Prediction of failure location in heterogenous PDMS and infarcted left ventricular extracellular matrix via deformation and structural imaging.

Cardiovascular tissue rupture events in-vivo are dangerous and challenging to predict. Material failures occur in locations of peak stress; however, estimating local stress in a heterogenous, anisotropic tissue is not generally feasible. Image-based data can often be acquired in a clinical setting. We therefore sought to create an algorithm to determine the location of failure in heterogeneous materials based on imaging data alone. Two materials were tested: PDMS with a soft inclusion and decellularized infarcted myocardium (LV ECM). Samples were biaxially tested while undergoing digital image correlation (DIC) and quantitative polarized light imaging (QPLI) and then pulled to failure. We tested a combination of metrics to predict failure location, including the sum squared difference in the deformation gradient (Dif) and the direction of deformation (Dir) from pre-failure DIC, and the intensity, which is a measure of tissue density (Dens), and the degree of anisotropy (Anis) from QPLI. We also considered data from various biaxial extensions. We computationally removed mesh edges until the sample was divided into two partitions denoting the failure. For PDMS, when all biaxial extensions were considered Dif produced the best failure predictions in all samples. For LV ECM, however, the most successful predictions occurred when either the equibiaxial loading alone or equibiaxial plus strip-uniaxial extensions were included rather than all extensions. Further, weighting factors Dens*Anis or Anis alone produced the best predictions. Thus, for fibrous tissue, the underlying structural data from QPLI was the stronger predictor of failure location. Tissue samples tended to fail either at the infarct edge or in thin regions. Similarly, PDMS samples failed at the inclusion, suggesting the variability in material properties influenced the location of failure. Ultimately, this image-based failure prediction algorithm may have utility outside ex-vivo testing, especially when tissue structure can be deduced.