A segmentation and tracking program was developed to analyze C.Elegans early embryogenesis in order to reconstruct automatically the cell lineage. The algorithm consists of two phases :
Segmentation : Like other segmentation procedure, since the image contains a lot of noise, the images has been filtered with a 3D Median with different radius varying with time. Then the challenge is how to threshold; if the threshold is too low, some of cells stick together, with a high thresholding certain weak cells will be deleted. For each time the image stack is segmented with a low threshold, then for each object detected an iterative thresholding is applied on the pixels value inside the object in order to check if the object was actually composed of multiple objects. If the separated object has certain criteria (like volume, sphericity, compactness), they are considered as nuclei and the loop continue until all objects are processed.
Tracking : Once we have segmented all the objects for all time point, another technique is applied to follow each object by tracking all objects of current time (t) point to others in previous one (t-1).
For movement, the spatial nearest neighborhood distances between objects at t and (t-1) are calculated and error is minimized using distance cost.
For division, when the number of object in two time successive are not equal, by the same way we find their nearest neighbor. The object which is close to two object of (t-1) is checked to see if it was divided or not by applying some biological rules; the newborns must have almost the same volume and the parent must be located in between them. if both conditions are not presented, that object considered as a parent is associated to the nearest child and the other one reinsert to the operation. The same procedure is repeated for the rest.
Presenting author: Jaza Gul Mohammed
Organisation: université pierre et marie curie- Cellular Modeling and Biological Imaging
co-authors: Supervisor: Thomas Boudier