I'm starting to use confocal microscopy and learning to perform image analysis, and when I started to study some of my images I found that in some channels (blue and red) I only have a few pixel intensities while in green these intensities are distributed throughout the histogram (as it should be, see attached image).
These intensity values are those that come directly from the microscope without any modification, so I understand that the problem comes directly from image acquisition.
Because my training in this field is scarce I ask for help to know the origin of the problem. Thank you very much in advance
Hi Alejandro,
I noticed your question looked quite interesting, and while I see you've already received some excellent suggestions, in particular by Paolo Pozzi, it's great to see when people are actively thinking about their microscopy data, so I thought I'd provide a bit more information for interests sake.
You're absolutely right to question those pixel values because a 3kx3k image with 5 pixel values? No way. But even in good working order HyD detectors don't produce a nice smooth histogram like PMTs. As Christopher mentioned it's to do with the design which I'll get to in a sec.
Firstly, those histograms. I've seen that a few times with HyDs/APDs (HyD detectors are a hybrid between APDs which are EXTREMELY sensitive but that made them equally as breakable and PMTs which have the lowest sensitivity but are nearly indestructible) and typically it occurs in one of two instances, either in the case of very low signal, but your staining and there's not a hint of cellular autofluorescence so we can safely rule that out. The other is when the detectors are nearing the end of their lives.
From those histograms that may be the case. A nice way to test this is to find a patch of cells that look kind of cool, play around with the laser power, scan speed, gain and averaging and take a really nice high resolution image using a PMT (gain ~700-800). It's better to disable smooth rendering to do this. It makes microscopes look better so there's often a smooth rendering feature enabled by default, Leica is no exception, right click on the image when you're in live or whatever and there should be the option to disable it.
HyDs are sold as being twice as good as PMTs (and you definitely pay a premium for it), so changing nothing except moving the HyD to where the PMT was, just adjusting the gain on the HyD (up to a max of 100), get a better image. If you can't do better with a more sensitive detector then by definition it's broken. You've also got some really nice images to send to whoever's responsible for the microscope to ask if they can explain/look into it.
Back onto their design though I've attached a couple of slides I had that have histograms of HyD detectors. I was literally the fist person to use them. That they are not smooth like the PMT's are isn't to say they're less accurate (When they're new). I've gone on for a bit so I'll sum it up in a nutshell. The way in which APDs and HyDs convert a photon into a measurable number of electrons so we can detect it is quite different to a PMT.
I see these are .lif files taken on a Leica confocal. Are the blue and red channels being collected on a HyD detector? I get similar results on HyD images. Something to do with the architecture of those detectors. Frame averaging seems to help give a more normal histogram. Or, switch to regular PMTs.
Hello Alejandro,
My Leica microscope does not have HyDs, so it could be a problem with those as Christipher said. Nonetheless, an alternative problem could be in the settings you used during acquisition.
At the moment you were probably working with 8 bit sensor bit-depth, which means the signal from your detectors would go from 0 for complete darkness, to 255 for a completely saturated detector. If in order to see an image you lowered the top scroller on the color bar that (at least for old Leica interfaces) is to the left of the image, you are simply "cutting" the top part of the range, but you are severely limiting the dynamic range of the detector. For example, if you move the top scroller to 1/10th of the total height of the color bar, your images will contain numbers between 0 and 25, and when you open them with Imagej, they will be scaled between 0 and 255, but in discrete steps of 10.
In order to avoid these problem, i would suggest to:
- In the settings of the microscope acquisition, switch to 12 or 16 bit mode. This will mean that the signal will be scaled between 0 and 4095, or 65535, which will give you much better data for image analysis and post processing, even if you tweak the color bar (the file size of the saved images will be much larger, though).
- Do not save your images as RGB, that would be automatically rescaled to 8 bit per color, and will in general reduce the information content for your image analysis. Save the raw data instead if you want to do any processing.
- When using the microscope, use the color bar as a last resort to make images brighter, instead prefer slowing down the scanning frequency, increase the voltage of the detector and average over multiple images if the signal is noisy, or slightly increase the excitation power the photobleaching is not terrible.
Dear Christopher, Paolo and Nicolas,
Thank you very much for your answers and for helping me with your experience. Indeed, after reviewing the configuration of the microscope in image acquisition, I had blue and red in hybrid detectors (until now I didn't pay much attention to that, but I see its importance).
As for the format of the images, I know that the ideal is to save them as composite instead of RGB, but when in the Leica program I export the images as TIFF (or when I import to ImageJ from Lif format) they appear in ImageJ as RGB images, I will have to investigate how to change it.
In conclusion thank you very much for your contributions, I have taken images again using only photomultipliers and thanks to that the reliability of my data has increased a lot.
Thank you very much and greetings!
Hi Alejandro,
I noticed your question looked quite interesting, and while I see you've already received some excellent suggestions, in particular by Paolo Pozzi, it's great to see when people are actively thinking about their microscopy data, so I thought I'd provide a bit more information for interests sake.
You're absolutely right to question those pixel values because a 3kx3k image with 5 pixel values? No way. But even in good working order HyD detectors don't produce a nice smooth histogram like PMTs. As Christopher mentioned it's to do with the design which I'll get to in a sec.
Firstly, those histograms. I've seen that a few times with HyDs/APDs (HyD detectors are a hybrid between APDs which are EXTREMELY sensitive but that made them equally as breakable and PMTs which have the lowest sensitivity but are nearly indestructible) and typically it occurs in one of two instances, either in the case of very low signal, but your staining and there's not a hint of cellular autofluorescence so we can safely rule that out. The other is when the detectors are nearing the end of their lives.
From those histograms that may be the case. A nice way to test this is to find a patch of cells that look kind of cool, play around with the laser power, scan speed, gain and averaging and take a really nice high resolution image using a PMT (gain ~700-800). It's better to disable smooth rendering to do this. It makes microscopes look better so there's often a smooth rendering feature enabled by default, Leica is no exception, right click on the image when you're in live or whatever and there should be the option to disable it.
HyDs are sold as being twice as good as PMTs (and you definitely pay a premium for it), so changing nothing except moving the HyD to where the PMT was, just adjusting the gain on the HyD (up to a max of 100), get a better image. If you can't do better with a more sensitive detector then by definition it's broken. You've also got some really nice images to send to whoever's responsible for the microscope to ask if they can explain/look into it.
Back onto their design though I've attached a couple of slides I had that have histograms of HyD detectors. I was literally the fist person to use them. That they are not smooth like the PMT's are isn't to say they're less accurate (When they're new). I've gone on for a bit so I'll sum it up in a nutshell. The way in which APDs and HyDs convert a photon into a measurable number of electrons so we can detect it is quite different to a PMT.
Dear\ Alejandro Febles
pls read this article :-
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4365987/
Article Fluorescence Microscopy
- http://zeiss-campus.magnet.fsu.edu/articles/basics/digitalimaging.html
Good luck
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training