Object Identification Problem


Here is a screenshot of my current settings for the ‘IdentifyPrimaryObjects’ Module, they are quite lenient:

My test module is also very simple, as follows:

Load Images
IdentifyPrimary Objects

Here is my problem: Of the four image sets which I am analyzing, all objects are correctly being segmented. However, only in two of the four images are the objects being accepted (green outline). The other two images have similar objects which are identified correctly with good accuracy, but for some reason are not being accepted as objects (yellow outline).

Any help with this would be greatly appreciated!
Thank you,

Hello, it appears that you are keeping objects that are touching the border so this should not be the source of the problem.
Is it possible for you to send your code with an example image for which you get the yellow outline?
Kind regards,

Hello vchernys,
Thank you for responding! Here are the modules I have selected to use for this pipeline. The purpose of this particular pipeline is to identify certain RCC cells which should be (+/-/+/+) for the four channels we used to image, this is why I have four: LoadImages and IdentifyPrimaryObjects.

Here is an example of the windows that come up when I hit AnalyzeImages, this is how the segmentation and ensuing object identification (or, lack thereof) occurs:

As you can see, the objects are segmented correctly in the CD45 channel, and yet none of them are accepted. Please let me know what I can try in order to correct this.

Thank you.


This looks like a bug in master that was fixed last week; if you haven’t re-pulled since then, can you do so now and see if that fixes your problem?

Also, do you mind if I ask why you’re using LoadImages rather than the typical load modules? We’ve been considering deprecating LoadImages so I’m just wondering if there’s a particular use case we don’t know about that isn’t well handled otherwise before we make that decision.

Hi bcimini,

Thank you, I will try that.

I chose to use LoadImages because I needed to group images. I could not find this feature in the typical Load Modules. Have I missed it? If not, could you somehow include that feature in the Load Modules?


You certainly can group images in those modules. Allow me to attach this from the CP help section on “Configuring Images for Analysis”.

Once you have used the Images module to produce a list of images to be analyzed, you can use the other Input modules to define how images are related to one another, give them a memorable name for future reference, attach additional image information about the experiment, among other things.

For more information on these modules and how to configure them for the best performance, please see the detailed help by selecting the module and clicking the button at the bottom of the pipeline panel, or check out the Input module tutorials on our Tutorials page.

Hi bcimini,
Thank you, I will try that. However, I still cannot get the IdentifyPrimaryObjects to accept the CD45 objects. Any suggestions would be helpful, thank you!

It’ll mean remaking your pipeline, but you may want to use the stable instead of the bleeding-edge release for the time being as we’re actively updating IdentifyPrimaryObjects for the 3.0 release at the moment- you can use git checkout stable to access that branch. If you want to make a bug report that will certainly speed the process of it getting fixed- please include your pipeline, images, OS, what commit you’re working from (just to help our software engineers trace it down) and anything else you think might be helpful.

Hi bcmini,

Do you know when the 3.0 release will be?


Sometime this fall; we don’t have a more exact date than that yet.

Ah, in the meantime, do you suggest working with an earlier version, but essentially using the same pipeline?

Yes, precisely- if you use our stable or beta branches you’ll get much more reliable performance than trying to work on the bleeding edge. There are some fun new features in 3.0, so I definitely recommend upgrading when we’re closer to release (or after release), but for now you may find it less frustrating to stay with stable.

Thank you! I will use one of those then and see what happens! :slight_smile:

Sorry, one more question about the NamesandTypes Module. For the purposes of my project, I will need to analyze images in groups, however, these images cannot be analyzed together, they will need to be analyzed separately for their signal in each channel. Is this possible to do with NamesandTypes? I just found that you could group images together, but when it came to measuring object intensity correlation between channels, there was no way that I could find to do it. Please let me know what I can do. Thanks!

Hi, sorry, didn’t see this was updated again until just now!

By “group”, do you mean something like “I have many channels from the same XY coordinates”? If so, yes, this is precisely what NamesAndTypes was meant to do, I recommend reading the part of the manual I linked to in a previous post to learn more about how to do this. Essentially in NamesAndTypes you add an entry for each channel, and tell CP what’s unique in each image’s name or extracted metadata to help it assign the images into “sets” (ie “DAPI” images contain “ch01.tif”, “GFP” images contain “ch02.tif”, etc)- the help article goes into it in more detail.

The Group module is something separate- it’s mostly relevant when you have many images that need to be sorted into discrete analysis runs (ie tracking objects between each frame of a given movie but then starting fresh when a new movie is loaded, calculating per-plate or per-well averages, etc). Again, this will be covered in the help section.