Image J difficulty in formatting for auto-analysis

Hello, I’m fairly new to ImageJ, and can not seem to find a threshold that will allow for me to auto-analyse the area of my cells accurately, just wondering if anyone has any advice? 1.pdf (1.4 MB)

Hi Aaron,

Would it be possible to a) post a tiff containing the image and b) in a second image, indicate where the cells are located? From the pdf it is hard to get the image data into ImageJ and there seem to be things floating in the ‘well’ that could be air bubbles, debris or cells… But then, I’m not a biologist.

Hi @eljonco ,

Thank you for getting back to me. I’ve attached a tiff and an image highlighting the relevant cells. yes there are a couple of air bubbles in the well, apologies, hopefully, the image clears it up, thank you 1.tiff (8.3 MB)

OK @Aaron1, that’s better. I notice that the magnification you use has a lot of overhead footage in the image. Maybe that is because you want the B3.B4,C3,C4 in the image. Unfortunately, this is not the best way to obtain a proper threshold but this image gives us the opportunity to mention a few options for improvement as generally speaking, image processing is ‘crap in, crap out’.

So first try to get the best of your microscope, camera and illumination. Your well doesn’t seem to give much colour information, so why not record in monochrome? That also saves 66% of data to store.

Next, if you look at the histogram of the well bottom, after converting your RGB to 8-bit grey, I see a peak from 190 to 220. That means your dynamic range is 220-190=30 grey levels, while you have 255 at your disposal. Try to manipulate your specimen, illumination, camera exposure time and microscope setting such that you obtain more grey levels from the well bottom.

The resolution of your image is quite low, despite the >8MB file size. Try to get more pixels per cell. Some cells you indicate have only a few pixels; leave out the unnecessary areas, focus (boom, boom) on the interesting part of the plate.

The image is currently scaled as 11.69x8.27 inches, which doesn’t seem quite right to me. Using Analyse>Set Scale, an appropriate line tool selection and a known distance, you can calibrate the pixel size more precisely.

Currently there is debris in your well while debris and cells have the same grey values. So make sure your medium and the plate bottom are free of any garbage. Polish!

Having said all that, on your sample image, do ‘Image>Type>8-bit’, select the bottom of the well using oval tool, do shift-D for Duplicate, then use that image for thresholding, using Image>adjust>Threshold, choosing the Default and have ‘Dark background’ unticked, should get most of your “cells” selected.

For me, the auto threshold selects 0 as lower and 193 as upper threshold limits. That gives the following image, which has most of your cells right. If for counting, this may suffice, if for size and (densitometry/intensity) quantitative analysis, this may be a starting point to inspect the (x,y) location of every result found using the Analyze>Analyze Particles tool.