1

we have to measure the intensity of the fluorescence in certain regions of images using imagej. we came up with the below steps to measure the intensity. while it does seem correct, my question is --> are we actually measuring intensity correctly using the following steps or are we erroneously measuring something else and believing that that value is the intensity?

  1. Make the image a 8-bit
  2. Threshold the image (Image > Adjust > Threshold) to outline all the regions and click Apply
  3. Open Analyze > Analyze particles. Make sure “add to manager” is clicked
  4. Analyze > Analyze particles > Show > Bare Outlines. This will open a new image.
  5. Open the color microscopy image. Then, Image > Overlay > From ROI Manager.
  6. Image > Overlay > To ROI Manager.
  7. In ROI Manager: press “measure.” (a Results window with individual data points will pop up)
  8. Right click in the Results window and click Summarize.
  9. Record mean intensity data

are we correctly measuring mean intensity data using the above steps?

user1911720
  • 11
  • 2
  • 2
  • 5
  • In you're previous question (http://stackoverflow.com/questions/13925995/options-to-convert-16-bit-image) you asked about converting to 8bit. This conversion changes the pixel values so that it won't be the same as your original. If the intensity is the original mean pixel value of the image then you will probably be getting an error from this step – Codey McCodeface Jan 03 '13 at 12:24
  • @medPhys-pl, this is for a different image set and in this set (thankfully!) all of the images are in 8-bit color and we have to "black and white" the images. To do that, we change them from 8-bit color to 8-bit. so would this still mean that there is an error because the conversion changes are remaining consistent and not changing the pixel values? – user1911720 Jan 03 '13 at 15:02
  • 1
    It's not clear to me what it is that you are trying to measure and if it is well defined. To convert colour to grayscale you have to add up the red, green and blue components and to sum them to get the gray value so 'gray = x*red + y*blue + z*green'. The problem is x,y,and z are not well defined so the final measurement isn't well defined. Have a read of the Wikipedia article http://en.wikipedia.org/wiki/Grayscale – Codey McCodeface Jan 03 '13 at 16:17
  • @medPhys-pl: Would there be any practical benefit to focusing on the original image's RGB channel that most closely the spectral peak of the fluorescent dye used? – trashgod Jan 05 '13 at 22:22
  • @trashgod: Speaking as a physicist as well as a computer scientist I would say that it still doesn't mean much to me. The question is as much scientific as it is a software issue. Perhaps we should close or move? – Codey McCodeface Jan 07 '13 at 10:47
  • @medPhys-pl: It's a good question, but I'm voting to close as off-topic. My experience is with data having a known spectrum, which is retained as metadata for later use. Sorry, I meant to say "most closely _matches_ the spectral peak" above. – trashgod Jan 07 '13 at 15:09
  • Sorry about the question being off-topic. Thank you everyone. I will try to focus on the original image's RGB channel and look into that method @trashgod and i do see you point medPhys-pl. if x y z are not well defined, that means that final intensity is erroneous. – user1911720 Jan 08 '13 at 14:43
  • @trashgod, where should i post the question so that it is not off-topic? (like another website?) thank you – user1911720 Jan 08 '13 at 14:44
  • @user1911720: I'm not sure about a better site. Looking at your previous question, can you do `Import > Raw` to preserve the dynamic range of the original? – trashgod Jan 08 '13 at 18:53

1 Answers1

0

The are some things you need to be aware of when measuring intensity in imageJ:

  1. ImageJ automatically converts images to 8-bit
  2. You should ALWAYS use RAW if available
  3. If your microscope cannot save files as RAW format you must use .tif, other formats create artifacts
  4. Save channels separately not as an RGB stack if you are using .tif format

Before threshold you must split the channels, especially if you are looking for florescence from a certain wavelength.

You also need to use the "Watershed" function after the threshold so it will outline individual cells allowing you to avoid outlining a group of cells globed together. This way you can measure their individual intensities. However it is not perfect so once ROI pops up and the cells have been outlined by the particle analysis you must go through and ensure it is measuring singular cells. Any outlines containing more than one cell should be deleted. Also look for odd shaped cells or micro-nuclei, which should also be deleted.

Now I assume you are measuring human cancer cells. You should use these settings:

enter image description here

After all of this then you apply the overlay. Whatever image you are overlaying on you must also split the channels of that as well and use the channel that contains the wavelength you are measuring (ie: Alexaflour 488 would be in the GFP channel).

If you have been collecting data without doing this I would trash all of it as the procedure done to collect it hasn't controlled for anything.

Also as for where to go to ask this stuff.

FrankyG
  • 182
  • 7