MeasureObjectNeighbors

Hi all,

I recently began using CellProfiler to quantify fluorescent intensity data from Immunohistochemical slides of nucleus pulposus cells. In order to quantify the percent of cells positive for the protein target of a particular stain I am trying to count the number of DAPI stained nuclei that are “neighbors” to the stain of interest. This seems to be working fine when the images have few stained objects. However, on images where there are many stained objects I am receiving the following error:

Traceback (most recent call last):
File “cellprofiler\pipeline.pyc”, line 1934, in run_image_set
File “cellprofiler\modules\measureobjectneighbors.pyc”, line 432, in run
IndexError: index (113) out of range (0<=index<112) in dimension 0

I have attached my pipeline and an example low stain density stained image and high stain density image. I believe that what I am trying to do with this pipeline is quite simple. I want to identify and count the DAPI stained nuclei and then determine how many cells are positive for the protein of interest by measuring how many nuclei are within a certain range of the protein objects (in this case either collagen X or osteopontin). Because I was unable to get good images of the cell morphology on brightfield I believe that this is the best way for quantifying percent positive.
I would greatly appreciate if anyone has any advice regarding my approach to counting percent positive or (more importantly) regarding the error message that I am receiving.

Thanks!








WKin.cppipe (52.9 KB)

I believe I found a solution for my aforementioned issue. Rather than using measureobjectneighbors I expanded the nuclei to a reasonable range (~5 pixels, what I would have had measureobjectneighbors use as a threshold), and then used relateobjects to make the red protein stained objects the children of the DAPI stained nuclei. Then I used classifyobjects to quantify the percent of enlarged nuclei that had at least one child object. In case anyone was wondering I could not have simply used relateobjects to begin with since the two stains did not always overlap within the cells.

I managed to replicate the bug and filed a report here: github.com/CellProfiler/CellPro … ssues/1179

Re: percent positive readout - We have a percent positive example pipeline available from our examples webpage that you might want to take a look at, if you haven’t already: cellprofiler.org/examples.sh … ntPositive
-Mar