Data-driven modeling of heterogeneous viscoelastic biofilms.

Abstract

Biofilms are typically heterogeneous in morphology, structure, and composition, resulting in nonuniform mechanical properties. The distribution of mechanical properties, in turn, determines the biofilm behavior, such as deformation and detachment. Most biofilm models neglect biofilm heterogeneity, especially at the microscale. In this study, an image-based modeling approach was developed to transform two-dimensional optical coherence tomography (OCT) biofilm images to a pixel-scale non-Newtonian viscosity map of the biofilm. The map was calibrated using the bulk viscosity data from rheometer tests, based on assumed maximum and minimum viscosities and a relationship between OCT image intensity signals and non-Newtonian viscosity. While not quantitatively measuring biofilm viscosity for each pixel, it allows a rational spatial allocation of viscosities within the biofilm: areas with lower cell density, for example, voids, are assigned lower viscosities, and areas with high cell densities are assigned higher viscosities. The spatial distribution of non-Newtonian viscosity was applied in an established Oldroyd-B constitutive model and implemented using the phase-field continuum approach for the deformation and stress analysis. The heterogeneous model was able to predict deformations more accurately than a homogenous one. Stress distribution in the heterogeneous biofilm displayed better characteristics than that in the homogeneous one, because it is highly dependent on the viscosity distribution. This study, using a pixel-scale, image-based approach to map the mechanical heterogeneity of biofilms for computational deformation and stress analysis, provides a novel modeling approach that allows the consideration of biofilm structural and mechanical heterogeneity. Future research should better characterize the relationship between OCT signal and viscosity, and consider other constitutive models for biofilm mechanical behavior.