Demo of automated cell or c-Fos counting with ImageJ / FIJI (2021)

[@janicetwixendorf6035]

You're a great help Kevin, thanks!

[@kruneuro]

No prob!

[@souheylabenbia4889]

Hello, thank you so much for this video, i need to quantify ovarian folicules (H E Staining slides), could this plugin is available?

[@kruneuro]

Hello! If you mean the general cell counting function that I'm using here, it should be available with default ImageJ. If not, you'll want to get the newer version: FIJI, a.k.a. ImageJ2. That said, there isn't a specific plugin for ovarian follicule counting to my knowledge. But, given that you're using H&E staining, you'll have to use color deconvoloation first to make the image not RGB, and thus able to be thresholded and then counted. I have other videos on my channel about how to do color deconvolution. Once you do separate out the color channel in which the follicules are most visible without other cell structures, you can use the threshold and then analyze particles functions. For the particle analysis in your case, I recommend checking "include holes" in your analysis if the follicules look more like rings than solid circles (I'm just guessing this might be the case).

[@souheylabenbia4889]

​@@kruneuro thank you so much for your response . i m a beginner in fiji so i will try that.

[@CRAPPYBETO]

Hello, thank you very much for the informative video! My question is since there is a forced conversion of the image to an 8-bit version of it, could this interfere with accuracy? I ask particularly because my cell images are taken at 10x magnification (16-bit, RGB) composited together in Photoshop (all while staying in a lossless format), and opened in ImageJ for cFos quantification (manually). Either way this has been extremely helpful to my lab, big thanks!

[@kruneuro]

Glad I could help. In regards to the 16 to 8 bit conversion, it seems like there *might* be loss of accuracy. Where this *might* happen is if your stained cells do not stand out from the background much (due to dim staining or other reasons), and 16 bit will allow for a wider margin than 8 bit when using the thresholding function. But, I don't expect that to actually matter too much if the cells are visually distinct from the background, and if their perimeter is in stark contrast to the surrounding background. Regarding that latter point: one exception might be if a cell has a gradient around it where the intensity goes from low to high. So, this would apply to cells or targets that have some sort of hazy exterior like a very loose extracellular matrix (maybe, for example). All things considered, I believe the potential color range loss that occurs from the 16 to 8 bit conversion will only be noticed in some very niche cases. Even if there's a visual difference, the thresholding function will treat both type of images rather equally.

[@shaileshvarade]

Hey, thanks for your video. It was wonderful to learn so much about ImageJ. I just had a query, how to measure the velocity of a rising bubble in any solution using Image J?

[@kruneuro]

Thanks! Your question is a bit tricky for me as I haven't used ImageJ for assessing active images (gifs) or video. In theory, one could take the frames of a video, threshold to isolate the bubble, and then mark its centerpoint (I believe this is done via Analyze -> Set Measurements -> Center of Mass option, then being sure to request summary of results when analyzing particles). Doing this over a successive series of images will show a progression of the centerpoint's coordinates, and the time can be determined based on video framerate and the length of time recorded. However, that approach seems awfully manual, so perhaps there are plugins that do this more automatically. I see in the Plugins menu that there are Tracking and Time Lapse options in the current version of ImageJ/Fiji, and I suggest seeing if there's documentation or video of that somewhere on the web. Sorry I couldn't give a direct answer.

[@susobhandas4480]

thanks a lot for the video!

[@milenarodriguezalvarez465]

Kevin, this is fantastic, thank you so much for a such detailed video. I am doing exactly the same using a different neuronal activator marker, I also have an experimental and control group. However sometimes the pictures are not exactly in the same position. Do you move the square around to find the area of interest analyzing a new image? Do you use the same threshold for every image as well? I have noted that the optimal threshold can be different for different pictures, any tips to get around that? Do use filters? Do you remove background? Thank you so much !!!!

[@milenarodriguezalvarez465]

Also, have you batched the counting process?

[@kruneuro]

​@@milenarodriguezalvarez465 You're welcome! To your questions: 1. I do move the box around, as images are frequently not centered over the exact area of interest. So, the x and y coordinates of the box will differ between images, but that isn't really an issue. But I maintain the width and height dimensions across images for the same brain region. Example: medial accumbens shell always uses the same box size, but lateral septum images might use a larger box size. In my 2020 disgust publication, we decided on box sizes for each region that generally encompass as much of that specific brain area without containing much if any of surrounding areas. 2. I try to keep the threshold similar across images, but it will likely vary to some degree between batches of staining. Background fluorescence might come from tissue overfixation or something weird happening with the staining process, so in some images the signal to noise ratio will be poorer and require different threshold values. I don't think this is an issue either, unless the thresholding shrinks the stained particles/nuclei/cells so much that the counter starts skipping them. So, be sure to document the threshold values you decide on for each image in case things need to be re-counted for whatever reason. 3. I try to get around the differing threshold issue, as well as any adjustments requiring dropping out the background or increasing global brightness of the brightest elements (hopefully the correctly stained stuff), by amping up the staining intensity through tyramide signal amplification. Maybe I should make a brief video on that... But ask me about that if you have questions. Essentially though, staining methods that improve signal to noise ratio circumvent the need to do any image processing beyond simple thresholding, and they also put me at ease knowing that I'm not messing with the original picture much. So, no more color curve shifting or playing with contrast settings, except for making images *slightly* better for publication. And, better staining (and maintaining decent quality of brain tissue) should prevent threshold values from differing much between images, even though they will always differ a little bit when using an epifluorescence microscope setup. 4. Not sure which filters you mean here. In ImageJ, or other software? I've never used them. 5. On that last question, I am unsure what batching the counting process means. Something like binning?

[@bhawnapandey4375]

Hi Kevin, I am trying to count bubbles in my image. So, can the process of counting particles be applicable to bright field images, i.e., not labelled?

[@kruneuro]

Hello - yes, it should still be applicable. Staining isn't required as long as your target (the bubbles) contrast with the background. Even if the image is in color, I think the regular thresholding will convert it to grayscale first. If it doesn't do that, change the image type to 8-bit first. Then, you would need to check on the "include holes" option before analyzing particles, and you should avoid doing the process -> binary -> watershed function as it will fragment the bubbles. Hopefully this is helpful.

[@ayahhamdan1649]

Hi! Thanks for this video. Do you have any tutorials on how to use imageJ to count co-localization of c-fos cells and another red-reporting cell?

[@kruneuro]

Hello there! I do not yet have my own protocol for colocalization, even though it is a measurement that is of much interest to many. From what I recall reading about it, it is definitely a bit more complicated, involving color thresholding. A brief search & viewing makes me think this help get you part of the way there: kzbin.info/www/bejne/qIPce6F8pJpqrrc . Unfortunately it seems the video focuses on percent overlap and correlations rather than discrete counts. For such counts, I think the thresholding needs to eliminate other single-color targets from the image, and then counts should be able to be done normally via the Analyze Particles tool from there (as the image should be binary black & white, then). Other videos seem to exist on the subject but may not give straightforward & quick answers for ImageJ... so that's a video I should do at some point. One alternative is to do regular thresholding and analyze particles on the individual green and red channels (after they are grayscale'd) to figure out how many red and green cells there are, then also taking note (in the same results) how much area each occupies. Then on the multi-color image, you can do this color thresholding to isolate the yellow, and then take note of the area measurement of that. You can then divide the "yellow area" number by the "red area" number to get a fraction, and multiply that fraction by the total number of red cells. That would roughly give you "the number of red cells that are also yellow". The same procedure could be done for green.

[@zahraahmadi2234]

hello. I have a question.would you please help me?I want to count the number of grafts inside the plates with the Fiji (filament detector). how can i do that(can i have your email address i want to send photos of analysis)

[@kruneuro]

Hi there, I am unsure how well I can assist you (I've never used filament detector), but I can try. We should converse over email, with some picture examples you can provide. I don't post my email in the youtube comments as I don't want to be harassed by bots, but you can find it via the website link in this video's description above.

[@zahraahmadi2234]

@@kruneuro thank you so much