On Feb 20, 2:25pm, Steffen Dietzel wrote:
Hi Steffen,
>
> I have a question concerning the resolution of the confocal microscope:
>
> Normally the resolution of a confocal is given as the distance by which two
> points must be separated to resolve them in two points (about 300x600nm x,y
> and z.). But this is not the type of experiment I am interested in. I want
> to determine the gravity centers of point-like Fluorescence in situ
> hybridization signals (cosmid or YAC-probes) and then (by 3D image analysis)
> the shortest distance of these gravity centers to a nearby surface of a
> large structure recorded in another color (a chromosome). This larger
> structure has a strong gray-level increase over several pixels at the border
> and in the core the gray-levels move around a mean.
First two answers to this and earlier postings:
At the university of Amsterdam Karel Strasters and I developed a reliable and
easy method to measure the point spread function of a fluorescence confocal
with fairly large latex beads (230nm diameter). In principle these beads are
far too large to measure a PSF directly. (For 'theoreticians' (not me!): their
spectrum has lots of zeros within the passband of the microscope. Even 50nm
beads are not good enough...) We overcome that problem by submitting the bead
image to an image restoration algorithm: out comes the PSF. You can find all
details in JoMi 178, 1995, pp165-181.
One detail is of interest to you: to increase the S/N ratio in the bead image
we
had to average several of them. In order to do so we had to align their center
of mass (CM). Like you, we wondered about the precision by which we could do
this. We needed a standard deviation (SD) of 10-25nm. So we did (noisy)
computer simulations and found that for these very bright beads the SD of the
CM is below 10nm. Smaller beads could not be used (at that time) because they
can't be aligned&averaged.
Sounds complex? The procedure is now automated, contact us if you are
interested.
Next, and this is what this long story is all about, we used this PSF to
restore our images, roughly of the type you have. This 'untangles' neighboring
spots, corrects for the weird PSF (symmetric, ha!) and improves resolution on
the whole.
>
> My questions are:
> (1) How accurate (in nm) can the gravity center of a point like signal can
> be determined (relative to other signals) and
See above for an upper limit. However, for your particular S/N level and object
size you should repeat the experiments (not very difficult).
> (2) How accurate can the position of a "surface" be determined, lets say at
> a particular threshold?
As far as I know such figures cannot be computed analytically, so I am afraid
you have to do simulations. Also not very hard.
For backgrounds see the recent paper by Geert van Kempen in IEEE Eng. in Med.
and Biol. (1996, pp76-83).
>
> I am working on a Leica TCS System and use mainly the 63x, 1.32NA PL APO lens
Should be no problem.
Regards to all of you, Hans.
--
--------------------------------------------------------------
dr. Hans T.M. van der Voort ([log in to unmask])
Scientific Volume Imaging BV,
J. Geradtsweg 181,
1222 PS Hilversum, The Netherlands.
World Wide Web: http://www.svi.nl/
tel: (0)35-6859405 or (0)653-345445
fax: (0)35-6837971
|
