Segmentation of the body axis into a series of repeating units is a canonical strategy in morphogenesis, and evidence of this can be seen in the skeletal system of all vertebrates. In vertebrates (and even some insects), the developmental process of segmentation is characterized by the rhythmic and sequential addition of segments (called somites) to an elongating body axis, and is regulated by an oscillatory mechanism -- the segmentation clock.
Understanding the development of the body plan in zebrafish requires high-throughput microscopy and image analysis of mutant and transgenic fish. We are developing automated methods for identifying and segmenting the somites from time-lapse microscopy images. To segment the developing somites, we use a constellation model where each individual part is represented by an active appearance model. We initialize each active appearance model using a random forest classifier, and make use of inference on graphical models to determine the optimal constellation.
Figure. Typical image data before and after semantic segmentation