Hi Guy
I'm going to have to disagree with some of what you said because we are
talking about live cell imaging. As I'm sure we all know, opening the
pinhole has a big effect on z resolution and that is the main reason why
signal goes up in 3D objects (adding the wings of the in plane Airy disk
alone only adds about 25% signal). It's the increase in the size of the
sampled volume that generates more signal. Now I can see that you think
of an out of focus object as a "noise" generator, and in that sense I
agree with you, but if we are talking about thin cells how many bright
well out of focus objects are there? I really don't think you can say
that opening the pinhole will let in a LOT more "noise "(your definition
here) unless you know what the object really is. There is always a trade
off between S/N and resolution, so if you can take a reduction in
resolution you will be able to turn the lasers down and that is what
this thread started with -keeping cells alive.
I'm quite inflexible in my view on this, the correct _live cell_
solution is to open the pinhole until the x,y,z resolution is as low as
you can accept and then get the needed S/N with minimum laser power
-this is what we teach on the live cell course in Vancouver -it's all
about the keeping the cell healthy while getting just enough information
to answer the experimental question (and not minimizing "noise" per se.).
Yes, I agree with you that if the object z extent is << PSFz, you have
enough information to get a reasonable solution to the convolution
problem. But in that special case I suggest you don't need a confocal at
all! For very thin objects a WF microscope with a good CCD (+/-
deconvolution) will give a better result than a confocal (even with the
pinhole wide open as the confocal detector/optics are less efficient).
But in most other cases, a confocal is used to reject significant out of
focus light (as you seem acknowledge in your description of "noise"
increasing with pinhole diameter) and that contribution cannot be
properly defined by just 3 planes within the PSF (that's sometimes
called "nearest neighbours deconvolution" -but it's not a correct
solution to the deconvolution problem). Now you may use some simple
software to take the 3 planes of data to increase contrast and reduce
noise in the center plane but the image is not _deconvolved_ at all but
rather "deblurred" I think this is the case because it is not possible
to _know_ that all _out of focus_ information has been correctly
estimated with just 3 planes -and that is because of _photon noise_.
For an intuitive approach I like to think of it this way, if 3
planes(=samples) can only supply enough information to solve the spatial
problem (2 for Nyquist plus 1 for outside plane of focus generated
signal), what extra information is available to identify/control noise?
If you sketch a problem as 1D on a piece of paper one can see how all
the information in 3 planes is used for the spatial problem (e.g. if the
object is A.sin(wx)+B you have 3 unknowns A, B, w so you need 3
samples). Now if you add Poisson noise so that you know that the mean is
equal to the variance, how do you get an estimate of the mean when all
the variance in your 3 samples is already assigned to the equation
variables you just solved? From this I think it's clear why noise
introduces a problem and kills the 3 plane deconvolution solution for an
extended object. Let me be quite clear here, having extra information
planes can always help improve the in plane image (I call it
"deblurring" or "denoising" depending on the intent/method of the
processing) but it is not possible to actually deconvolve until
sufficient information has been captured to solve the spatial problem
_and_ get some estimate of noise. On the other hand, if the problem is
so constrained that you _know_ there is no data outside the plane of
focus (so that you have all the information needed to solve the
noise/spatial problem with 3 planes -and on this I think we agree) then
you shouldn't be using a confocal anyway IMO.
Cheers Mark
> On Sat, 14 Aug 2010 23:02:18 +1000, Guy Cox wrote
>
>> It's a bit challenging to disagree with Mark but .... we are interested
>> in signal over noise. Opening the pinhole beyond the diameter of the
>> Airy disk will let in a little more signal (from the outer rings)
>> and a LOT more noise. In almost every case it will make things worse.
>> Photons are precious - if they come from where we want - other
>> photons are something we need to exclude at all costs.
>>
>> As to the question about 3 sections - Mark is quite right, of course,
>> if we are dealing with a thick sample, but if I've followed this thread
>> correctly we are dealing with thin cells where the information is
>> largely in one plane. If this is so we should be able to do pretty good
>> deconvolution with 3 sections.
>>
>> Guy
>>
>> Optical Imaging Techniques in Cell Biology
>> by Guy Cox CRC Press / Taylor & Francis
>> http://www.guycox.com/optical.htm
>> ______________________________________________
>> Associate Professor Guy Cox, MA, DPhil(Oxon)
>> Australian Centre for Microscopy & Microanalysis,
>> Madsen Building F09, University of Sydney, NSW 2006
>>
>> Phone +61 2 9351 3176 Fax +61 2 9351 7682
>> Mobile 0413 281 861
>> ______________________________________________
>> http://www.guycox.net
>>
>> -----Original Message-----
>> From: Confocal Microscopy List [mailto:[log in to unmask]]
>> On Behalf Of Mark Cannell
>> Sent: Friday, 13 August 2010 8:52 AM
>> To: [log in to unmask]
>> Subject: Re: Optical slice thickness and number for PSF and
>> deconvolution
>>
>> Hi All
>>
>> I'm sorry but this advice is wrong. The pinhole is a control that
>> _should_ be used when decreased (mainly z) resolution is acceptable.
>> The
>>
>> lasers can then be turned down and, if desired, decon. can be used
>> to help clean up the image. The problem is that many users want a
>> "pretty picture" but pretty pictures may not be needed for
>> quantification of
>> (say) number of mitochondria. As we say on the Vancouver course, "Every
>>
>> photon is precious" and you may also increase signal by accepting a
>> wider spectral band or using an LP filter. The key to good experimental
>>
>> work is to understand what measurement you want and then to pick
>> conditions that allow you to get sufficient data with sufficient
>> (not too many) time points to answer your question. Do you need a
>> full 3D image or will a couple of slices suffice? Use a high NA
>> lens. As others have said, consider using widefield with a high QE
>> CCD if you really don't need the maximum possible resolution in 3D...
>>
>> My 2c
>>
>> Mark Cannell
>>
>> Vincent wrote:
>>
>>> *commercial interest*
>>>
>>>
>>> Dear Jan,
>>>
>>> The amount of the signal in images is mostly judged just after image
>>> acquisition. Based on this it is often decided to use a wider pinhole.
>>> As you probably know, when deconvolution is properly performed you
>>>
>> will gain not
>>
>>> only an increase in resolution but also in signal. Therefore, we
>>>
>> advise to close
>>
>>> the pinhole and use deconvolution for increasing the signal (to noise)
>>>
>> before
>>
>>> determining the quality of the image.
>>>
>>> As with imaging the object of interest it is important to follow the
>>>
>> Nyquist
>>
>>> criteria for imaging the bead images.
>>> We have a Nyquist calculator on our website
>>>
>> (www.svi.nl/NyquistCalculator) to
>>
>>> determine these rates. You can also create a picture here of your
>>>
>> theoretical
>>
>>> PSF to get an idea of its dimensions.
>>>
>>> In general it is best to really match the Nyquist criterion in xyz.
>>>
>> Else you can
>>
>>> go for 2x more. This however may introduce other problems like e.g.,
>>>
>> bleaching.
>>
>>> If the bead images are differently sampled it requires interpolation
>>>
>> for
>>
>>> matching that, making the process of deconvolution more
>>>
>> computationally
>>
>>> demanding. Thus Nyquist is okay. Another important thing to keep in
>>>
>> mind is that
>>
>>> you need to image enough planes to cover your PSF.
>>>
>>> I hope this answers your questions.
>>> Best regards,
>>> Vincent
>>>
>>> ***********************************************************
>>> Vincent Schoonderwoert, PhD
>>> Scientific Volume Imaging bv
>>> Hilversum, The Netherlands
>>> [log in to unmask]
>>> [log in to unmask]
>>> Tel: + 31 35 646 8216
>>> ***********************************************************
>>>
>>>
>>>
>>>
>>>
>>> Jan Trnka wrote:
>>>
>>>> Dear list,
>>>>
>>>> this is probably a trivial question but so far I haven't found a
>>>>
>> good answer.
>>
>>>> When taking 3D images of subresolution beads in a confocal
>>>>
>> microscope (for PSF
>>
>>>> construction) does the number and thickness of slices in the z-stack
>>>>
>> need to
>>
>>>> be exactly the same as that of a sample to be deconvolved? I
>>>>
>> understand the
>>
>>>> x-y dimensions need to be the same but how does it work for z? Would
>>>>
>> a higher
>>
>>>> number of thinner slices (finer z resolution) of the bead improve
>>>>
>> the
>>
>>>> construction of the PSF? My actual samples are imaged with a rather
>>>>
>> wide
>>
>>>> pinhole setting to limit the exposure of the sample (live cells) and
>>>>
>> thus
>>
>>>> provide quite thick optical sections.
>>>>
>>>> Thanks,
>>>>
>>>> Jan
>>>>
>>>> Jan Trnka, MD, PhD
>>>> Department of Biochemistry
>>>> 3rd Medical Faculty
>>>> Ruska 87
>>>> 100 00 Praha 10
>>>> Czech Republic
>>>> [log in to unmask] <mailto:[log in to unmask]>
>>>> Tel.: +420 26710 2410
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>>
>> No virus found in this incoming message.
>> Checked by AVG - www.avg.com
>> Version: 9.0.851 / Virus Database: 271.1.1/3032 - Release Date: 08/12/10
>> 04:34:00
>>
>
>
> Dr. Sudipta Maiti
> Associate Professor
> Dept. of Chemical Sciences
> Tata Institute of Fundamental Research
> Homi Bhabha Raod, Colaba, Mumbai 400005
> Ph. 91-22-2278-2716 / 2539
> Fax: 91-22-2280-4610
> alternate e-mail: [log in to unmask]
> url: biophotonics.wetpaint.com
>
|
