Section: New Results

Spatio-temporal registration of embryo images

Participants : Grégoire Malandain, Léo Guignard [Virtual Plants] , Christophe Godin [Virtual Plants] .

This work is made in collaboration with Patrick Lemaire (CRBM).

Current imaging techniques can capture temporal sequences of 3D images with very high time resolution over several hours. Comparing sequences covering the same time period opens the way to the study of developmental variability. Stitching together sequences captured from different embryos may help producing a sequence covering the whole development of the animal of interest. For this, it is necessary to align two sequences in both time and space.

We developed a method to align two 3D+t time series, based on the detection and pairing of 3D+t landmarks. These landmarks, which correspond to periods of fast morphogenetic change, are deduced from the analysis of the non-linear transformations that allow to co-register pairs of consecutive 3D images in each sequence (see figure 7 ). [12] .

Figure 7. Spatio-temporal registration of two time-series of embryo. Enlarged renderings indicate the registered timepoints. Notice that the temporal registration is not linear since the interval length between two registered time is different from one embryo to the next.