How to generate PSF of iSIM using PSF generator

Hello,
I have troubles to generate a theorical PSF for instant SIM (iSIM) using PSF generator.
For example, the iSIM mentioned in the paper ’ Guo, M. et al. Rapid image deconvolution and multiview fusion for optical microscopy. Nat Biotechnol, 38, 1337–1346 (2020).', as follows:

iSIM imaging. The iSIM system has been previously described13. For all experiments, a ×60, NA = 1.42 oil-immersion objective (Olympus PlanApo N 60× Oil) was used, resulting in an image pixel size of 55.5 nm and a lateral resolution of ~150 nm. Fluorescence data were acquired with a pco.edge 4.2 sCMOS camera, and the exposure time was set to 40 ms per image frame. The imaging axial step for beads, immunolabeled mitochondrial samples and transfected endoplasmic reticulum (ER) samples was set to 100 nm, 100 nm and 500 nm, respectively.

I multiplied the excitation PSF by the emission PSF generated by PSF generator, but did not get the resolution mentioned in the paper.

Thank you very much for your input!

Update 2024/1/24:
I followed the steps in the paper: ‘Zhovmer, A. & Combs, C. A. A Step-by-Step Guide to Instant Structured Illumination Microscopy (iSIM). in Confocal Microscopy (eds. Brzostowski, J. & Sohn, H.) vol. 2304 347–359 (Springer US, 2021).’ to generate the theorical PSF of iSIM.

1 generate the excitation PSF using the following parameters:


The measured FWHM is about 201.67 nm in the lateral direction.

2 generate the emission PSF by only modified the wavelength to 515 nm.
The measured FWHM is about 204.80 nm in the lateral direction.

3 generate the idealized PSF by multiplying the excitation PSF and by the emission PSF.
The measured FWHM is about 147.96 nm in lateral direction.

Now the resolution of the generated PSF approximates the resolution mentioned in Guo’s paper (i.e., ~150 nm).

I want to get a same lateral resolution as the iSIM system in Guo’s paper.

I am a beginner of microscopy. If I’d done anything wrong, please let me know.

Hi @K.K and welcome to the forum. Can you elaborate briefly on exactly how you generated your PSFs (e.g. specify the parameters you’re using), and clarify what you mean when you say “did not get the resolution mentioned in the paper”? Exactly how were you calculating resolution, what did you get and what were you expecting?

Hello @talley,
Thanks for your reply.
I have made some update, please take a look.

Thanks for the added details, helpful!
Now, I’m still a bit confused about what you’re trying to achieve. You’ve said that the paper reported 150nm, and when you take the product of em x ex PSF, you get 147.96.

Are you wondering why the paper states 150nm and why you got 147.96? Or am I missing your point?

First, I’m not exactly sure that this is the correct way to generate an iSIM PSF.
Second, as you say, I am wondering why the paper states 150nm and why I got 147.96 nm, and what makes the difference.

Second question first, since it’s easier:

I am wondering why the paper states 150nm and why I got 147.96 nm, and what makes the difference.

A) as far as a can tell, the paper simply says “a lateral resolution of ~150 nm”, and i don’t see where they state their wavelength assumptions, and 148 is absolutely “about 150n” :slight_smile:
B) That number will change based on the wavelengths you used. So if you bumped your emission slightly from 515, you’d get to 150 on the nose.

long story short, it doesn’t matter, 148 is effectively the same thing.

I’m not exactly sure that this is the correct way to generate an iSIM PSF

Yes, it’s an appropriate estimate.
All Image Scanning Microscopy (ISM) techniques (like iSIM, AiryScan, SoRa, etc…) can be modeled (prior to any deconvolution) roughly as the product of the excitation PSF and the detection PSF. In other words: roughly similar to what you would get in confocal microscopy if you had the pinhole closed all the way down (but without all the signal loss).

If you want a deep understanding of why that is, it’s a bit harder to know at which level to begin discussing it, depending on the level of understanding you’re looking for.

If you’re very interested, I recommend the supplement of Li et al 2015 (though, fair warning: it’s not for the faint of heart :joy:, it’s very dense, but it’s all there)

https://www.science.org/doi/suppl/10.1126/science.aab3500/suppl_file/li-sm.pdf

The key equation to grasp there is 71 (which they derive over a number of steps):

which can be paraphrased as “The effective ISM image (D) is the Fluorophore distribution (S) convolved by the overall PSF (H), summed over all detector pixels”, where (equation 69):;

“The overall PSF (H) is roughly equal to the product of the excitation PSF and the detection PSF centered at pixel image”.

And note that the detection PSF itself is influenced by the pinhole size (eq 67):

image

i.e. The detection psf is the convolution of the ideal detection PSF with the pinhole function. Note here, that as the pinhole closes all the way down to “nothing” (in technical terms, as it becomes a delta function), then the equation above reduces down simply to the detection PSF (because convolution with a delta function does nothing).

So, this is why it’s fair to say that the effective PSF of iSIM (or any other ISM with a very small pinhole pixel element) is roughly equivalent to the product of the excitation and emission PSFs.

1 Like

Thank you so much for your assistance!
As you said, “148 is effectively the same thing”, it may be enough to use current PSF to do the following processing. I will follow the current method to generate the PSF and dive into the provided supplement for mor information.

1 Like