I am new to using CellProfiler and am tring to use it to count the number of small mitochondria within a cell.
When I run the IdentifyPrimAutomatic module It has a problem where it is splitting mitochondria that are obviously attached.
THe following image is the parameters that I have set up fro IdentifyPrimAutomatic.
Any ideas of what the problem may be and how to solve it would be greatly appreciated.
Thanks for your time
There are several things you can do to approach this issue:
(1) Have you tried modifying the de-clumping parameters? Further down the module, there are two options: “Method to distinguish clumped objects” and “Method to draw dividing lines”. Usually, you want to set the first to either Intensity or Shape, and the second to Intensity or Distance. Both of these affect whether single objects are split apart and if so, how.
(2) Using Test mode. The Help has more info on how the options in (1) work, but one additional useful setting is the very last one, “Do you want to run in test mode…” where the various combinations of the two parameter settings are displayed. Which combination you pick depends on which one gives you the best results.
(3) Smoothing: Whatever you choose for these options, how the de-clumping takes place is also dependant on the amount of smooth applied to the image to remove spurious noise that can impact the results. Under “Size of smoothing filter”, the default is Automatic, but you can set this to whatever value to want (the value determined by Automatic is shown in the module window). Increasing the value will lead to less splitting, but possibly more false negatives, and decreasing it will split more, but with likely more false positives.
(4) Set “Try to merge…” to “No”: This parameter should be used with caution since you may run into a situation where the module is trying to split objects apart and then trying to merge them back together. The help for IdentifyPrimAutomatic has some further explanation on the use of this parameter, but in general you probably want to set it to ‘No’ and make sure that your other settings work well.
However, from the looks of your image, it looks like you are using a binary (i.e., black and white) image as your input. Is this the case? Ordinarily, IdentifyPrimAutomatic is used on a grayscale image, using the gray values to perform the identification and de-clumping. If it is binary, that will also affect how well the comments in (1-4) work for you.
Hope this helps!