http://confocal-microscopy-list.275.s1.nabble.com/Deconvolution-of-3D-SIM-data-tp6251420p6268624.html
I'm not going to give an answer, rather to raise some theoretical aspect of deconvolution and SIM and give my non scientific opinion.
As far as I know, in noise free images, making deconvolution would be as easy as making a division of the image with the PSF in the frequency domain, as convolution in frequency domain is a simple multiplication of the real image with the characterization of the objective...PSF. Unfortunately, due to noise, this is not straightforward but the goal should be the same.
In SIM, images from the frequency domain containing the Moiré patterns are put back together using the interference of typical 9 different angles/phase shifts, from which high frequency information is retrieved. Therefore, the "behavior" of each dot as a PSF for classical deconvolution makes no sense to me (eventually with 9 PSFs...and deconvolve each before reconstruction...). Furthermore, SIM only works for features which are in the interference range, i.e., resolution limit, boosting on those cases the MTF (Modulation transfer function), having now a new local maximum and change the supposed linearity of deconv. Also, as it was already posted here, it is possible to use non linear Fourier reconstruction to go a little further.
Therefore, I would be very, very skeptical either on any deconvolution and quantification using SIM.
Tel . +351 4464538
Fax. +351 4407970
> *****
> To join, leave or search the confocal microscopy listserv, go to:
>
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy> *****
>
> To elaborate a bit on that, why isn't it possible to go beyond the
> diffraction limit with deconvolution ? I don't see the theoretical reason,
> in a ideal case (subresolutive beads sufficiently spaced). Of course in a
> real sample, S/N issues and fluorophore density impose a limit on the
> precision, but why would this limit precisely be at the diffraction limit ?
>
> Christophe
>
> On Wed, Apr 13, 2011 at 09:44, Andreas Bruckbauer <
[hidden email]> wrote:
>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>>
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy>> *****
>>
>>
>> Hi Dan, Brian
>>
>> thanks for the comments, i know that for real live samples everything what
>> exceeds the resolution limit of the system is an artefact, however for
>> sparsely separated beads i would expect that i can determine their position
>> with high accuracy only limited by the signal to noise ratio. This is
>> exactly what is done in localisation microscopy (PALM, STORM...). One
>> approach is to calculate the cross correlation between the measured 3D PSF
>> and the 3D image (see e.g. A. G. York et al. Nature Methods 8, 2011, 327).
>> While this is not exactly what deconvolution does, it is at least related to
>> deconvolution.
>>
>> So i was wondering if this special case, a limited number of well resolved
>> beads which are small enough to qualify as point objects, could tell us how
>> good the deconvolution method actually works, or in the case of SIM, what
>> the image restauration actually does. At least theoretically, the size of
>> the beads (110 nm in my case) should not be a problem as by using the same
>> beads to measure the PSF we trick the deconvolution into thinking that they
>> are point objects. This is similar to the obove mentioned localisation
>> microscopy which does not depend much on optical resolution or pixel size.
>>
>> best wishes
>>
>> Andreas
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> -----Original Message-----
>> From: Daniel James White <
[hidden email]>
>> To:
[hidden email]
>> Sent: Mon, 11 Apr 2011 10:53
>> Subject: Re: Deconvolution of 3D SIM data
>>
>>
>> *****
>>
>>
>>
>>
>>
>> To join, leave or search the confocal microscopy listserv, go to:
>>
>>
>>
>>
>>
>>
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy>>
>>
>>
>>
>>
>> *****
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> Hi Andreas,
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> On Apr 9, 2011, at 7:02 AM, CONFOCALMICROSCOPY automatic digest system
>> wrote:
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>> Date: Fri, 8 Apr 2011 11:48:46 -0400
>>
>>
>>
>>
>>
>>> From: Andreas Bruckbauer <
[hidden email]>
>>
>>
>>
>>
>>
>>> Subject: Re: Deconvolution of 3D SIM data
>>
>>
>>
>>
>>
>>>
>>
>>
>>
>>
>>
>>>
>>
>>
>>
>>
>>
>>> When i generate a PSF for deconvolution using suitable beads, then image=
>>
>>
>>
>>
>>
>>> the same beads again and deconvolve the image, i would expect to get
>> real=
>>
>>
>>
>>
>>
>>> ly tiny dots. See e.g.
http://www.svi.nl/BeadsDeconvolutionExample=20>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> real tiny? The images of sub resolution objects dont get very much smaller
>> after
>>
>>
>>
>>
>>
>> deconvolution...
>>
>>
>>
>>
>>
>> what you really get is much higher contrast, so the features look "sharper"
>>
>>
>>
>>
>>
>> You also get a little bit more resolution.. but thats not really the main
>> point.
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> Its really about contrast.
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>> I would expect something similar for the OMX when the reconstruction is
>> pe=
>>
>>
>>
>>
>>
>>> rfect and includes proper deconvolution. However we get larger features,=
>>
>>
>>
>>
>>
>>> still within the expected OMX resolution 120 nm in width and 300 nm in
>> z,=
>>
>>
>>
>>
>>
>>> but no dots.
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> No reconstruction or deconvolution will give images that contain features
>> that
>>
>>
>>
>>
>>
>> are smaller
>>
>>
>>
>>
>>
>> than the resolution limit of the system.
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> What do you mean by "dots" ?
>>
>>
>>
>>
>>
>> Single 40 nm pixels?
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> In the OMX it's twice the conventional resolution,
>>
>>
>>
>>
>>
>> so you cant get objects that appear smaller than 120 nm.
>>
>>
>>
>>
>>
>> (actually... API say the OMX system sometimes seems to outperform theory
>>
>>
>>
>>
>>
>> slightly... but only a bit)
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> You should not an must not get single pixel object with the reconstructed
>> pixel
>>
>>
>>
>>
>>
>> spacing of 40 nm (thats the spacing we have with the EM CCDs after 2x
>> resolution
>>
>>
>>
>>
>>
>> increase from the physical 80 nm pixel spacing)
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> If you get smaller objects, its wrong, and an artifact. No?
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>> I think there is still some improvement possible either in=
>>
>>
>>
>>
>>
>>> the setup of the instrument or the software.
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> Actually, in my hands it seems to do what theory predicts and no more.
>>
>>
>>
>>
>>
>> There is no good reason to expect it to make smaller objects than 120 nm.
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> Certainly the careful alignment of the optics and the highest quality
>> objective
>>
>>
>>
>>
>>
>> lens are critical,
>>
>>
>>
>>
>>
>> as is the measurement of suitable SIM PSFs and careful
>> calibration/measurement
>>
>>
>>
>>
>>
>> of the parameters for the reconstruction...
>>
>>
>>
>>
>>
>> the angles and the phases.
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> So, other than making it faster and more tolerant to sub optimal input
>> data,
>>
>>
>>
>>
>>
>> I dont think there is any improvement to be made in resolution.. so I dont
>> see
>>
>>
>>
>>
>>
>> how you think there is room for resolution improvement?
>>
>>
>>
>>
>>
>> Maybe you can share your ideas?
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> cheers
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> Dan
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>>
>>
>>
>>
>>
>>
>>> best wishes
>>
>>
>>
>>
>>
>>>
>>
>>
>>
>>
>>
>>> Andreas
>>
>>
>>
>>
>>
>>>
>>
>>
>>
>>
>>
>>>
>>
>>
>>
>>
>>
>>> =20
>>
>>
>>
>>
>>
>>>
>>
>>
>>
>>
>>
>>> =20
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> Dr. Daniel James White BSc. (Hons.) PhD
>>
>>
>>
>>
>>
>> Senior Microscopist / Image Visualisation, Processing and Analysis
>>
>>
>>
>>
>>
>> Light Microscopy and Image Processing Facilities
>>
>>
>>
>>
>>
>> Max Planck Institute of Molecular Cell Biology and Genetics
>>
>>
>>
>>
>>
>> Pfotenhauerstrasse 108
>>
>>
>>
>>
>>
>> 01307 DRESDEN
>>
>>
>>
>>
>>
>> Germany
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> +49 (0)15114966933 (German Mobile)
>>
>>
>>
>>
>>
>> +49 (0)351 210 2627 (Work phone at MPI-CBG)
>>
>>
>>
>>
>>
>> +49 (0)351 210 1078 (Fax MPI-CBG LMF)
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
http://www.bioimagexd.net BioImageXD
>>
>>
>>
>>
>>
>>
http://pacific.mpi-cbg.de Fiji - is just ImageJ (Batteries
>> Included)
>>
>>
>>
>>
>>
>>
http://www.chalkie.org.uk Dan's Homepages
>>
>>
>>
>>
>>
>>
https://ifn.mpi-cbg.de Dresden Imaging Facility Network
>>
>>
>>
>>
>>
>> dan (at) chalkie.org.uk
>>
>>
>>
>>
>>
>> ( white (at) mpi-cbg.de )
>>
>>
>>
>>
>>
>>
>>
>>
>>