How do you normalize images with different background fluorescence with ImageJ/Fiji?

I have been using imagej for quantification of my fluorescence signal and found that while thresholding my images, one tissue sample will capture all of my signal while another will capture my slim to none of my signal. I know what my signal looks like compared to my background. I have determined my threshold value, by thresholding a negative control so no signal is detected. Unfortunately, this excludes weaker signal and has, therefore, does not give me the best and most accurate results. I have tried to subtract background, but that doesn’t fix my actual problem of normalizing my pictures so I can compare them. Thank you for any help.

I have attached a few of my images and I’d really appreciate any suggestion.

WT%20TRITC%20GM|625x500

@nkashyap

I don’t quite understand … so bear with me. You are able to threshold your images fairly well - but are still ‘missing’ some lower signals? Have you tried any other pre-processing steps before thresholding to improve contrast? How exactly are you setting your thresholds?

Another important question is did you acquire all these images on the same system, on the same day, using the exact same acquisition parameters? That’s truly the only way you can compare a group of images… otherwise, you are manipulating datasets that are not truly comparable.

Hi there! I have tried pre-processing steps, such as subtract background, enhance contrast, and unmask sharp. I am assuming that there are pre-processing steps that I should do, but I was hoping someone had any other suggestions. I believe that the tissues are inherently giving off different background brightness that throws off my thresholding and quantification. For example, in order to capture all my signal for one tissue slide, I used 140, 65535 as my minimum threshold and maximum threshold, while I would have to use 180, 65535 or 100, 65535 for other tissue. Despite having a consistent protocol for dissection, embedding, sectioning, and immunohistochemistry, I was still having issues. However, I couldn’t do all my slides in one day as I have 100 slides that I had to do. I had to perform my experiments over the cross of several weeks, but during each experiment, I kept the time for each step consistent so I wouldn’t be manipulating datasets.

I uploaded three pictures that do not have a caption, but are from the same tissue and slide, with each image taken on the same day. As you can see, the tissue has different brightness level of background. I need a way to account for that.

Hi @nkashyap,

If I understand properly then you set the threshold manually, is that correct?
If so, you could alternatively try the “Auto Threshold” method in Fiji and use the “Try All” function on a couple of your images.
Maybe you can identify a threshold method that works well for all your images.

1 Like

Hi there, Thank you so much for your suggestion. I have been playing around with imageJ and using the autothreshold to see if I can find something that works for me, and I was kinda running into the same problem. I thought I would try to show you what I mean. I have highlighted in red filters that highlight as much signal as possible without background. I highlighted in yellow my negative controls that show little to nothing. I added an extra negative control here. The green boxes shows a common filter among all the pictures, which sounds promising; however, that filter still shows some signal in my negative controls. Some of which have more signal in my negative controls that images with actual signal.
DKO CCN3


DKO NEGCRL

NEGATIVE CONTROL - MDX

CCN3 - MDX - IMAGE 3

CCN3 - MDX - IMAGE 4

1 Like