Relate possible update

Hello,

it would be very useful if you could include in the Relate module the option (checkbox) of “not including children if no parent is found for them”. In other words, not including any object whose father is “0”.

Best.

Joaquin.

Hi Joaquin,

At the moment, the “Mean” values calculated by the Relate module only include objects which have parents. Objects without parents are disregarded. Is this what you are looking for?

Mike

I think I did not post the question correctly. Let’s assume I am detecting
100 cells in one image. Now, let’s assume that I am detecting only 7 out of these 100 cells containing certain other object inside. Finally, assume I ONLY want to save the features from the cells that contain these other object inside and not from the other 93 cells. As I cannot reject objects that do not relate with other object, I am generating a huge ammount of data that I do not really want. That is why I think it could be extremely useful to have the possibility of “filtering by relate”. I mean, to filter objects if they do not relate with other kind of objects. Looking forward to hearing your opinion.

Best.

Joaquin.

Hi Joaquin,

I understand what you are aiming for, and as you know it is not currently supported. I think it is a good idea, but difficult to implement. The way relate works now, it sums up all of the subobject measurements per parent object and gets the average per parent measurements. This is necessary because if you identify 500 speckles and you only have 20 nuclei, you have a problem when you try to export this data in to an object table of a database (if you have 20 rows for nuclei, you lose 480 rows of speckles data, and if you have 500 rows for speckles, what do you fill in for the other 480 rows of nuclei data). This is something we have debated quite a bit, and still don’t have a good answer for. The point being, if you wanted to exclude objects based on their containing subobjects, you would have to do the relate early in the pipeline before you do any measurements, and you would lose average measurements of those subobjects. Of course, I guess you could just have another relate module at the end of the pipeline as well. What we have been thinking about is having subobjects exported as a separate table, but this gets messy in terms of analyzing the data.

We are welcome to any ideas on how to deal with this better :smile:

Regards,
Mike

Here is something to try, if you are using the Developer Version and don’t mind making a minor change to the code:
Right now you can use the FilterObjects module to filter the parents by their AreaShape, Intensity, etc. This requires you to choose from a dropdown menu in the FilterObjects module that lists AreaShape, Intensity, etc.

So, go into the FilterObjects code and add the ability to choose the number of children option (this means you would need to add the word “Children” to the dropdown list). Then you would be able to use the FilterObjects module on the parents, to filter the parents by the number of children, using feature #1, which will be named subobjectCount, where the word “subobject” is actually replaced by whatever you called the children. Then, use your measure modules only on the filtered parents (whatever you decide to call them). Let us know if this works, and we will make the change in the main code for next release.

Anne

@Anne_Carpenter and @Mike_Lamprecht . This is a useful post and might be useful in solving my problem Relating 3 objects
but I don’t see where you can name children in the RelateObjects module in CP Version 3.0. I think that is a step I am missing would this also then allow you to get the numbers that @JGoni was wanting correct?

Thanks This did the trick! I was able to solve my problem with this. CP3.0Pipelinetest.cppipe (17.2 KB)