We are developing a system for the in vitro detection and tracking of cells in confluent cell culture. As a particular example we used human endothelial cells. The cells were exposed to fluid shear stress in the BTF-system and life cell imaging (phase contrast) was performed. As a consequence cells migrate, elongate and align with the direction of flow.

The task of detecting and tracking cells is characterized by the following challenges: Firstly, at the used resolution a cell has no visible individual features to distinguish it from the other cells. Secondly, neighboring cells are sometimes difficult to separate. Thirdly, the detection of cells must be robust to floating cell debris and image noise.

Our method is based on a statistical model which describes the correlation between one measurable and four non-measurable quantities. The measurable quantity is a sequence of grey scale images, i.e. a movie. The non-measurable quantities are constituted by: A segmentation of the image sequence. A position labelling determining the center position of each cell in each frame. A shape labelling determining shape parameters, where the shape of cells is approximated by ellipses. A transition labelling, which establishes a matching between detected cells of consecutive image frames. From the most probable configuration of these hidden variables, given an image sequence, one can recover the trajectory of each cell.

The trajectories can be used to study several parameters of cell locomotion, cell morphology an cell-cell interaction. Migration of endothelial cells is an important step in angiogenesis and therefor plays a critical role in wound healing. Thus through the acquisition of detailed quantitative data, particularly after molecular manipulations, this tool might contribute to a better understanding of these processes.