Wiki‎ > ‎

Confocal and Widefield Image Processing

posted May 7, 2019, 7:37 PM by Danny Xu   [ updated May 7, 2019, 8:35 PM ]
Z-Stack Step Size


Microscope Calibration

Have you gone through the entire process of aligning the optical path? Are your phase plate and annulus properly matched? Is your Kohler illumination aligned? Are your samples sitting flat on the stage? Is your camera in the correct alignment in the optical port?
The microscope itself is professional grade. You should be able to take clear images in any mode so long as it is in working order.

Image Overlay

3D reconstruction of 2D confocal image stacks

Chimera software (http://www.cgl.ucsf.edu/chimera/) for 3D visualisation of confocal stack (after processing with ImageJ).

Amira for 3D visualization. It is a very expensive software. But they generously enable all the features for the trial version.
ImageJ (Fiji, just a new name for ImageJ) has many plugins for 3D visualization like 3D Viewer, Volume Viewer.


Bitplane Imaris vs. Amira
That depends on what kind of analysis do you need mostly. Amira is good at 3D/4D visualization, segmentation; you can easily generate workflow and apply to other dataset. Imaris has lots of isolated modules you can buy for particular applications, such as deconvolution, co-localization. Imaris can also connect to MATLAB, if my memory is correct. If you do not really need some particular functions/algorithms of Amira or Imaris, I suggest you start from Imagej or Fiji. It's free and has covered most common image processing and analysis algorithms, and highly flexible and extendable, if you are little background of programming.

Cell Counting with Imaris

Deconvolution

There are some for ImageJ, but it is suggested that the best bling algorithm is implemented in commercial software AutoQuant. The list of interesting blind and nonblind algorithms can be found here http://deconvolve.net/DNLinks.html
The one of blind methods I like most among mentioned is LASIP.
But why do you prefer blind deconvolution? Nonblind ones usually give better results and are less computational intesive. The best among commercial is Huygens Suite (week trial available in www.svi.nl), the best among freeware is COSMOS (http://cirl.memphis.edu/cosmos.php , new release will be ready in several weeks). For WF microscopy the mentioned ones give very high image quality improvement.



The point spread function (PSF) is the hardest to get. Without a good PSF deconvolution is meaningless.
I am skeptical of blind deconvolution in general (estimating the entire PSF image using an iterative process)


First create the PSF-file using your exact settings (mounting, oil, objective, NA,...) and use this for the deconvolution.

One generally deconvolves z-stack images. You can first acquire a z-stack of a stuck bead under the same magnification to generate the PSF. Open the image z-stack after this and then go to the 3D-Deconvolve function. Select the image stacks there: one sample stack and the other PSF genrated. Play with the parameters and you will get your deconvolved image from this

"Two plugins can be used in ImageJ to perform 3D deconvolution, DeconvolutionLab and Parallel Iterative Deconvolution.
They both need a stack containing the PSF, they are based on iterative algorithms, that is to say they first compute the estimated original signal. Then this estimated original signal is convolved with the PSF, so it should reproduce the deformation induced by the system and give the actual image. Different algorithms are then used to minimise this difference. A commonly used algorithm in fluorescence microscopy is Richardson-Lucy".


Free:  ImageJ/Image2, FIJI, ICY


co-localization plugins are available for the open source java application ImageJ (http://rsbweb.nih.gov/ij/) and its various distributions, including Fiji (http://pacific.mpi-cbg.de/wiki/index.php/Fiji). Two plugins that implement all the co-localization methods described above are JaCoP (ImageJ) and Coloc_2 (Fiji)."

EpiDEMIC plugin on icy (http://icy.bioimageanalysis.org/). It is a blind deconvolution (ie no need of a PSF) algorithm for epifluorescence

The implication that "ImageJ is not appropriate for deconvolution" is not correct. ImageJ/Icy and other open source programs can produce results as good or better as commercial deconvolution if used properly. In fact in the most recent Deconvolution Grand Challenge, one of Ferreol's algorithms beat all the commercial entries (the best commercial entry was actually the Olympus implementation by David Biggs). In a recent paper, several algorithms from Deconvolutionlab2 were tested against Huygens and produced comparable results

I encourage anyone interested in ImageJ and/or Icy deconvolution implementations to post questions on the respective listservs and forums, and share images. If you interact with the developers and community you will be able to obtain very good results.



Commercial:  Huygens (SVI, Inc), Imaris (Bitplane), Autoquant X (Media Cybernetics) and Volocity (Perkin Elmer).

Comments