How does ImageJ calculate intensity (mean gray value?)

intensity
Tags: #<Tag:0x00007fa304695128>

#1

I use ImageJ to calculate intensities of cells by using the multi-point tool, selecting a bunch of cells with a large green circle, and then control T then control M to measure them all at the same time. Recently I’ve noticed that the intensity values can be highly variable. I’ve done tests and clicked the same cell multiple times, and gotten very different intensity values. I’m hoping that someone knows how ImageJ calculates intensity so that I can adjust my protocol to make it more consistent. If anyone does something similar and has suggestions for how to get more consistent results, that would be greatly appreciated too. Thanks


#2

Good day!

First of all ImageJ cannot calculate intensities but gray-values only. Intensities are physical entities that can only indirectly represented in images.

Gray-values depend on the image format, i.e. if you have 24bit color, 8bit gray, 16bit gray or 32bit gray images, you have different ranges of values available.

The mean gray-value obviously depends on the image area you are measuring. Generally the area is defined by the number of contained pixels. If so, the mean gray value is the sum of the values at all pixels divided by the number of pixels.

However, the area may be given in scale units (Anlayze >> Set Scale…) which, of course, gives different results.

HTH

Herbie


#3

Hi Herbie,
Thanks for your response. Do you know how I can check the image format you mentioned (24bit color, 8bit gray)? And would you recommend importing my pictures in grayscale for the color mode? Currently I import on default.

I always use the same icon for the multipoint tool - extra large circle. That should mean that the area (# of pixels) should be the same every point right? Do you know if it uses all the pixels in that circle, or maybe it just uses exactly where I click?

Thanks again.


#4

Good day,

please study the ImageJ-user guide:
https://imagej.nih.gov/ij/docs/guide/index.html

And would you recommend importing my pictures in grayscale for the color mode?

It depends on what you want to do.
If color information is required for your analyses, you need to use the 24bit RGB-color image—no?

Do you know if it uses all the pixels in that circle, or maybe it just uses exactly where I click?

The latter but why not try yourself with a little test image?

Regards

Herbie


#5

Hi cbrusco1,
I think you are using the multipoint tool incorrectly:
The multipoint tool selects points and points do not have area. If you set Mark width you are not changing this , but you are telling the tool to just draw a circle of the corresponding width on your image and thus you are changing its grey value. That is why you are receiving different results on multiple clicks - you are selecting different yet close points and are changing their value (for refference :

Blockquote Mark Width If greater than zero, a mark of the specified diameter will be permanently drawn in the current foreground color (cf. Color Picker… [K]↓). Note that marks modify the image (it may be wise to work with a copy) and color marks are only available with RGB images (see 18: Embedding Color Annotations in Grayscale Images↓).https://imagej.nih.gov/ij/docs/guide/146-19.html#toc-Subsection-19.5

So in short you are not measuring areas but individual pixels and very likely you are modifying them doing so. Not to mention that in the multipoint tool all points are part of tge same selection so on your first click you have selected and measured one point, on the second two (and the mean is the mean of both), on third - three etc.
Instead of multipoint you should use a selection with constant size (Edit>Selection>Specify…) and reuse it for all your measurements: drag selection over area , then hit "T " to add it to RoI manager, drag over next area , hit “T” again etc. In this way you will have separate RoIs with the same area from your different cells and when you hit Measure all, you will receive measurements from each separate RoI.
Good luck!