Identifying germinated vs. non-germinated pollen

Hello,

I am trying to come up with a method to count the number of germinated pollen grains as well as the total number of pollen grains (both germinated and non-germinated) present in the image.

My goal is to have Image J calculate these values which I can then use to calculate the percentage of viable pollen grains in each image.

I have had success in counting the round, non-germinated pollen grains using Process >> Find Edges, followed by Image >> Adjust >> Threshold and finally Analyze >> Analyze Particles.

While this works well on circular, non-germinated pollen, I am unsure how to count germinated pollen which has a large pollen tube protruding from the pollen grain.

Does Image J have a plugin that can detect abnormally shaped objects? Do I need more specialized software such as CellProfiler?

I have attached an image for reference.

Thank you

Hi,
Separating germinated from non-germinated seeds shouldn’t be too challenging if you can easily threshold the image. Once you have a well thresholded image containing all your objects, it should be possible to separate the germinated seeds (long processes) from the non-germinated seeds (roundish) based on their shape parameters (roundness) using the Particle Analyzer or Extended Particle Analyzer.
It probably could be done with your sample image, but there are a number of issues with the image that make the task more challenging:

  1. The illumination is rather uneven. A background subtraction will help with that, but perhaps the imaging/illumination could be improved.
  2. Same areas of the image look rather washed-out/over exposed. Perhaps it would be possible to adjust the illumination to achieve a slightly better/more consistent contrast across the image.
  3. Some non-germinated seeds appear to be touched by pollen tubes from other germinated pollen, which is more challenging to differentiate. Perhaps a slightly lower pollen density/equal pollen spacing could help with that to avoid contact. This would also be helpful if in the future you decide that you might want to analyse the length of your pollen tubes.

Cell Profiler or some machine learning approach (e.g. Ilastik) could probably also do it, but it will always be easier if the image quality can be improved.

I hope this helps and provides you with some ideas,
Volko