Hi CPA Team
Using CPA to identify hits in an HCS assay I encountered an obscure but highly annoying bug (see attachments for documentation).
*** PROBLEM ***
Using the regular image filenames (eg PANDORA_080402180001_A01f00d0.DIB), the CPA image viewer always displays the image with ImageNumber 1, no matter what image I try to load. Even if I try to load image 1005, CPA stills loads the very first image (ImageNumber = 1). Strangely, it is only the image that is incorrectly displayed. When I click on “View->Show Nucleus Numbers”, the numbers labeling the objects appear in the right places – only the image is wrong. Except if I load image 1 of course. Additionally, the channels all look the same, only the colors differ.
However, if I change the first image’s name in the database and the file system to something else (eg i1h2ax.DIB), all images are loaded correctly. Image 1 is still loaded correctly, but now image 1005 is loaded correctly as well. As before, if I click “View->Show Nucleus Numbers”, the numbers appear in the right place, but now they actually match the images. Furthermore, now the channels work. Channel 0 (eg H2AX) is not merely the same as Channel 3 (eg DNA) but there is really a difference in the displayed morphologies.
*** DOCUMENTATION ***
Please find the following files attached:
• A segment of the csv files used to populate the mySQL database (both from the per_image and per_object table)
• mySQL setup script
• CPA properties file
• Screenshots documenting the bug
• MainThread console outputs for both cases (original filename and changed filename)
*** SYSTEM CONFIGURATION ***
Mac OS.X 10.6.8 (64 Bit)
mySQL Community Server 5.5.15
*** PROPOSED SOLUTION ***
I think there might be something wrong with the load_using_bioformats routine. The “changed file name”-console output states the following every time I attempt to load an image:
Failed to load /Users/johnny/Research/cellprofiler_test/cpa_test/Kin07_6hr/images/PANDORA_080402180001_G24f02d2.DIB with bioformats. Use PIL instead
Traceback (most recent call last):
File “cellprofiler/modules/loadimages.pyc”, line 2807, in provide_image
File “cellprofiler/modules/loadimages.pyc”, line 2957, in load_using_bioformats
File “cellprofiler/utilities/jutil.pyc”, line 439, in method
File “cellprofiler/utilities/jutil.pyc”, line 377, in call
JavaException: Index: 5, Size: 0
It says “Use PIL instead”. Well, PIL seems to work
On the other hand, the “original file name”-output does not show any warnings/exceptions at all, but the image viewer does NOT work.
This bug could be connected with the DIB file format somehow, as the PNG files from the CPA example work well. It could also be the image names.
Alright, that’s it from my side. Hope that was helpful .
Computational Systems Biology Initiative
Koch Institute for Integrative Cancer Research
Massachusetts Institute of Technology
documentation.zip (3.19 MB)