Is there a way for cellprofiler to differentiate btw colors?


#1

Hi, I have an image of ki67 cells that I really need help with. I have been able to count the total number of nuclei in the image using cellprofiler by inverting the image and then asking it to count the total number of nuclei by using primautomatic.

However, as you see in the image, there are brown and gray nuclei. Is there a way for cellprofiler to pick out just the brown nuclei and give me that number?

There’s the link to the picture.

Thanks!

img222.imageshack.us/img222/7295 … ityjv4.jpg


#2

Hello,
You can use the ColorToGray module and select “split” to separate the blue, green and red channels. To try to pick out the “brown” nuclei, you can try combining 2 channels (say–blue and green) by using the Combine module and subtract it from the third channel using the Subtract module.

~martha


#3

Hi Martha,

I tried your method and I couldn’t get it to work. A combined picture of Red and Blue or another combination subtracted from the original usually cut out the brown nuclei along most of the gray nuclei. So I tried measuring the intensities of the cells instead and then used the filterbyobjectmeasurement module to get the brown cells.

Thanks for your suggestion though.

I do have another question, I see that the number of each type of nuclei is shown uner Number of Objects if i do Measureareashape. Can cellprofiler take the ratio of those two numbers for me somehow?


#4

Hello,
There are 2 ways to filter: Use the FilterByObjectMeasurement to filter by size (area or perimeter), then use the MeasureAreaShape again for each type of cell (larger/brown and smaller/gray). The second is much easier to do, but it is outside of CellProfiler. If you import your data to a database (which could be as simple as an Access database) and filter using a query. We generally do any filtering in MySQL databases, rather in CellProfiler itself.

Hope this helps!
martha


#5

thanks for all of your help!