, have relatively low specificity. It is therefore of interest to extract new biomarkers to help better identify potential responders. Here we explore one example of a potential biomarker, the regional myocardial work , a measure of cardiac efficiency, using a computational model of cardiac mechanics optimized to patient specific data using a high level adjoint based data assimilation method.
Left ventricular (LV) geometry was obtained from 4D echocardiography, and the segmented chamber was modelled as an incompressible, continuous hyperelastic body described via an transversely isotropic material law. Active force development was modeled through additively decomposing stress into passive and active stresses, the latter added along the cardiac fiber direction, defined by a rule based architecture.
The model was fit to 4D imaging of the LV through the cardiac cycle using an adjoint-based data assimilation technique, which automatically solves for the gradient of the solution with respect to local active stress, for highly efficient minimization of model misfit against collected data. Simulations were optimized both globally and regionally in 17 delineated segments. With these simulations, the amount of mechanical work performed between time point tm and tn could be regionally calculated through -
W(tm, tn) = ∫ S: ∂tE dt = ∑i S(ti-½): dE(ti-½)
S(ti-½) = 0.5*(S(ti)+S(ti-1))
dE(ti-½) = E(ti)-E(ti-1)
Here subscript t indicates the time point, S is the Second Piola-Kirchhoff stress tensor and E is the Green-Lagrange strain tensor.
We tested the method on healthy control subjects and patients suffering from left bundle branch block (LBBB). The results show an excellent fit between measured and simulated strain (R^2 = 0.8) and volume (R^2 = 1.0). The estimated regional myocardial work, assessed in these segments, shows clear differences between the healthy and diseased patients (e.g Mid Septal longitudinal wasted work ratio: 1.45 (LBBB), 0.24(Healthy)) and can potentially be used as a biomarker to map regional cardiac dysfunction.