Automated cell segmentation

I’m working on segmentation in c elegans embryos. The dataset looks something like this:
I basically want to mark the cell membrane. I’ve tried several types of thresholding and morphological operations. Thresholding usually results in discontented cell boundaries. Morphological(using MorphoLibJ) operations usually results in over-segmentation or under-segmentation. I tried Trainable Weka Segmentation also[Problem: discontented cell boundaries].

The major problem is discontinuous cell boundaries in results.

Can anyone suggest some methods to achieve the required segmentation?

PS: The final aim is to track each cell (to get to know about their position vector, velocity vector, shape changes etc).

Hello @geekSiddharth,

It seems like an appropriate problem for the Morphological Segmentation plugin but you said you didn’t get proper results? Can you post here one of your images?

Another option would be to use the Trainable Weka Segmentation to obtain the probabilities of boundaries (to correct the discontinuous ones) and then apply Morphological Segmentation.

1 Like