PSF of SPIM systems and deconvolution

Hello,
I was wondering what is your experience for deconvolution of ligh-sheet data.
in particular what is the best way to compute/estimate the PSF.

For a classic light-sheet with gaussian beam the light-sheet thickness changes and this also changes the PSF?
Also does someone of you know why the PSF in a light-sheet looks so different than an EPI? It looks V shaped and I guess the reason is separation of illumination and detection but is there a theoretical model for it.

Thanks

Antonio

what is your experience for deconvolution of ligh-sheet data.
in particular what is the best way to compute/estimate the PSF.

big topic, many opinions, here’s mine:

In my opinion, the “best” way to measure a PSF used for deconvolution is to (carefully) measure it directly with a bead. But, for the reasons explained below, it’s important that you do this using the exact same sheet (NA, wavelength, etc…) that you use to capture your raw data, and that the sheet be in the detection plane similarly between the PSF and the raw data.

For your other questions, the important thing to remember (and the “model” for it all) is that the effective “system PSF” in a light sheet (indeed, pretty much any microscope) is the product of the excitation PSF (in this case the sheet), and the detection PSF (your “widefield” PSF that you would see with epi illumination).

A good reference for this concept, if you’d like to read more, is Engelbrecht & Stelzer (2006)

Keeping that simple model in mind, the answer to the rest of the questions becomes clearer

For a classic light-sheet with gaussian beam the light-sheet thickness changes and this also changes the PSF?

Yep. The system PSF (the product of the detection and excitation PSFs) will change if you change either the sheet itself or the detection parameters.

does someone of you know why the PSF in a light-sheet looks so different than an EPI? It looks V shaped

That usually means your sheet is a bit out of focus. In that image above, I showed the “widefield PSF” and the light sheet profile both centered in the middle of the image… but that doesn’t happen for free, it requires alignment of the system, and if your light sheet is slightly outside of the detection plane, then the product of your sheet and your detection PSF will be skewed to one side. Here’s some read data I took to demonstrate that concept, shifting the sheet slightly out of focus. (note, the funny additional side-lobes are because this is a lattice not a gaussian sheet)

going back to the question of deconvolution, you can imagine now, that if you collect raw data with your sheet in a different state of “defocus” than the PSF you used for deconvolution, you’ll get artifacts. So, extra care care is required during both acquisition of the PSF and the raw data.

10 Likes

Dear @talleym
thanks for your insight. In principle I should observe something rather symmetric. I will try to realign the sheet and see how far I get. Could it also be that the Widefield PSF has some spherical aberrations for instance due to the GEL where the beads are? This could lead to a more asymmetric PSF.

Antonio

absolutely. if you have dramatic spherical aberration in the detection PSF, then the product of the detection PSF with even an in-focus light sheet profile may still show an asymmetric profile. It would be useful to look at the widefield PSF on its own. Depending on the degrees of freedom in your equipment, you can do this by:

  • fixing the light sheet position on the bead and taking a z-stack only moving the detection objective (if you can move the detection objective)
  • moving the light sheet in synchrony with the sample to keep the bead constantly illuminated as you move the sample through the focal plane (if your objectives are fixed and you can only move the sample)
  • if both objectives and the light sheet itself don’t have any motorized control, then you can illuminate the entire sample volume with widefield-like illumination (you’ll have a lot more background that way, so may want to prep a very sparse bead sample)
1 Like

Everything Talley said is correct however I would like to add a couple of points. The way the GEL/beads mix is prepared is also VERY VERY important. Bead concentration and even distribution is key but the gel you use must be perfect. Many people do not melt their agarose well enough and you end up with bad quality agarose. If you use agarose make sure the stock you use is for microscopy use only and not open everywhere with every kind of dust particles in it. Not every agarose are the same low melt is best and below 1% final concentration is optimal. Finally, check in which part of the volume you take your measurement as the deeper you look (detection axis) and further into the sample laterally (light sheet) the worst the condition get for PSF measurement.
This also depend on your set up (1 LS, 2LS, wavelenght???) as agarose absorbs in the UV range.
Happy to help with protocol if requires

1 Like

Practicing and trial and error with the agarose and beads regardless of application is time consuming. There are many protocols, which skip smalls details learned from making mistakes. As @Emmanuel_Reynaud mention you get one small thing wrong…it can become frustrating, especially when you are “almost” finished with a custom system trying to determine what’s. When you finally get it right the sample that gives you slightly better results…it doesn’t last very long and you have to make another. Some years ago, the local distributor for a company called Argolight let me borrow a test slide…

There are some limitations with their standard slides regarding water. However, I am wondering if anyone has tried their custom offering for Lightsheet/SPIM?

I didn’t had the chance to test it. We are currently trying to design 3d optical phantoms for LSFM systems using our 3D bioprinter technology and will hopefully send some around to test and improve

1 Like

Happy if you could share your protocol. There are some Lightsheet/SPIM builders around our lab. Had tried helping by giving them advice for sample prep. It was based on experience with 2P and confocal systems. A whole jar of agarose and a lot of beads later, not sure they have exactly the results they are hoping for…

The problem is we use a 3D bioprinter so not common tool in lab and it is a fluid phase one so may not work but happy to Zoom a sample prep class anytime

1 Like

Hi @talley,
this is what I measured once. To me it does not really look like a defocus but more like RI mismatch.
What do you think?

antonio

Have a look at this paper:
https://www.nature.com/articles/s41598-019-53875-y
it contains not only a theoretical model, but also a link to software which in principle allows to simulate PSF and do deconvolution.

Beads are great if you work with water, but I haven’t found a good method to experimentally characterize PSF in a solvent used for tissue clearing.

definitely looks like spherical aberration… whether there is also defocus of the sheet is a bit hard to say at that contrast level (and without knowing more about what you’d expect to see, in terms of how thick your sheet is relative to the axial resolution of your detection PSF)