Problematic neighbor counts


I have a set of images coming from a single microarray scan of cells, and different images correspond to different wells on one glass slide. I am trying to get some statistics of each cells’ close neighbours using the following settings:

However, when we looked at the analyzed data, in Well #1 (=Image #1), the number of neighbours are all 1. In Well #8, the number of neighbours are all 2 (with a single exception). This was extremely unreasonable. Especially for #8, it is almost impossible to have such a perfectly even distribution in an unpatterned well. Moreover, while there is no intrinsic differences between the wells, the neighbour counts on other wells do vary from 2 to 200.

Hence we took a more careful look at the data, and found further evidence that something goes wrong (at least in well #1):

From the 3 highlighted lines, we could derive the following information:
Cell #1, whose closest neighbour is 24.5 pixels away, is considered to have 1 neighbour within the defined range (I think the range is 15+15=30 pixels, as both identified objects would be expanded equally)
However, Cell #8, which is 24.5 pixels from Cell #1 and 5.8 pixels from Cell #11, is also considered to have 1 neighbour within the range.

Is that actually a bug? Or am I misunderstanding something?
I have also attached the images, the pipeline, and a copy of the output excel sheet. The software version is 2.1.1. Thanks very much.

Input Images and Pipeline.rar (1.94 MB)

Hi Luke,

I’ve managed to replicate this issue with your pipeline/data using 2.1.1. However, it doesn’t occur with our “bleeding edge” unreleased versions. I’m hard pressed to account where it come from or how it was solved; the closest bug report on it is this one: … ssues/1179. But it doesn’t seem to correspond to your issue.

You won’t see ithe fix until the next release, but if you’re daring you can download our latest trunk build from this page ( … TrunkBuild) subject to the caveats mentioned.


Hi Mark,

Thanks for your reply :smile:
So we’ll just wait for the new release.