Abstract

Active contour models are used extensively in biologic image analysis for object segmentation and description, allowing to represent a variety of complex structures within a common parametrization. For dynamic morpho-topological phenomena at intra-cellular, cellular, and supra-cellular levels for in vivo imaging, motion and deformations are necessary to provide a quantitative basis for time lapse analysis. To this end, optical flow techniques can be used to estimate motion vector fields in time lapse image sequences. We analyze the assembly of multi-cellular structures in vertebrate fish embryos regarding dynamics of cell adhesion, shape and cells organization, combining parametric active contours with optical flow. We use a semi-automated cell segmentation and motion estimation in images from time lapse spinning disk microscopy. First, cell membranes are segmented at a given time with active contours; second, optical flow vector fields are computed and used to predict the location of the contours within the next time frame; finally, a new active contour iteration adjusts to the contours to the cell membranes. Our iterative scheme allows a precise morpho-topology analysis over consecutive time frames minimizing the manual labour for cell segmentation in different in vivo models.