One more comment on the geometry issue. One problem that we have
encountered on confocal is that images may look just fine and lead you to
believe that the images are accurate reflections of the "true" locations
of fluorochromes in the original image. In reality, they are at least as
vulnerable to bad optics as deconvolution. So called "Lensing" effects
and other distortion in the optical path (like spherical abberations)
will effect, sometimes greatly so, images collected on a confocal even
though the image "looks OK". This incidious nature of confocal can be
quite embarrasing when it is pointed out and most people are totally
unaware of it. I suggest you look at a bunch of fluorescent beads on top
of each other sometime.
 
Ciao, I'm sure we'll be talking more later.
 
________________________________________________________________________________
 
 
Paul Goodwin
Image Analysis Lab
FHCRC, Seattle, WA
 
On Thu, 30 May 1996, Eric A. Shelden wrote:
 
> On Wed, 29 May 1996, Dennis McKearin wrote
> >         I am interested in learning the pro & cons of confocal vs
> > deconvolution methods of microscopy. Specifically, I wish to know how the
 
 
>
> Well, just to get things rolling on this thread, I'll put my $.02 in.
> Recognize that both technologies are in the process of evolution, so what
 
much deleted
 
>
> Confocal microscopes rely on the geometry of emmited light paths to
> filter out out-of-focus illumination from a specimen. Usually this has
> been implemented by scanning a laser beam across the specimen with a
> series of scanning mirrors, with emitted light descanned through much of
> this pathway. The detectors at the end of all this are usually
> photomultipliers. Generally, there is a significant trade-off between
> image quality, scanning speed and sensitivity, as expected. The beauty of
> the technique is that the formed image is "real" in the sense that it is
> not computationally derived from estimates (although very good ones) of
> what light "should" be doing in your optical system. A side-by-side
> comparison of images of the same specimen using the two techniques
> suggests that confocal microscopes can attain slightly higher resolution,
> partially due to a phenomenon achieved by using coherent laser light
> called "super resolution" - see A Handbook of Confocal Microscopy,
> incidentally), and partially due to the difficulties in estimating
> (mathematically) the spread of light very close to an in focus point. As
> a result, microtubules in confocal microscopes (properly adjusted and
> used) seem to be thinner and more clearly resolved than can be achieved
> by (commercial) deconvolution software. Limitations, particularly for
> issues related to multispectral imaging, are related to the use of lasers
> to illuminate multiple fluorophores - you can do this but it won't come
> cheap and may in fact involve buying (and maintaing) more than one laser,
> and the relatively skewed spectral efficieny of photomultiplier tubes
> used for imaging - some of the confocal microscopes I've used see
> fluorescein-like dyes MUCH better than rhodamine-like dyes. Multispectral
> imaging seems to be particularly difficult so implement fast and
> efficiently, in some implementations on currently available instruments,
> multiple fluorophores are illuminated with the same excitation spectrum,
> chosen so as to excite more than one fluorophore - although neither are
> excited at anything like peak efficiency, thus requiring higher
> illumination levels and longer/slower scanning times to obtain an image
> of the same quality that could be achieved by scanning the specimen with
> filters optimized for individual fluorophores. Scanning is also slow,
> although higher scan rates are accomplished by either scanning at lower
> resolution or by scanning smaller areas.
>
> Deblurring (can) use standard epifluorescent illuminators and captures
> all of the emitted light coming from the specimen, forming images with
> standard imaging devices such as SIT camaras or CCD arrays. The
> techniques relies on the ability to mathematically estimate the
> contribution of adjacent areas in the specimen to a given image and
> subtracting this information.A consideration is that the technique is
> only as good as the starting images - noise images from low-light level
> imaging devices do not work as well as images collected over long
> integration times with cooled CCD camaras. Also images from specimen
> areas burried deeply in a specimen that's not very transparent are also
> going to be a problem, while confocal microscopes (won't get much light
> from this area either) but may be more likely to build up an acceptable
> image over time. For a discussion of these methods you are of course
> refered to articles by Agard and Sedat, UCSD? or Fay at UMass Worchester,
> (my sincere apologies to others I am not mentioning at the momment
> because -hey I AM on vacation and my brain doesn't have to work that
> well). In principle, deconvolution could be used on images which are
> captured VERY quickly - much faster than could be achieved by scanning
> techniques, and with the judicious use of filters and optics could obtain
> multispectral images simultaneously or nearly so. In practice, of course
> whis would be quite difficult to do with a living specimen (which is the
> only thing you really need to do this with) because of the need to obtain
> relatively noise-free images. Although deconvolution is computationally
> demanding and can require hours of computation to implement some
> algorithms, this can all be done after the fact, allowing rapid
> collection of data. In my experience, imaging devices can also be
> substantially more sensitive than confocal microscopes allowing much
> lower illumination levels to achieve a reasonable image. Deconvolution
> can also be implemented somewhat less expenxively than buying a confocal
> microscope - although after adding the cost of the microscope, software
> and camaras up I'd hesitate to call the technology cheap.
>
> Both technologies have limitations in dealing with out-of-focus
> information in thick specimens, ie if your area of interest is located
> behind a brick, neither technique works well. Deconvolution is
> particularly susceptable to neighboring bright objects punching
> subtractive holes in one's image.
>
> So, ideally I'd like to have both around :-). If you need to get images
> of lots of different fluorophores from the same specimen and budgets are
> tight I'd have to give the deconvolution systems a close look. My
> personal feeling is that confocal microscopy is easier to use for the
> uninitiated and perhaps more powerful for penetrating thick and
> semi-opaque specimens, but slower, more expensive and less adaptable in
> the hands of an experienced digital microscopist. (You might for example
> see what Dr. S. Inoue has been up to).
>
> Since I've stuck my neck out really far here, I'll be very interested in
> other peoples opinions.
>
> Regards,
>
>
> Eric.
>