Compare Images taken with a different zoom

Hello, I’m new to ImageJ/ Fiji and have a small question regarding the analysis of two different groups of images, that were accidentally taken with a different zoom applied to the microscope (zoom in function).

Group 1: No zoom
Group 2: Zoom differs between the value of 2 and 4 (I have the exact value for each neuron).

I’m currently analyzing the raw integrated density and I am comparing their areas. The Problem is, that is that due to the applied zoom, the Area of group 2 is much bigger than the Area of group 1, messing up the results. Taking this in consideration, is there any way to compare the two groups with each other (without losing information of the pixels,etc.)?

If anyone could help, I’d really really appreciate it!

Information images

Probably, sort of, assuming all else was equal and the flatness of field is good. But it is not normally good practice unless you were careful to keep things like the scan speed, the number of pixels, etc constant.

If you know the pixel sizes, you should be able to do everything in square microns. In FIJI, this can be set in the Image->Properties… tab. Use “um” for microns. That is where you can put the correct “microns per pixel” value that microscopy images should come with (or be known when taking the image).

Or convert areas in pixels to areas in microns at the end of whatever measurement you are doing. Multiply out the pixel size (or relative pixel size) at the end of collecting the data.

Or, take the image with the smaller pixel size (zoomed in), and increase the pixel count proportionally. Image->Adjust->Size will allow you to go from 1024x1024 to 2048x2048. I recommend no interpolation, but to a certain extent, you can’t completely recover or compare the images perfectly.

Another rougher approximation might be to multiply the integrated density by the zoom so that they match. It is essentially the same assumption as above, that each pixel in the zoomed out image is 4 or 16 pixels in the zoomed in images. The value of each of the four or sixteen pixels that the single pixel turns into will be the same, so multiplication will give a similar result. It may not exactly be the same result depending on how your are thresholding/generating your ROIs.

Anything you do at this point will be something of an approximation, if nothing else, due to how rough the ROI is that you will have to create on the zoomed out images (it will be blocky). Thin dendrites on the zoomed in images will be completely absent without the 4x zoom, etc. You cannot fix the experiment in post :stuck_out_tongue:
Hopefully those are not how the images look raw, as much of the cell looks washed out and over the threshold, which would make the integrated density meaningless.

Thank you for the detailed answer! It was really helpful :slight_smile:

1 Like