A computational framework for simulating the dynamics of keratin filament networks

• Saransh Vora
• Dmytro Kotsur
• Reinhard Windoffer
• Julian Mattes

In this work, we propose a biomechanical computational framework, which allows us to perform numerical simulations in order to mimic the behaviour of keratin filament networks as observed in microscopy data. The developed automated computational methods allow us to: (1) generate, by a randomized process, a filament network topology with predefined properties; (2) simulate the motile behaviour of the filament network; (3) synthesize image data based on the produced simulations. The synthetic image sequences can then be used to evaluate filament extraction and tracking algorithms as the underlying ground truth is available.

The core of our simulation module consists of a spring-mass-damper discrete mechanical system, which allows simulating elastic behaviour of the filaments. Additional stochastic external forces are also incorporated, which allow exhibiting the effect of Brownian motion and the influence of the actin network resulting in an inward-directed filament flow.

Based on the produced simulations we are able to synthesise artificial image data mimicking the behaviour of the filament network. Artificial images can be exposed to additional microscopy-related imaging artefacts such as Poisson/Gaussian noise, convolution with a predefined point spread function.