Hi,

I’m not at all sure how relevant this will be without seeing an example image, but the Process -> Find Maxima… command can be very good for counting within in dense regions. The main thing you need is for each cell to have a distinct maximum (a pixel of cluster of pixels all surrounded by lower values). To improve the chances of success, you can adjust the ‘Noise tolerance’ value with ‘Preview point selection’ clicked.

If you set the output type to be ‘Point Selection’, you can then manually refine the result by adding / removing points:

http://rsbweb.nih.gov/ij/docs/guide/146-19.html#sec:Point-Tool

If you draw a ROI first, Find Maxima… will just find points within that. Press M when you are done to measure all the points, and from the Results Table that appears you can read off how many points there were.

To be practical in your case, you will probably need a few preprocessing steps:

- Convert to grayscale; either run ‘Image -> Type -> 8-bit’ or ‘Image -> Color -> Split Channels’ and choose the channel that seems to have the best contrast.

- Edit -> Invert (needed to apply Find Maxima… useful to brightfield images rather than fluorescence)

- If you get too many maxima, try running a small Gaussian filter first with Process -> Filters -> Gaussian Blur…

Anyway, I hope something in there helps - good luck!

Hi,

I’m not at all sure how relevant this will be without seeing an example image, but the Process -> Find Maxima… command can be very good for counting within in dense regions. The main thing you need is for each cell to have a distinct maximum (a pixel of cluster of pixels all surrounded by lower values). To improve the chances of success, you can adjust the ‘Noise tolerance’ value with ‘Preview point selection’ clicked.

If you set the output type to be ‘Point Selection’, you can then manually refine the result by adding / removing points:

http://rsbweb.nih.gov/ij/docs/guide/146-19.html#sec:Point-Tool

If you draw a ROI first, Find Maxima… will just find points within that. Press M when you are done to measure all the points, and from the Results Table that appears you can read off how many points there were.

To be practical in your case, you will probably need a few preprocessing steps:

- Convert to grayscale; either run ‘Image -> Type -> 8-bit’ or ‘Image -> Color -> Split Channels’ and choose the channel that seems to have the best contrast.

- Edit -> Invert (needed to apply Find Maxima… useful to brightfield images rather than fluorescence)

- If you get too many maxima, try running a small Gaussian filter first with Process -> Filters -> Gaussian Blur…

Anyway, I hope something in there helps - good luck!

Hey,

See if this helps:

Open your image on Fiji and subtract background and apply a Gaussian blur of radius 1.

Then convert your image into a binary image (process->binary->make binary). Be careful about how you threshold your image prior to making your image binary!!

Then go to analyze and hit analyze particles. Retain the default settings of 0-infinity size. make sure to click the summarize button and then analyze your image. This will essentially give you the total area of the pixels on your image. Now again return to your binary image and using the wand tracing tool, click on any given cell and measure the size of that cell (Analyze->measure). If you do this over say 50-100 cells, you will get back the average area of a cell and then you can divide your total area by the average area of a cell to find out the total number of cells. Though this aint perfect, I found this method gives back values very close to a manual count, atleast in the cortex!!

Hope this helps!!

I came across the same problem in my masters. I had used DAPI to stain nuclei in neuronal cryosections, and then batch analyzed hundreds of images using a free program called CellProfiler. It took me ages of internet searching to find this program! Nothing I found in ImageJ seemed to do the trick quite as well, at least at the time.

It has an amazing ability to detect and count overlapping cells in dense sections. We also used it to look at the overall area of the tissue penetrated by blood vessels, and then compared cell density (or count) to the areas of high blood vessel density. It worked great.

It will not only count the cells but also output other metrics such as area, diameter, roundness which may be useful for your study. I loved it the most for it's ability to batch process, so you can select your folder with all your images to analyze (it can handle hundreds at a time) and then just leave it to run. Hope this helps!

http://www.cellprofiler.org/

Thank you for all your suggestions.  Thanks especially for the cellprofiler link Leslie.

You are very welcome! I'm glad you found it useful. There are so many free open-source programs out there to make these labour-intensive processes easier, they are just really hard to find as they don't advertise.

Hi Leslie. As you suggested I am also trying to count Nissl stained neurons in a large cortical section after imaging. It seems I cannot find anymore the pipelines online. Is there another way to find them ? I am interested in the: Cell/particle counting, and scoring the percentage of stained objects. 

Hi Roberta,

When I used cell profiler I did not use any pre-made pipelines. I created my own based on my particular needs. However I was able to find a link to these pipeline examples that you can download and test out.

Good luck!

http://cellprofiler.org/examples.html