MorpholibJ - 3D Watershed Segmentation Help

Hello ImageJ! I am a high school intern who is working in a lab over the summer. I have been tasked with the challenge of identifying the volumes of plant cell walls in 3D image stacks. To do so, I have been using the plugin “MorpholibJ - Morphological segmentation” with the hope to make each cell stand out. However, I am having limited success. The software seems to not detect all the cells, or groups cells together that should not be grouped in the first place. Is there a better plugin I should use? Or maybe a way to manually detect the cells? Or a way to edit the original image to have more distinct cell walls?

Thank you in advance!

Hello and welcome to the forum!

It would be really helpful if you can post here (or with a link to dropbox/google drive, etc) an example of the images that are giving you trouble.

That being said, if the result provided by the Morphological Segmentation plugin is not satisfactory, you can always manually correct it using its post-processing panel or the Label Edition plugin.


1 Like


A complementary solution is to pre-process the image before segmentation.

Image stacks of plant cell walls may have a low signal to noise ratio, and some smoothing filters may make disappear some parts of the cell wall, making the segmentation more difficult. Directional filters, such as based on 3D Hessian can be useful. See for example FeatureJ, but you need to combine the results manually to compute a “plate-like” or “plateness” index.

1 Like

Thank you for your response and your warm welcome!

The first link I have attached is a link to the protocol I am using, and the second is a link to one of the .tiff pictures I am using. I did practice using the post-processing panel as you requested, but it seemed to have very little effect on the final result. I will also try the Label Edition plugin and tell you how that works!


Thank you! The Hessian filter seems to have fixed the problem! The pre-processing filter combined with the post-processing process of merging the label of cells seems to have fixed the problem! Thank you all very much!

1 Like