An approach to quantification of biaxial tissue stress-strain data

Delineation of the mechanical properties of biologic tissues is one of the cornerstones of biomechanics. Abundant data from uniaxial tests exist but these cannot be extrapolated to describe three-dimensional properties of tissue. Biaxial stress-strain studies have been performed using skin, blood vessels and pericardium. Quantitative description of tissue properties in these studies has employed either polynomial or exponential strain-energy functions. Interpretation of these data, however, is difficult because of wide variability of the estimated coefficients of these functions. This variability has been attributed to experimental noise, numerical instabilities in the algorithms, or to strain-history dependence. No systematic method has been proposed to evaluate the variability. This paper describes a statistically based approach to assessing the sources of and accounting for variability of coefficients in describing biaxial stress-strain data. Our data are from canine pericardium subjected to various combinations of simultaneous biaxial stretching. We first determine a suitable strain-energy function with the least number of free parameters that will fit the data reasonably. We then perform residual analysis to see if standard statistical methods can be used to assess the variability. If not, we use a nonparametric method called bootstrapping that is suitable for assessing the uncertainty in the coefficients. Using a five-parameter exponential strain-energy function, pericardial tissue is found to be strain-history dependent and anisotropic. These findings cannot be attributed to either experimental noise or instability in the numerical algorithms.