Détection et suivi d’objets ponctuels dans des séquences d’images de fluorescence Detection and tracking of punctual objects in sequences of fluorescence images

We present a new method for automatically detecting and tracking, in sequences of 2D fluorescence images, punctual objects which are intermittently visible. We demonstrate the ability of our approach to track, in live cells and at a molecular level, membrane proteins specifically attached to quantum dots, which are new fluorescent probes with great prospects for ultrasensitive in vivo imaging. The fluorescence 2D image stack is considered as a 3D image in a spatio-temporal space. By applying a detection process based on a fluorescence image model, we first obtain an unstructured set of points in the 2D+T volume. The trajectories of moving points are then viewed as minimal paths in a metric which depends on the 3D image features. These trajectories are retrieved with a perceptual grouping approach based on the Fast Marching algorithm. Our method, which requires few parameters, is illustrated by studying the dynamics of glycine receptors in the neuronal membrane.