Tracking the dynamics of keratin filaments

D. Kotsur, R. Yakobenchuk, R. Leube, R. Windoffer, J. Mattes. Tracking the dynamics of keratin filaments. pages Poster No.11, 2, 2018.

  • Dmytro Kotsur
  • Roman Yakobenchuk
  • Rudolf E. Leube
  • Reinhard Windoffer
  • Julian Mattes
SeitenPoster No.11

The networks of keratin filaments play an important role in motion, dynamics and mechanical properties of epithelial cells. The topology of the keratin network is highly dynamic and has significant variations in different cell types. An accurate manual quantification of the spatio-temporal network properties would require a lot of effort and would be extremely time-demanding. In order to process a large amount of keratin image sequences, we have developed an image analysis framework for automated localization and tracking of individual keratin filaments. The framework consists of two main routines: detection and tracking of filaments. The first step for both routines is the preprocessing of image sequences. At this step the image noise is reduced and image artifacts are removed. Using Hessian ridges enhancement and gamma correction, the contrast of keratin filaments is improved. Afterwards the image is binarized, using two-level thresholding and the skeleton of the network is computed. The detection routine is based on the analysis of the skeleton produced during the preprocessing step. It incorporates the filters which allow to select filaments with certain prescribed properties. The tracking routine uses the results of filament segmentation and follows the detected filaments through the image sequence without the explicit segmentation in the other frames of the sequence. It can also use manually labeled initial positions of the filaments. Initial filaments are represented as parametric curves, which are fitted to the next consecutive images using active contours approach together with a precomputed optical flow. The proposed framework facilitates precise analysis and classification of different types of filaments within living cells and allows us to link their occurrence to specific cellular conditions.