Search the CONFOCAL archive at
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
Jim, Mario et al.,
please use www.powermicroscope.com and let me know!
All my best
ALby
On Venerdì, mar 12, 2004, at 18:15 Europe/Rome, James Pawley wrote:
> Search the CONFOCAL archive at
> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>
> Carl,
>
>
>
> A generic quick response: for specimens that have very few copies of
> fluorophores, and/or a weak signal, the confocal "iris" can be opened
> to acquire more photons. Your point that "With confocal images, out of
> focus light is already essentially eliminated," is true but it is also
> the reason that conventional confocals are so inefficient at
> collecting light. Light is not reassigned, it is thrown away. Although
> opening the iris makes the images brighter, there is a loss of
> resolution especially along the z-axis. In effect, you are sampling a
> bigger volume/voxel, hence the better signal. With proper calibration
> and good software, deconvolution can restore the lost resolution, and
> when properly done, correct for spherical aberration to some extent.
> The latter could be of value even for bright images that provide good
> S/N that normally one would not think to bother deconvolving.
>
>
>
> I cannot stress the importance of doing accurate calibrations for
> deconvolution. Also, the algorithms, are not all the same. You mention
> Autoquant but they offer a few different methods, as do other vendors.
> My prejudice is that the deconvolution routine be energy conserving
> (uses all the signal from the object), and that the algorithm fits a
> PSF function to calibration data from sub-resolution fluorescent
> spheres. Some routines use the raw bead data for deconvolution but I
> think this increases the noise problem, which for the type of sample
> this discussion applies to is an issue. This kind of detail is usually
> not known by the sales rep. so you might talk to tech. support to see
> why it hasn't worked for you. You probably will need to acquire more
> out of focus frames, as well. This is important for deconvolution, but
> not usually done for confocal.
>
>
>
> If you have the calibration data and an image stack to test, I may be
> able to try a different routine. Just send me a note directly if you
> are interested.
>
>
>
> Mario M.
>
>
>
>
> I am a great fan of ALWAYS deconvolving 3D confocal data!
>
> There are two reasons beyond those mentioned so far: improving
> resolution and removing out of focus light.
>
> These other two are
>
> 1. The Nyquist sampling rule is not just about recording data but also
> about "reconstructing" the analog signal from it. A crucial part of
> the latter process is the condition that, before being displayed,
> the reconstructed analog signal should be passed through an
> '"amplifier" with the same bandwidth as the amplifier from which the
> original data was sampled.
>
> In microscopy, the "input amplifier" is the microscope optical system
> that limits the spatial bandwidth of the signal. We all recognize by
> now that, if the "bandwidth" of the optics is accepted as the Abbe
> resolution, then you need about 2.5 pixels between the center of the
> Airy peak and the first dark ring or, in other words, that the image
> of a point object should be recorded as a blob about 5 pixels wide.
>
> Many assume that, as only low bandwidth info is digitized, there can
> be no high bandwidth info present in the digital data and so it is not
> necessary to remove it after the numbers are turned back into an
> analog version.
>
> This is not true for signals containing a lot of Poisson noise. If the
> peak signal is only 10 photoelectrons/pixel on average, this signal
> will contain 30% Poisson noise and adjacent pixels fed a "constant
> signal" may easily vary by 55%, (from 7 to 13 PE) giving entirely
> artefactual single-pixel features. These "features" will be one pixel
> wide or about 5x smaller than the smallest features that could had
> passed through the microscope.
>
> Short of defocussing the electron beams on your CRT to reduce the
> resolution of the monitor by 5x, the only way to mimic the "maximum
> spatial bandwidth criterion" imposed by Nyquist reconstruction is to
> use deconvolution. Unlike defocusing the CRT, deconvolution has the
> added (and essential) advantage that it can be done in 3D, not just
> 2D. This makes it the right and proper way to implement Nyquist
> reconstruction. (Note: As the PSF in confocal is almost a Gaussian,
> you can do almost as well by using a properly-dimensioned 3D Gaussian
> filter.)
>
> 2. Deconvolution essentially averages the signal over the many (125?)
> voxels needed to sample the signal from a single point object sampled
> at Nyquist frequency. As far as S/N is concerned, this is almost the
> same as trying to collect 125x more data in the first place (Kalman
> 125x?) but doesn't take so long or damage the specimen (nearly!!!) as
> much. (Keep this in mind those of you who image Calcium ion
> concentrations: smooth you data before you ratio it.)
>
> Of course, when you deconvolve confocal data, you will find the
> contrast drops and the image doesn't look so "sharp". You should be
> glad!! Those "sharp" objects were noise artifacts and the contrast can
> always be raised to match the characteristics of the display monitor
> to your personal optical sensing system.
>
> Failing to do this is not only condemns you to lower S/N than needed,
> it can let structural artifacts into your final images.
>
> Happy Spring break!
>
> Jim Pawley
>
> --
>
> **********************************************
> Prof. James B. Pawley, Ph.
> 608-263-3147
> Room 223, Zoology Research Building,
> FAX 608-265-5315
> 1117 Johnson Ave., Madison, WI, 53706
> [log in to unmask]
> 3D Microscopy of Living Cells Course, June 12-24, 2004, UBC, Vancouver
> Canada
> Info: http://www.3dcourse.ubc.ca/ Applications due by March
> 15, 200i
>
........................................................................
....................................
... non basta, resistere, resistere...resistere
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Alberto Diaspro
Department of Physics, University of Genoa
Via Dodecaneso 33, 16146 Genoa, Italy
voice: +39-0103536426/480 fax 010314218
[log in to unmask], http://www.lambs.it
http://wiley.com/cda/product/0,,0471409200,00.html
|
