Fast Tracking Algorithm of GFP-Transfected Living Cells Based on the Chan-Vese Model

Warning

This publication doesn't include Faculty of Education. It includes Faculty of Informatics. Official publication website can be found on muni.cz.
Authors

MAŠKA Martin MATULA Pavel MUNOZ-BARRUTIA Arrate ORTIZ-DE-SOLÓRZANO Carlos

Year of publication 2011
Type Conference abstract
MU Faculty or unit

Faculty of Informatics

Citation
Description We present a general, robust, and fast approach for tracking GFP-transfected living cells in time-lapse series acquired using a confocal fluorescence microscope. The proposed tracking scheme involves two steps. First, the coherence-enhancing diffusion filtering is applied on each frame in order to reduce the noise and enhance flow-like structures. Second, enhanced cell boundaries are detected through a minimization of the Chan-Vese model that divides an image domain into two possibly disconnected regions of minimal variance. To speed up the second step, final contours from the previous frame are taken as seeds in the next one. The minimization of the Chan-Vese model is implemented using a fast level set-like algorithm (Maška et al. 2010) achieving near real-time performance in 2D. Furthermore, this algorithm has been integrated with a topology-preserving constraint based on the simple point concept from digital geometry in order to preserve known topology from the first frame throughout the entire time-lapse series. Such constraint provides a simple and inherent mechanism for keeping the boundaries of two cells separated even if the cells get touched in subsequent frames. The potential and preliminary results of the proposed tracking algorithm are demonstrated on 2D as well as 3D time-lapse series. The image data used in our experiments was acquired using a Zeiss Cell Observer Spinning Disk confocal microscope equipped with a 20x Plan Apo (0.85 NA) objective lens. The execution time of the proposed tracking algorithm is about 0.55 seconds per a 2D frame of the size 512x512 pixels and about 30 seconds per a 3D frame of the size 512x512x42 voxels.
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.