FIJI (ImageJ): 3D Image Deconvolution [Theoretical PSF]
ฝัง
- เผยแพร่เมื่อ 3 ต.ค. 2024
- Learn how to use FIJI (ImageJ) to do deconvolution on confocal images acquired in 3D or as optical sections (z stacks). This tutorial uses the PSF Generator to model the imaging process of a microscope via a theoretical point spread function (PSF) and the DeconvolutionLab2 plugin to help restore the effective specimen representation (maximize observed resolution and signal) for 3D confocal microscopy images.
You can download the plugins from these sites:
PSF Generator: bigwww.epfl.ch/...
DeconvolutionLab2: bigwww.epfl.ch/...
Both plugins were developed at EPFL - Biomedical Imaging Group (BIG).
SUBSCRIBE to have first access to new video tutorials: / @johanna.m.dela-cruz
![FIJI (ImageJ): 2D Image Deconvolution [Theoretical PSF]](http://i.ytimg.com/vi/BwKbL4yD2-0/mqdefault.jpg)
Hi. Thank you so much for the video. I just want to ask , can this plugin be used for determining the PSF of light sheet microscopy? If yes, then are the parameter should be from the detection objective?
Hi @TNTT1234. Light Sheet image quality typically varies across the image because of the spatially variant thickness of the excitation sheet. A variable PSF is generally suggested for deconvolution. I haven't tried deconvolving lightsheet data using ImageJ, but you can always try. However, there are much better options for deconvolution, in particular Huygens light sheet deconvolution software, which incorporates a variable PSF that uses the specifications of your light sheet (microscopy parameters).
These are such great videos!!! Would it matter which model to use for PSF and/or Deconvolution Lab? Or are these defaults ones ok? Any insights on that? I have z-stacks of cells stained for Mitochondria.
Hi @Cris_Sci. Thanks for watching. The default models usually work, although you can always check the other models. They might give better results, depending on your images. DeconvolutionLab2 incorporates several algorithms of deconvolution with their own parameters. Tuning these parameters on your own (or you may be given automatic estimates) could be difficult and might also give you disappointing results.
The instruction in this video allowed me to greatly improve the quality of my image. For that, I want to say thank you! I do wonder, however, if there are any other plugins or methods you are aware of to speed this process up. I am working with 4 channels and large z stacks. Right now, it is taking me ~30 minutes per channel to deconvolute. Let me know what you think and thank you again for the fantastic instruction!
Hello. It would be additional work, but have you perhaps tried breaking up your z stacks into 2 or 3 groups? Or maybe you could scale down your images? You have bead images from which the psf is derived, I assume. I’m certain there are other ways to deconvolve images, although I haven’t explored these yet. One that I’ve heard about is CLIJ, which implements a GPU approach (instead of CPU).
@@johanna.m.dela-cruz I will try to break up the stacks to see if that helps! And, I will look into CLIJ. Thank you for your suggestions and response!
@@johanna.m.dela-cruz As a secondary question, I noticed you are working with a 16-bit image in this tutorial and have 16-bit selected in the PSF Generator window. If I were working with an 8-bit image, would I want to select 8-bit in the PSF generator window?
@@dsergio-xw3nw I don't think it matters that much...you can do 8-, 16- or 32-bit. The final display of the deconvolved image would still be 32-bit.
Hi thank you for the effort and passion that goes into creating these videos. I have followed the instructions of your tutorial, and my deconvoluted image is improved but when i use the deconvoluted image stack to generate a 3D script interactive animation my cells are somehow "compressed". I i use the image stack that i did not process with the theoretical PSF the cells look normal but i dont have the detail i am looking for. The Z-step values i entered are correct you know what could cause the compression ? It has to be an error with the Z-Step information?
I have found out, that when you save the deconvoluted image as a tif it will equalize your XYZ to 1 pixel. In my original data i had an XY resolution of 45 nm and a Z resolution of 186. That is the reason for the compression of my cells when i open the .tif deconvoluted stack. The question is now what should be the go to file option i should save my data in so my properties are maintained if its not tif ?
Hello @Carsten Mattes. You should change your calibration settings so that they are the same as your original image. Check the pixel and voxel size of your original image stack in Image>Properties. You can check Global settings so that even your deconvolved images will have the same spatial calibration. Save your files as .tif as well.
Would you advise against using 3D deconvolution if the z-stack was acquired with a widefield fluorescence microscope WITHOUT a motorized stage? I obtained a series of z-slices by controlling the objective position manually with the fine focus knob. The theoretical interval between two slices was 2 microns, which is not precise in my case. Is this issue serious for 3D deconvolution?
Hi Feline. Perhaps it would be best to first do stack focusing via Extended Depth of Field (EDF), then if stll needed, you can try 2D deconvolution on the output. Check out my tutorial on EDF here: th-cam.com/video/-B-p776s6VA/w-d-xo.html
Thanks for this video! I just had a quick question: when picking your dimensions XYZ in the PSF Generator, for Z you chose 25. What does this value represent? I thought it might correspond to something related with your z-stack size, but it is not super obvious to me. Probably me just not understanding something correctly. Any feedback would be greatly appreciated! Thank you
Hi @Juan Morales. The z can be any size. I don’t think it is necessary to make the width, height, and depth of the PSF image match the image stacks. As long as the PSF fits in the resulting (PSF) stack, then that’s fine.
Thanks for following my series.
Thank you so much for your great work and the video! It helps me a lot. I want to ask that do I need to preprocess the images, such as subtracting the background, before performing the PSF generator?
Hi @rong liu. The algorithms used in the generation of PSFs mostly take care of noise and filtering, so I don't think it is necessary to preprocess your images.
Thanks for your patronage.
Just wondering how would you do this for multiple 3D images using the same PSF?
Hi @Salim Ibrahim. Batch processing would be ideal for this. You would need to write a macro/script for the sequential deconvolution of all stack files in a folder.
@@johanna.m.dela-cruz Thanks I'm quite new to this type of way i.e. doing it via macro/scripting are there any resources out there or from your videos that will help?
@@salimibrahim459 You could ask for guidance from experts at the Image.sc forum (forum.image.sc/).
@@johanna.m.dela-cruz tahnk you
Hi Johanna, I am wondering what values are filled in the field size xyz? Are these the dimensions of the image in pixles and the number of slices?
Thanks and thanks for all your tutorial videos.
Hi @theophil777. Did you mean the size xyz in the PSF generator? You are right - this would be the dimension of the PSF image in pixels and the number of slices.
Thanks for subscribing. I appreciate it.
Hi Johanna, thank you for your answer. Best regards