waehlby cell clump splitter

Hi guys, I'm a bit lost using this node. Hopefully someone can point me in the right direction. First of all it take a tremendous amount of time before this node finishes completion. ImageJ's watershed function does this usually within a second, while this node usually takes around a minute or more.

And once it is complete the results are really poor. I tried a bunch of settings but this problem keeps popping up. Here are some links to screenshots I took. Putting in an image in a post on this forum doesnt work with uploaded images, i gave up on trying that.

http://prntscr.com/97ggs1

http://prntscr.com/97ggxx

http://prntscr.com/97gh1c

These were made with the following settings: Distance merge threshold: 1 Gauss size: 100 Size merge threshold: 1. I did this within the counting cells example workflow, with the thresholder set to manual and the value to 125.

Are there any settings I should change, or is there a better node, or perhaps a ImageJ macro to do this better and faster?

I also just tried the Seeded watershed node. However its performance is better, no splitting through the middle of particles anymore, its computation time is still really horrible.

Something else I noticed is that some particles receive two or even three labels. I think this is also one of the reasons why analysing takes so long, however I can't figure out where this comes from.

Hi Luuklag,

first about the WählbyCellClumpSplitter: Indeed, the current implementation is (a) very sensitive to the parameter settings and (b) not the fastest. We are working on (b) at the moment, such that the node is way faster. There is an example workflow on the example server for CellClumpSplitting, maybe this helps.

About the Seeded Watershed: Actually the WählbyCellClumpSplitter bases on this node, therefore it's not suprising that it is slow ;-)

Here is what I suggest until the node is faster:

* Did you try the predefined ImageJ Macro 'Watershed'? You can apply it on any BitType (BitMask) image and it should split your cells.

* You can also try to use the Morphological Labeling Operations (Erode). Does this help?

Christian

Hi Christian,

Using the standard Watershed Macro really helps, although this brings me back to an old problem I had with that feature, that it makes splits were it shouldn't. http://forum.imagej.net/t/watershed-making-unwanted-splits/396/6

Would there be a way to get the watershed from the biovoxxel toolbox working as a macro?

Regarding eroding, I already do this in pre-processing as to remove some background noize etc. However I am interested in the size of the particles, so every erode has to be followed by an equal dilate, as not to compromise particle size too much.

 

Luuk

Got it working :)

perfect! let me know if I can help further!

 

> However I am interested in the size of the particles, so every erode has to be followed by an equal dilate, as not to compromise particle size too much.

See Morphological Image Operations Open/Close for this purpose.