Tracking, counting, and measuring the intensity of protrusions

'm currently working with a team that is studying cell motility. We have obtained a stack of images from our microscope. We’re specifically interested in measuring the size, number and pixel intensity of the protrusions relative to each cell.

I first used the pixel classification module to create a model of what’s a protrusion, what’s a cell cytosol, and what’s our background. I’ve been trying to use the tracking with learning module to track the protrusions of each cell. What we want to extract from the data is 1) how many protrusions does each cell have, 2) where and how big each protrusion is 3) what is the ratio of the maximum intensity relative to the cytosol of the protruding cell.

I would appreciate guidance on the tracking module. I guess my general question is how can we track and measure the features of a cell (protrusions) that come into the field of view and then disappear.

Thank you again,


Hi Ali,
you have not posted any images, but I suggest that you should explore morphological operations (erosion and dilation) for the generated models. You will need to erode your images until the protrusions disappear and after that dilate it using the same number of iterations. This will give you cell body; difference with original image will produce areas of protrusions.
You can count number of obtained protrusions per cell and measure image intensity under protrusions and relative to cell body.
It should be quite effective but depends on your particular images
Hope this may guide you further,


Hi Ilya,
Thank you so much for getting back to me. I’ve attached a sample of the types of cells that we are dealing with. One of my biggest concerns is that some of the cells (and their protrusions) are not as bright as other cells. What software would you recommend for morphological operations that could track the protrusions of bright as well as the not-so-bright cells? I was thinking that I can crop the stack into pieces that it only shows one cell. I’m open to any suggestions that you might have.
Thank you again,
ME3 RBD DMSO copy-1.tif (5.3 MB)

Hi Ali,
well your task is rather complex and I suggest you to contact an image specialist at the local imaging facility, if you do not have large experience with image processing.
I think the CellProfiler is the best tool for doing the task. Below I show you the required steps; I will use MIB ( for the illustration purposes.

To start, you should have segmented model of your cells, I would expect that it is the most difficult thing in your case.
You’ve mentioned that:

I first used the pixel classification module to create a model of what’s a protrusion, what’s a cell cytosol, and what’s our background.

So, you should hopefully have already the segmented cells.

  1. Below an image with the model; all cells belong to the same material (have the same index in the model) and it is important that adjacent cells have gaps (I highlighted one of the gaps with a green circle)

  2. the next task is to separate cells to individual objects, where each cell will have its own index in the model. In MIB, right click over the material ‘1’ and select Get Statistics:
    In the statistics window, press the Run button to get areas of individual cells (If there are some small noisy objects those can be removed at this stage)

    To split object, select with the LMB the 1st and the last (Shift+LMB) objects and press the right mouse button to call a popup menu and select Objects to a new model:
    As result, each cell got an own index and you can see that they all have distinct colors:

  3. Now it is possible to get image intensity of each cell. To do that, convert the model into a different type: Menu->Models->Convert type->65535
    Next reopen the Get Statistics dialog as in step 2 above
    In the Get statistics window, select Objects->Model (if it is not there, reopen the window); Mode->Intensity and change MinIntensity to MeanIntensity and press Run. Now you should get mean intensity of each cell.

    You can export this by pressing the Export button

  4. The protrusions can be obtained using morphological operations. There are few ways, but the easiest is this:

  • Select Exterior in the Segmentation table:
  • Press Alt+A to select it (the corresponding area turns green)
  • Invert selection: Menu->Selection->Invert selection
  • select the proper step for erosion in the selection panel->Strel size. I put “3” and pressed Erode 2 times (z-key shortcut). As you can see that the green area shrank
  • Now we need to restore the size of the cells, to do that we need to apply dilation the same number of iterations as erosion (i.e. 2):
  • Remove the green area from the cells: press “S”
  1. Finally, we can calculate intensity and size parameters of the obtained protrusions:
  • Start “Get statistics” dialog, as in step 2
  • Make sure that: Objects->Model, Shape->2D, Multiple checkbox->ticked
  • Press the Define properties button and choose properties that have to be calculated
    The problem is that at this stage MIB calculates combined properties for all object belonging to the same material. So you will get area or mean intensity of all protrusions. If you really want to have them all split by individual object you have to choose corresponding object in the Objects dropdown and recalculate (I would normally write a plugin to do that automatically). At this point, you will also see that there are some very small object (few pixels in size) and it would be the best to filter them out.

Based on that I think CellProfiler would be the better choice but I am not totally sure whether you can get statistics for individual objects within multiple objects…

Best regards,

I added cellprofiler tag, so perhaps someone from CellProfiler team can comment on the task.
Best regards,

dear @Alikhoddam,
a part from the great suggestions of @Ilya_Belevich, I just want to suggest you to check this good plugin that some of the researchers of my institute use to analyse similar data (if I’ve understood well your request)

it’s Adapt a fiji plugin.

Emanuele Martini

1 Like