Abstract

Spatio-temporal analysis of multi-cellular systems at the single cell level has emerged as one of the fundamental techniques in cell biology, developmental biology, and experimental medicine. The vast amount of image data obtained by tracking experiments requires computerized approaches, either fully unsupervised or semi-automated to facilitate an in-depth analysis of the underlying biological principles. Automated cell tracking methods need to deal with different modes of cell migration, touching and overlapping of cells, as well as with changing cell numbers due to mitosis. In this paper, we explore the applicability of a rather general tracking paradigm based on fluid registration of the image sequence coupled to a level set segmentation. We demonstrate that a consistent combination of registration and segmentation can overcome limitations of previous registration based tracking methods and yields encouraging results in terms of tracking reliability and mitosis detection.

Keywords: Cell biology, Cell tracking, Fluid mechanics, Image registration, Level set segmentation, Variational methods.