What I believe Guy is refering to is that images that are say a full 3D
mass of fluorochrome with areas that exclude dye, say a block of
fluorescent plastic with bubbles in the plastic, do not deconvolve well.
This is because the potential solutions are almost infinite. For samples
such as you mention, our deconvolution system would be very well suited
for. As for your software, there certainly are good and poor
implementations of the algorithms and the quality of the resultant images
will differ significantly with the quality of the software. This is one
reason why we wanted to create some agreed upon definitions of things
like resolution and linearity so that we could evaluate not only confocal
versus deconvolution, but also different deconvolution methods.
 
________________________________________________________________________________
 
 
Paul Goodwin
Image Analysis Lab
FHCRC, Seattle, WA
 
On Fri, 31 May 1996, Tracy Richmond McKnight wrote:
 
> On Fri, 31 May 1996, Guy Cox wrote:
>
> > Deconvolution varies from quite good (isolated point objects)
> > to completely ineffective (very extended objects).  As with
> > everything else in this game, we have to trade off one desirable
> > property to do better on another!
>
> Can you elaborate on this?  I have been dealing with this issue
> for the past few weeks.  I have images that I have collected with
> the confocal (MRC600) of a labelled receptor in the membrane/cortex
> of an oocyte.  The distribution is uniform - that is to say, "diffuse" -
> but has an underlying punctate patterning along with with larger bright
> regions. It looks like a multi-scale patterning that could be teased out
> with fourier analysis.  However, the diffuse images are not useful for
> such analysis so I applied a 2D deconvolution to try ans clean-up the
> images first.  The resulting images looked VERY MUCH like a simple
> threshold had been performed.  In fact, I could not tell the difference
> when I did an eyeball comparison.  Mind you, I wrote this algorithm
> myself and it may have a few bugs, but I couldn't help wondering
> if these images did not contain enough high-frequency information,
> i.e. were not punctate enough, for deconvolution to be effective.
> I've seen images similar to mine in the literature (P.J. Shaw, Fay, etc.)
> but they are nothing like the chromosomes of Agard and Sedat.
>
> What information do you have on this?  Can you point me to some
> articles on this subject?
>
>
>
>
>
> <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
> Tracy Richmond McKnight                 email:  [log in to unmask]
> Department of Human Physiology          TEL:    (916) 752-5584
> School of Medicine, MS-1A               FAX:    (916) 752-5423
> University of California
> Davis, CA 95616
> >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>
>
> > Jennifer Kramer wrote:
> >
> > >One important point here not mentioned is the presence of the pinhole or
> > >slit in the back focal plane of the objective lens.
> > >This can exclude up to 99% of all emitted light, requiring use of high
> > >probe concentrations (which can be potentially toxic).
> >
> > This sort of statement always mmakes me see red!  The only light the
> > confocal pinhole excludes is the *out of focus* light!!  The Airy
> > disk at the point of focus more or less all goes into the final
> > image (depending on the pinhole setting).
> >
> > The reason for the better light budget of deconvolution systems has
> > nothing to do with the pinhole and everything to do with serial vs
> > parallel collection of images.  In terms of light budget, a one
> > second scan of a confocal image (collecting 256K pixels) is equivalent
> > to a 4 microsecond exposure of a widefield image with the same
> > light intensity.  Looking at it the other way round, a single
> > widefield video frame (1/25 sec) is equivalent to scanning a
> > confocal image for 10,000 seconds!!
> >
> > Of course in confocal we try to overcome this by using brighter
> > light (a laser) but because fluorochromes saturate it isn't a total
> > solution.  And if we crank up the light beyond the saturation point
> > we get severe bleaching, as Jennifer points out.
> >
> > So, yes, deconvolution does have a much better light budget, but
> > NOT because "we are throwing away 99% of the light" in confocal.
> > On the other side of the coin, confocal is (to a very rough
> > approximation) object independent in its 3D imaging capability.
> >
> >                                         Guy Cox
> >
>