I have a set of binary images (black and white) of cells containing mitochondria, I need to count the number of mitochondria in these cells. What function & version of Imagej or Fiji do I use? is there any paper that covered similar work? Please point me to the right resources.
1) how are you defining "a mitochondrion"? Usually they're a network, or at least reticular to some extent, rather than neat little individual countable units.
2) do you have any other image data (DAPI channel, cell surface marker etc) that could help refine your analysis? As far as I can see, even if we assume mitochondria are strictly countable (see 1, above), you're going to have a hard time excluding nuclear spaces, and an even harder time determining which cell which mitochondrial signal belongs to. Especially with cells this confluent.
3) have you considered other methods? Like western blotting for mitochondrial proteins, or qPCR for mtDNA?
Hello Dear Hasan
I am a material science engineer.
Can you describe to me which particles are mitochondria :)?
Then I can help you.
Regards
While it is certainly not the only option, I like to use ImageJ (or Fuji). As such, my suggestions below are specifically related to this program. I find that the open source nature of the software means that I have the flexibility to do what I want and that there are a lot of plugins, macros, codes, and message boards (https://list.nih.gov/cgi-bin/wa.exe?A0=IMAGEJ) out there to rely on.
On the surface, your question seems straightforward, but your image actually has quite a lot going on. This may be difficult for some image processing software to handle. You will need to be careful in defining your conditions and the complexity of your image may require a fair amount of trial-and-error. While your question uses the word “binary,” your image is actually greyscale meaning that it has pixel values ranging between 0-255. Therefore, your trial-and-error will likely involve a finding a series of steps to make your image binary (pixel values of only 0 and 255, which are black and white, respectively) in such a way that you can measure/count whole, individual objects of interest (i.e., mitochondria). This can involve steps such as thresholding your image (e.g.,: Auto Local Threshold using Mean with a radius of 50 and a Parameter 1 of 1) and then further cleaning it up to remove speckles and define the edges (e.g., Process then Noise and Remove Outliers followed by options in Process’s Binary menu). Note that these “e.g.,” actions are suggestions, and your actual workflow may end up quite different. Whichever steps you choose, once you have a satisfactory image that shows the structures as whole, individual units, the Analyze Particles function will allow you to define objects you wish it to measure using size and circularity. The mitochondria in your image look like they are relatively uniform in these aspects. If you have many images to evaluate, then the time put into this trial-and-error can be very worthwhile. However, if you only have a few images, and aren’t familiar with programs such as ImageJ, then it may be easier and faster to just count them manually.
Should you decide to go the image processing route, you should also look into plugins. ImageJ has plugins on their website (https://imagej.nih.gov/ij/plugins/index.html) and other users will post theirs as well (a simple internet search will find you many of them). For example, the Cell Profiler plugin is powerful but may be limited by the complexity of your image. Another option is to create your own macro. ImageJ’s macro recorder is a wonderful tool that records your steps (i.e., clicks) into code that can be saved as a macro or made into a plugin that can be used to batch process many images. These steps are figured out ahead of time by the user (see above about trial-and-error). For this to work, the steps that you take must be able to make good decisions about all of your images automatically. There are certain checks that you can build into it (like saving certain outputs), but ultimately you’d want something that can work consistently and automatically. A big benefit is that it can measure your images in a more unbiased manner, but a drawback is that it can miss or overcount on some images.
Is this a cell culture that you are imaging or is it a cross section of a tissue? If the latter, you should also take into account cross-sections are like sub-samples of a 3D structure. In this case, taking advantage of stereological methods could be advantageous to your study if you don’t want to go down molecular routes such as finding mitochondrial copy number. It is certainly not a new concept (e.g., Loud 1968 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2107397/), and some will argue its applicability and practicality, but it may be worth looking into for you especially as the demand for quantitative over qualitative data increases. Of course, the accuracy of any particular stereological method will depend on its prerequisites and assumptions (see Mayhew & Lucocq 2015 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4386931/). There are stereology plugins available for ImageJ that may or may not be able to handle your complex image, just have to try them and see. So again, you may have to resort to doing some counting manually (see Ferguson et al. 2017 https://doi.org/10.1007/s00418-017-1564-6).
I know that this post is rather long (I tend to do that), and that I’ve given several options. I regret that I can’t give you a simple do-this-then-do-that kind of answer. I hope that someone may come along that can (I’d be interested in hearing it too!). I hope that it helps at least a little bit though.
Good luck!
- Melissa
- Badges
- Science method
- Similar topics
- Methodology
- Crystallography
- X-ray Diffraction