Fluorescence Detection in Retinal Image Help


With reference to the image below: GFAP_Iba1-E84L1-20x-2 2 - Copy.tif (3.4 MB)

Could you please describe an optimal way to measure the fluorescence (mean grey area and SD)?
I.e: in the ROI circled below, what would be the best steps to quantify the fluorescence?

Currently the following steps are being undertaken: 1) subtract BG ; 2)convert to 8 bit 3)measure mean and SD 4)set threshold based on mean and SD. The mean values are perceived to be too low
GFAP_Iba1-E84L1-20x-2 2 - Copy - Copy.tif (3.8 MB)


I don’t understand the workflow sketch you provided. Is that to generate a segmentation?
Because an intensity measurement should be performed on data without image processing.
What do you mean by the mean values are perceived to be too low?

If the measurement does not reflect the average intensity in the area you want to measure than this could be a problem with the segmentation. If it is not accurate enough and you segment also background than it could underestimate the actual intensity.

1 Like

Thank you for your reply. Apologies for the vagueness in the initial question.

The objective is to measure the fluorescence/bright spots in the encircled region. Any method is welcome.

Yes, the process described is for segmentation and the issue is with efficiently removing background during segmentation without compromising the region of interest too much.

In what structures are you then interested in? Only in the thin part within the outline?

A manual segmentation can also give good and quick results.
I would suggest using a segmented line:
Double click on the line ROI in the Fiji toolbar.
Select the segmented line.

Then you can select a line width to expand the area of measurement:
Right click again on the line ROI.
Select a width of the ROI.
Select the spline fit to smooth out the segmented line ROI.


For documentation and curation you can add this ROI to the ROI manager and save it.

1 Like

I’m interested in the thin part, as outlined by the segmented line (which was very helpful, thank you!).

How could the fluorescent part of this ROI be quantified most accurately (mean and SD)?

Yes, mean and standard deviation. Maybe integrated density could be also a parameter worthwhile looking into.

You can also have a look at median as well as the min max values. To make it complete. I always recommend to measure and initiatlly plot as many parameters as one can get to get a feeling for the data. Then one can always settle for the most appropriate parameter later.