Doubts on Airyscan

Hello,

I am a PhD student, and recently begun to use the confocal microscope (LSM710) with Airyscan, and I have some doubts/curiosities—hopefully, you can help me out:

  1. The first question is related to the detector: I notice in Zen that the files after the acquisition have below the z-position a slider for “phases” and there are 32-phases; what are they, what they represent from the actual image? I imagine that they would have data collected from the sample (therefore they represent the 32 honeycomb detectors), but if that is the case why can it be displayed (the image displayed is black)? This is intriguing to me because only when I have phase-1 it’s when I can see an image of my sample.

  2. Second and third questions are related to each other: what is the correct way to look at airyscan data to then analyze: acquired the raw airyscan super-resolution data and go with it (analyze in ImageJ or CellProfiller), or the file needs to always be pre-processed in ZenBlack to be analyzed? Using the latter I noticed that my images are reduced to ~4 mb from 130 mb, I am curious what kind of compression is done here that guarantees no data loss (or if its just the case that data is allocated but nothing is stored in the raw file?)

  3. a. Also, is there another way to process airyscan data besides Zen black?

  4. If I acquire images with Airyscan super-resolution and perform the “processing airyscan” can I compared intensities between my samples? When the acquisition settings are the same. Is this fundamentally wrong? Aren’t the detectors measuring photons in a comparable way to a the T-PMT detector?

  5. What is the guideline when using airy scan super-resolution and the thickness of focus is different between two channels: e.g. green is 0.42 µm and DAPI is 0.38 µm (arbitrary numbers for these example), is it ok to use different thickness or they should be matched?

A bit of context: in my experiment I am trying to count foci of a target protein in the nuclei of cells, using a pipeline in CellProfiller to segment and measure the objects in the images; but, in case that does not work well I was thinking if instead the intensities could be compared. But given the fact that the there’s a lot about the Airyscan detector and data processing that I don’t know, I am not sure if that is right to do.

Lastly, what do people mean when they say that airy scan super-resolution (quote) “is not confocal”?

Thank you very much for your time.

cheers,
leo

Hi,
You my want to ask all the acquisition-related questions in https://forum.microlist.org/

About your image analysis question, if you just want to count foci, in FIJI the tool find maxima ( Process > Find Maxima) might help you.

If you provide a sample image we might be able to better help you.
Good luck!

thanks for sharing that, I had no idea.

I have been exploring the pipeline of “speckle detection” demo for CellProfiller, and so far I think it’s not failing. The bad results I get from there are related to the sample, because mouse embryonic stem cells grow in colony and it’s hard to get a good image for segmentation.
But finding maxima means that the algorithm will look for intensities right?

I am afraid I can’t :frowning:

Thanks

Hi @lsilva.m,

you asked many questions at once but we sticked oir heads together to try to answer them all.

  1. When you open an unprocessed (raw) Airyscan dataset in Blue, FiJi etc, it will show 2 channels per track: chA is the raw data from the 32 detector elements, each on a different “phase”; chA# is the pre-processed image, and is only present on phase 1. Each single element, or channel, may appear as dark, but there is effectively information stored. This will only become truly visible after processing of the data.

  2. You should run the Airyscan processing before doing any further analysis. The processing combines the 32 phases into 1 image by a pixel re-assignment process as described in our White Paper Basic priciple AiryScan. This is why the file size is reduced by a factor of 32. By the way: if the raw data was 8bit, then the reduction would only be by a factor of 16, as the resulting image would still be a 16bit image.

  3. The processing can be done in ZEN Black or ZEN Blue (for standard Airyscan data sets). Some third-party software (Huygens, SimFCS) can handle the unprocessed data sets for some specific applications.

  4. The intensities can be compared between different images, but it is recommend to use the same Wiener filter (“Strength” in ZEN Black) for all images. You can run the processing in “Auto” mode for the image with the lowest SNR, then look in the info tab to find the strength used, and then use the same for all other images (for example using “Run Batch”). For multi-channel images, I’d suggest doing the processing in ZEN Blue, as it allows you to set the “Strength” for each channel individually.

  5. It is ok to use different optical section thicknesses as, generally, only for very precise 3D localisation matching the optical section thicknesses is sometimes required. Airyscan defines by default (this can be overrun) the internal zoom-optics as such as to provide highest resolution and sensitivity possible. The deviation is around or below 10%, which is within the resolution limits and will not affect your tracking accuracy.

Airyscan is a confocal detector, it just uses several elements each acting like individual pinholes (with varying degree of off-axis detection). So while Airyscan SR is still confocal, it is enormously more efficient than closing a confocal pinhole to 0.2AU to achieve super-resolution. By closing a physical pinhole to 0.2AU, one would lose a lot of signal and reduce the signal-to-noise drastically, which needs to be compensated for artificially.

Airyscan on the other hand creates 32 virtual pinholes that each ‘see’ the corresponding signal from a 0.2AU pinhole, but all together they collect the light as if you opened the pinhole to 1.25AU. So while each detector helps in providing the super-resolution image, all detector elements together provide a 4-8x increase in signal-to-noise.

4 Likes

I recommend this recent video posted on the confocal listserv (another excellent source of information).
*And I just noticed that the video was already posted in the cross-post!


The first few minutes might help with the information you are looking for. There is also a section on processing hte images.

It is geared towards the 880, but I found it more easily comprehensible than the white paper (though I am somewhat of a beginner, no fault of the whitepaper).

@sebi06 Note # 4 is interesting, had never heard that!
@lsilva.m In the video linked above, they recommend using Auto, but that is also more core facility oriented.

2 Likes

Hi @sebi06, great! thank you very much for you comprehensive answers.

So, regarding the raw file: once its processed it’s possible to revert? In other words, if you had to store a raw file from your data what would be the file the Airyscan raw, or the processed Airyscan? (I am asking this because I am thinking on hard-drive space management).

On point-4: (before I book the microscope again) to make it clear, when you say

image with the lowest SNR

do you mean a processed Airyscan picture that has more background than the other pictures collected?

Also, (I am curious) is it possible to achieve similar results with Wiener filter using Python? One more, is it possible to see in Zen all the iterative steps of the Wiener filter?

I will keep that! Thanks

also found another .pdf of some university somewhere (don’t recall) suggesting the same

I am guessing the Auto is the “easiest” answer, but maybe not the most correct one. I am learning quite a bit about Airyscan this week between the microscopy forum and the confocal listserv.

I am checking my files, and I can’t see in the info tab the “strength” that @sebi06 is referring to; but I used “Auto” and run in a batch all the files–maybe I missed a step here (or a click somewhere to see that detailed information)

If you open a Zen Black base image in Zen Blue and process it in Blue, it should show up at the bottom of the Info tab:
image

If you process the file in Zen BLACK, and then try and open it in Zen Blue, it does not show up. I do not know why, some sort of incompatibility. Seen a bit of that between Blue and Black with Airyscan stuff.

If you process in Black and open the file in Black, you should see it listed as:
image

ah ok, that is my problem then, because I am opening in Zen Blue 3.2 lite.

probably, but shouldn’t this be like metadata? Wait…

5 minutes later:
it is in the info when you open the processed file in ImageJ (but why not in Zen Blue?), look:


edit: I guess #1 and #2 refer to the two channels used

1 Like

Yes, it is definitely in the metadata since you can take it back to Zen Black to find the value after opening it in Zen Blue. Zen Blue just doesn’t read Zen Black’s entry. Might just be that they changed the name of the entry from Airyscan Mode to Airyscan parameter.

If FIJI is easier than Zen Black, that works as well.

it is not. I am not a regular user of Fiji (it is not intuitive to me), if I needed to analyze microscopy data regularly I would rather learn python to analyze images with code than learning Fiji.
The issue is that I don’t have a license of Zen Black to run in my computer (only the computer that controls the microscope has…)

Ah, I figured since Zen Black comes bundled with Blue in Zen Lite that would be the easiest solution.

Unless they have stripped Black out of Lite in recent releases? The Airyscan module itself might not be included, but you should be able to use Zen Black to read the files if you are curious about the AS processing strength.

I believe that is the case. I don’t see the option to install Zen Black (but I guessed that when I first downloaded it installed both; but I lost the installer’s .iso)