Shearing Interferometer with Ti-Sapph
Locked
 
|
0000001ed7f52e4a-dmarc-request |
Re: Sampling rate for confocal microscopy
|
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Why 16 bit when I only expect a few 100 photons in each pixel? Best wishes Andreas From: [hidden email] Sent: 24/09/2016 20:23 To: [hidden email] Subject: Re: Sampling rate for confocal microscopy Dear Friends: The proper way to calculate sampling rates in widefield and confocal microscopy can be found in this paper:Heintzmann, R. (2005). Band limit and appropriate sampling in microscopy. In Simons, K., S. J. V., Hunter, T., Shotton, et al. (Eds.), Cell biology: a laboratory handbook (pp. 29-36). Amsterdam: Elsevier. The example given in the paper show that (for NA 1.3, ex=488nm, em=500nm) 47.5 nm is what you want for X,Y sampling, and 93.8nm in Z. Also make sure to set your digitizer to the highest setting, 16 bit mode if possible. Guy Hagen On Sat, Sep 24, 2016 at 9:12 AM, <[hidden email]> wrote: ***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/ |
|
|
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
I’ve been trying to keep out of this one but ….
Most of this thread is based on a fundamental misconception, which is detailed quite clearly in Microsoft Technical Memo #6 (1995) by Alvy Ray Smith “A pixel
is not a little square, A pixel is not a little square, A pixel is not a little square, (and a voxel is not a little cube)”. This is the actual title. Pixels are samples–ideally they represent points, in practice they represent Airy disks. Nyquist theory
is based on fitting sine waves to your samples –not making boxes out of them. All scanning probe microscopes do this as a matter of course, no confocal ones do. Sampling at Nyquist is all you need (2.3 pixels per resel if you want to meet the Rayleigh criterion). Personally I recommend 3 pixels per resel since then
you can do a mild (cruciform) median filter and reduce noise substantially without reducing resolution. But then you must map your image to your samples, not just plonk them on the screen. Most of the suggestions posted here for higher sampling rates seem
to be based on avoiding this step and getting a decent image by oversampling instead. But this is killing your fluorochrome and worsening your signal to noise ratio. Read Microsoft Technical Memo #6. Read Microsoft Technical Memo #6. Read Microsoft Technical Memo #6.
Guy Guy Cox, Honorary Associate Professor School of Medical Sciences Australian Centre for Microscopy and Microanalysis, Madsen, F09, University of Sydney, NSW 2006 From: Confocal Microscopy List [mailto:[hidden email]]
On Behalf Of Andreas Bruckbauer ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Why 16 bit when I only expect a few 100 photons in each pixel?
From:
[hidden email] ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Dear Friends: The proper way to calculate sampling rates in widefield and confocal microscopy can be found in this paper: The example given in the paper show that (for NA 1.3, ex=488nm, em=500nm) 47.5 nm is what you want for X,Y sampling, and 93.8nm in Z. Also make sure to set your digitizer to the highest setting, 16 bit mode
if possible. I can provide a PDF of the paper to anyone who contacts me.
cheers,
On Sat, Sep 24, 2016 at 9:12 AM, <[hidden email]> wrote: ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Dear all, Many thanks for the replies, it looks like opinion is divided with George suggesting 100 nm and Giulia from SVI suggesting the 46 nm of the Nyquist calculator.
It would be best to do the test to see if there is any improvement with the 40 nm sampling, does anyone know of papers looking into this? I would have thought that there is a sweet-spot and further reducing the pixel size (at the same total acquisition time)
would just lower the signal to noise ratio and any useful signal would be buried in noise. But I guess the deconvolution algorithm nicely sums up the pixels to get the signal back. Is there any advantage of acquiring e.g. one big pixel instead of four smaller
ones? I am thinking of something similar like minimising read-out-noise of CCD cameras by binning, just for the confocal case, or does the PMT and read out electronics behave like CMOS where there is no advantage of binning?
best wishes Andreas -----Original Message----- ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Hello Andreas, Please do not hesitate to contact us if you have any further questions, we are happy to help you! Best regards,
Giulia 2016-09-22 17:37 GMT+02:00 Tobias Rose <[hidden email]>: ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Just quickly: * optimize laser power (low), detector settings (modest gain, maybe higher than you think), scan speed (fast is good, resonant scan is faster and therefore
better) with averaging (median would be even better, if in your software or you are willing to make the extra effort). Why median?
Tobias On 9/22/2016 7:22 AM,
[hidden email] wrote:
|
|
|
Peter Rupprecht-2 |
Re: Shearing Interferometer with Ti-Sapph
|
|
In reply to this post by Stephen Smith
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Dear Stephen, >> The divergence and beam waist will shift between mode locked and CW operation too. Is this true? I'm using routinely CW mode of a TiSa laser to align a 2P microscope setup - do you think that these alignments might be systematically wrong? Or do you know how strong these changes in divergence are? Best, Peter |
|
|
0000001ed7f52e4a-dmarc-request |
Re: Sampling rate for confocal microscopy
|
|
In reply to this post by Guy Cox-2
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
One old AFM software, which I used 15 years ago did actually do this, it moved to the next sampling point and then took a measurement averaging over several cycles before moving to the next point (taking no measurement while moving). I guess today's confocals scan smoothly along the lines, averaging the output signal of the PMTs while doing so and then attribute the average signal to a certain sampling point (pixel)? For most applications there should not be much difference, but would it matter for highest resolution imaging (not the topic of my original post)? How accurate are the scanners (any jitter?) and is the averaging always done correctly? Are there differences between manufacturers or old vs. new model of confocals?
best wishes
Andreas -----Original Message-----
From: Guy Cox <[hidden email]> To: CONFOCALMICROSCOPY <[hidden email]> Sent: Sat, 24 Sep 2016 23:20 Subject: Re: Sampling rate for confocal microscopy *****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
I’ve been trying to keep out of this one but ….
Most of this thread is based on a fundamental misconception, which is detailed quite clearly in Microsoft Technical Memo #6 (1995) by Alvy Ray Smith “A pixel
is not a little square, A pixel is not a little square, A pixel is not a little square, (and a voxel is not a little cube)”. This is the actual title. Pixels are samples–ideally they represent points, in practice they represent Airy disks. Nyquist theory
is based on fitting sine waves to your samples –not making boxes out of them. All scanning probe microscopes do this as a matter of course, no confocal ones do.
Sampling at Nyquist is all you need (2.3 pixels per resel if you want to meet the Rayleigh criterion). Personally I recommend 3 pixels per resel since then
you can do a mild (cruciform) median filter and reduce noise substantially without reducing resolution. But then you must map your image to your samples, not just plonk them on the screen. Most of the suggestions posted here for higher sampling rates seem
to be based on avoiding this step and getting a decent image by oversampling instead. But this is killing your fluorochrome and worsening your signal to noise ratio.
Read Microsoft Technical Memo #6. Read Microsoft Technical Memo #6. Read Microsoft Technical Memo #6.
Guy
Guy Cox, Honorary Associate Professor
School of Medical Sciences
Australian Centre for Microscopy and Microanalysis,
Madsen, F09, University of Sydney, NSW 2006
From: Confocal Microscopy List [[hidden email]]
On Behalf Of Andreas Bruckbauer
Sent: Sunday, 25 September 2016 7:13 AM To: [hidden email] Subject: Re: Sampling rate for confocal microscopy ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Why 16 bit when I only expect a few 100 photons in each pixel?
Best wishes Andreas From:
[hidden email]
Sent: 24/09/2016 20:23 To: [hidden email] Subject: Re: Sampling rate for confocal microscopy ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Dear Friends:
The proper way to calculate sampling rates in widefield and confocal microscopy can be found in this paper:
Heintzmann, R. (2005). Band limit and appropriate sampling in microscopy. In Simons, K., S. J. V., Hunter, T., Shotton, et al. (Eds.), Cell biology: a laboratory handbook (pp. 29-36). Amsterdam: Elsevier. The example given in the paper show that (for NA 1.3, ex=488nm, em=500nm) 47.5 nm is what you want for X,Y sampling, and 93.8nm in Z. Also make sure to set your digitizer to the highest setting, 16 bit mode
if possible.
I can provide a PDF of the paper to anyone who contacts me.
cheers,
Guy Hagen On Sat, Sep 24, 2016 at 9:12 AM, <[hidden email]> wrote:
***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Dear all,
Many thanks for the replies, it looks like opinion is divided with George suggesting 100 nm and Giulia from SVI suggesting the 46 nm of the Nyquist calculator.
It would be best to do the test to see if there is any improvement with the 40 nm sampling, does anyone know of papers looking into this? I would have thought that there is a sweet-spot and further reducing the pixel size (at the same total acquisition time)
would just lower the signal to noise ratio and any useful signal would be buried in noise. But I guess the deconvolution algorithm nicely sums up the pixels to get the signal back. Is there any advantage of acquiring e.g. one big pixel instead of four smaller
ones? I am thinking of something similar like minimising read-out-noise of CCD cameras by binning, just for the confocal case, or does the PMT and read out electronics behave like CMOS where there is no advantage of binning?
best wishes
Andreas
-----Original Message-----
From: Giulia De Luca <[hidden email]> To: CONFOCALMICROSCOPY <[hidden email]> Sent: Thu, 22 Sep 2016 16:55 Subject: Re: Sampling rate for confocal microscopy ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Hello Andreas,
As imaging specialist at SVI I am happy to answer your question. We always suggest to use our Nyquist Calculator, that you nicely link, to calculate the ideal x, y and z sampling specific for your imaging settings. This will assure that you optimally preserve the spatial frequencies in your images for further image restoration with deconvolution. Such small sampling distances may arise the worry for bleaching. However, because the restoration procedure will improve the Signal to Noise Ratio of the images dramatically, this should not worry you! You can proceed in lowering the laser power or the exposure time/acquisition speed and then proceed with deconvolution afterwards. You will see that the noise will be taken care of in the deconvolution process. Please do not hesitate to contact us if you have any further questions, we are happy to help you!
Best regards,
Giulia
2016-09-22 17:37 GMT+02:00 Tobias Rose <[hidden email]>:
***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Just quickly:
* optimize laser power (low), detector settings (modest gain, maybe higher than you think), scan speed (fast is good, resonant scan is faster and therefore
better) with averaging (median would be even better, if in your software or you are willing to make the extra effort).
Why median?
Tobias
On 9/22/2016 7:22 AM,
[hidden email] wrote:
|
|
|
George McNamara |
Re: Sampling rate for confocal microscopy
|
|
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Hi Andreas, Yes, to most of your questions. What matters the most for you is your confocal: test it! If you have a non-photobleaching specimen, you can test repeatedly on that (pollen grains are a lot less expensive than http://argolight.com/argo-hm , if you can afford the latter, great. If your instrument is working great, no laser fluctuations over time (if the lasers or electronics are not cooled well: good luck). If your vibration table is not floating, then your building may cause more jitter than your hardware (a good test: knock hard on the table with your knuckles). Back to your max photons ~100, why go 16-bit (0 ... 65535) when 8-bit is sufficient (0..255). You should deconvolve, so that will "suck all the photons" (really all the numbers from the photo-electrons) to where there belong. As a simplified point spread function I will use, in focus: 25 50 25 50 100 50 25 50 25 sum = 400 and at appropriate distances, symmetrical above and below in focus planes of: 12 25 13 25 50 25 13 25 12 sum of each = 200, above + below = 400 If we know the object of interest is sub-resolution, it would be great to "suck all these counts" into the central pixel, total = 1200 counts. If we knew the conversion efficiency of everything (and had zero photobleaching, and always used the same laser power, scan speed etc), quantum yield of the fluorophore(s), degree of labeling of the antibody or oligonucleotide (and pretend either was uniform, and no self quenching). we could convert to: total fluorophores = 3000 (which should be the same regardless of instrument). number of biological molecules = 1. If using Expansion Microscopy, see for example, Fig 4 of Tsanov et al 2016 NAR http://nar.oxfordjournals.org/content/early/2016/09/05/nar.gkw784.long ExM has practically no background autofluorescence (compare to
fig 2 negative control columns, which have both autofluorescence
and occasional non-specific binding of one or very few oligos). Of
course after ~4x expansion per axis, that's a lot of voxels (4^3 =
64x more than at 1x specimen size). The way the oligo FISH probe
sets work, the expectation is that not every oligo will bind its
target site on the RNA. So, if 48 oligos are used to tile one RNA,
maybe 20-40 will bind, because of RNA binding protein(s) and/or
ribosomes, masking some binding sites (I'm ignoring splicing and
SNPs: more colors can be used for those). If we have reasonable expectation that one biological molecule is in the volume, the data could be simplified to a list similar to single molecule localization (STORM, PALm, etc acronyms), so simply: X,Y,Z = 1200 counts or = 3000 fluorophores or = 1 molecule (with twice as bright implying 2 molecules), with X,Y,Z being either integer (a bit lazy, but very compact),
or single precision float (ex. if the fluorophore is not centered
on one of Guy's & Alvy Ray's pixels). Depends on what you need
to do. Likewise, segmentation of the plasma membrane (Na+,K+-ATPase,
EGFR, CD45, CD3, CD4, CD8, c-Kit, etc), nuclear envelope (nuclear
pore complex, see panel 3A, right side, red dots, also gaps in
green of
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2916659/figure/F3/ )
etc, could be reduced to compact sets of numbers. This could save
a lot of bytes compared to keeping the raw Expansion Microscopy
voxel data. For more on Expansion Microscopy, use PubMed search "expansion
microscopy" (10 references so far, 1-8 are relevant). // Guy's favorite Microsoft technical memo is online at http://alvyray.com/Memos/CG/Microsoft/6_pixel.pdf More memors, titles at top, abstracts and download links, below http://alvyray.com/Memos/MemosCG.htm (Texas is not about the state I currently live in).
enjoy, George On 9/25/2016 8:04 AM,
[hidden email] wrote:
***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** -- George McNamara, PhD Houston, TX 77054 [hidden email] https://www.linkedin.com/in/georgemcnamara https://works.bepress.com/gmcnamara/75/ http://www.ncbi.nlm.nih.gov/myncbi/browse/collection/44962650 |
|
|
In reply to this post by Heping Yuan
*****
To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** To align a laser 2P I always use a stationary laser (633 nm) and a mirror under a 10X objective to align the pinholes through which will pass the femto laser. I always add an additional 1M lens focal length. And then i use the reflection mode to visualize the beam on one pmt. |
|
|
In reply to this post by George McNamara
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Hello: As for the other questions from Andreas, remember that the scan rate is constant over the field of view of a confocal. Otherwise, the image will look distorted, stretched, warped, or something - one can check it with a stage micrometer or something similar. Also remember there are 2 kinds of averaging. The integration of the signal during the dwell time and the averaging from repeating the scan multiple times. Both of these are expected to be reliable, except that you could get a noise spike that ruins the measurement in a particular pixel. Also keep in mind that the PMTs in confocals are used in analog mode. You are not counting photons but rather digitizing a voltage due to a photocurrent. I would say the biggest difference between older generation and newer confocals is that much better detectors are now available. (I guess some of the newest models can be run in photon counting mode(?)). Guy Hagen On Sun, Sep 25, 2016 at 8:32 AM, George McNamara <[hidden email]> wrote: ***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/ |
| Free forum by Nabble | Edit this page |