European Molecular Biology Laboratory
Question
Asked 12th Dec, 2016
How can I analyze fluorescence intensity of epifluorescence images?
Hi,
Attached is a sample picture of Leishmania parasites that are stained with mitotracker. This dye gets accumulated into the mitochondria in a potential dependent manner. So, accumulation of the dye in the mitochondria is a measure of the mitochondrial membrane potential. Each parasite has a single mitochondrion which is pretty big in size compared to other eukaryotes. I have such images for wild type parasites as well as from different mutants and I am interested to see if there is any difference in fluorescence intensity.
Particularly, I would like to use ImageJ for this puropose. However, I am confused about some technical aspects. I would feel great if anybody could explain those to me!
So, my plan is to measure the fluorescence intensity from individual cells and then take the average of them. As the size of the mitochondria may vary cell to cell, how would I maintain the same area of interest across different cells? Is it necessary to maintain the same area or for each cell I will have to draw the line around it's own mitochondrion?
Most recent answer
I think all the points above are very good.
In practice, after taking images with same settings etc from all the samples, perhaps just substract from each image the background by calculating mean intensity in an area of the image where no cells exist. So first draw a box/rectangle (same size in all images, maybe always at the left or right corner of the image where no cells would be), measure the mean intenisty of that box and substract that mean value from the whole image (ie from each pixel).
Then from there background substracted images calculate both the mean and total mitotracker intensity per cell, either by using software to automatically segment the cells or by manually sketching out the cell boundaries based either on the mitotracker dye or - even better, but not neccessary based on the image you had - another dye which would dye the whole cells/ boundaries.
In imageJ you can set the measurements, so measure both mean and total (integrated) intensities, it is difficult to say which parameter would be better to analyse in the end. I think would be good idea to have a control (say a drug treated parasite which should for sure decrease/increase the mitotracker dye) for example FCCP or similar. This would also give some kind of indicatioin of the dynamic range of this method.
1 Recommendation
Popular answers (1)
University of Aberdeen
Dear Sumit,
I have few other suggestions:
1) I think that one key important thing is that the acquisition parameters are the same across the images. I know that is an obvious point, but I saw so many people measuring and comparing fluorescence from image acquired with completely different parameters.
2) Be sure you are not saturating the pixels on your CCD, otherwise you will lose power in your quantification. It does not seems the case in the sample image you posted, but I thought it was worth to mention.
finally for the quantification... I think that the answer on your question depends on what you would like to quantify.
1) The answer above are all valid, however, personally I would stay away from normalisation, Adam is correct in saying that there is variation between coverslips and samples, but I usually deal with this variability making (when possible) more sample for each condition and averaging them across.
2) another point to consider is if you are interested in the total amount of fluorescence, I mean a bigger mitocondrium with the same amount of marker will have a lower mean intensity... You can use the total (or integrated) fluorescence intensity but in this case you have to be careful and remove the background, (a bigger object have a bigger integrated background).
Hope this help,
good luck
Max
3 Recommendations
All Answers (4)
Northwest A & F University
Hi Sumit, in order to produce reliable quantitative results, you should automate the image analysis. In your case, this should be fairly straightforward. The two key steps are thresholding and object detection. Before that, you should optimize the dynamic range (Image>Adjust>Brightness/contrast). Be sure to use exactly the same max and min values for each image. For the first step is, thresholding: (Image>Adjust>Threshold). Note that color images generally have to split into their red, green and blue channels (Image>color>Split channels). Various additional commands, such as erode and dilate, are available to help the computer to recognize objects. In ImageJ, object detection and quantification can be performed via the “Analyze particles” command (Analyze>Analyze Particles).
2 Recommendations
University of Aberdeen
Dear Sumit,
I have few other suggestions:
1) I think that one key important thing is that the acquisition parameters are the same across the images. I know that is an obvious point, but I saw so many people measuring and comparing fluorescence from image acquired with completely different parameters.
2) Be sure you are not saturating the pixels on your CCD, otherwise you will lose power in your quantification. It does not seems the case in the sample image you posted, but I thought it was worth to mention.
finally for the quantification... I think that the answer on your question depends on what you would like to quantify.
1) The answer above are all valid, however, personally I would stay away from normalisation, Adam is correct in saying that there is variation between coverslips and samples, but I usually deal with this variability making (when possible) more sample for each condition and averaging them across.
2) another point to consider is if you are interested in the total amount of fluorescence, I mean a bigger mitocondrium with the same amount of marker will have a lower mean intensity... You can use the total (or integrated) fluorescence intensity but in this case you have to be careful and remove the background, (a bigger object have a bigger integrated background).
Hope this help,
good luck
Max
3 Recommendations
European Molecular Biology Laboratory
I think all the points above are very good.
In practice, after taking images with same settings etc from all the samples, perhaps just substract from each image the background by calculating mean intensity in an area of the image where no cells exist. So first draw a box/rectangle (same size in all images, maybe always at the left or right corner of the image where no cells would be), measure the mean intenisty of that box and substract that mean value from the whole image (ie from each pixel).
Then from there background substracted images calculate both the mean and total mitotracker intensity per cell, either by using software to automatically segment the cells or by manually sketching out the cell boundaries based either on the mitotracker dye or - even better, but not neccessary based on the image you had - another dye which would dye the whole cells/ boundaries.
In imageJ you can set the measurements, so measure both mean and total (integrated) intensities, it is difficult to say which parameter would be better to analyse in the end. I think would be good idea to have a control (say a drug treated parasite which should for sure decrease/increase the mitotracker dye) for example FCCP or similar. This would also give some kind of indicatioin of the dynamic range of this method.
1 Recommendation
