*****
To join or leave the confocal microscopy listserv or to change your email address, go to:
https://lists.umn.edu/cgi-bin/wa?SUBED1=confocalmicroscopy&A=1
Post images on http://www.imgur.com and include the link in your posting.
*****

Hi Jeremy,

thanks for making this point:

    you need a lot of photons - but an average of 1000 photons still has a standard deviation of around 33 photons(3%), while for 255 photons the standard deviation is 16 (6%) and for 100 will be 10%. Normally users have no idea how many photons are present.

Most of new new confocal microscope users here at JHU SOM are graduate 
students or postdocs. Some have no idea what bit depth is or where 255 
(2^8 - 1) or 65535 (2^16 - 1) come from (I never spend time on 12-bit 
mode on the SP8). Some (when I visit their later sessions) operate in 
8-bit mode, even though I stress in training they need to spend the time 
to go to the bit depth drop down list and select 16-bit mode. (Leica 
makes this worse by having LAS X not re-load all the settings correctly 
on the settings configuration file I save for each user for each 
fluorophore set).
In principle, understanding image histograms (see start of thread) would 
help, but I don't think most users understand "IH's".
Back to 255 ... if the user is in 8-bit mode on the Leica photon 
counting mode (or similar now that other vendors have this), while they 
are acquiring (post-training, without me to remind them to use 16-bit 
mode) they likely have no idea that they "hit saturation" and that 
signal might have been 300, 400, 1000, etc if they had operated the 
microscope correctly (this sentence reminds me of all the bad and/or 
reckless drivers on the roads, at least here in the U.S.).
So, if the user hits 255 in two different channels, and then figures out 
how to do image correlations (ImageJ doesn't care how good or bad the 
input is), Green=255 and Red=255 will be wonderfully correlated.
Of course researchers do much worse things than bad setting on a 
microscope, for examples, see the current DFCI scandal,
https://forbetterscience.com/2024/01/02/dana-farberications-at-harvard-university 
... my favorite segment is the WB especially the "? lanes" in Hetz, 
though there micrograph, Mn-SOD ~ mtBAX ~ Merge is pretty uninformative 
(and lacks scale bar, so N=2 cells art work, not real data)
If you prefer Nature news version (which rehashes the above blog), see 
https://www.nature.com/articles/d41586-024-00202-9

George
p.s. I am a big fan of image histograms, essentially never train any 
user on it because they are only interested in acquiring data. The very 
rare time a new or experienced user, or their P.I. has asked about image 
channels correlations, I (try to not laugh out loud) encourage them to 
figure it out. Of course I don;t trust any image correlations in the 
listerature (I do like Keremy's papers on C's and other topics).



On 2/2/2024 3:46 AM, Jeremy Adler wrote:
> *****
> To join or leave the confocal microscopy listserv or to change your email address, go to:
> https://lists.umn.edu/cgi-bin/wa?SUBED1=confocalmicroscopy&A=1
> Post images on http://www.imgur.com and include the link in your posting.
> *****
>
> Noise can badly distort quantitative measurements - user should be encouraged to test this before acquiring large datasets. If you can use images with higher noise then you  can acquire more images in the same session and maybe get more useful data e.g. counting nuclei.
>
> But usually noise degrades measurement - Correlation measurements used to quantify colocalization between images are particularly dependent on noise- 2 images and each is noisy.
> The measured correlation always has a reassuring numbers of decimal places but is inevitably an underestimate - repeating the measurements with images based on more photon reduces the gap between the ground truth and measured correlation but it is still there.
> Importantly correlation is not greatly improved by using a larger number of pixels in the calculation - noise is still a killer.
> But as George hinted at earlier,  you need a lot of photons - but an average of 1000 photons still has a standard deviation of around 33 photons(3%), while for 255 photons the standard deviation is 16 (6%) and for 100 will be 10%. Normally users have no idea how many photons are present.
>
> With photon counting you at least know how good the image is, whereas with an integrated analog signal with variable gain, users have little idea how noisy the image is - the images look fine on the screen and have large intensity range -
> you can estimate the analog noise by looking at how the image changes as it rescanned - with some simple software this could even be instantly quantified but as far as I am aware no manufacturer of our expensive microscopes has bothered.
>
> In the limited area of correlation measurements there is a solution that does not require large numbers of photon.
> J. Adler, S.N. Pagakis and I. Parmryd (2008)
> Replicate Based Noise Corrected Correlation for Accurate Measurements of Colocalization
> J. Microscopy 230(1),121-133.
>
>
> Jeremy
> ===============================================
>                      B i o V i s   P l a t f o r m of  Uppsala University
>                     Light & EM microscopy / FlowCytometry & Cell Sorting / Image Analysis
> ===============================================
> Jeremy Adler   PhD - Senior research engineer
> Light, Confocal Microscopy, Image Analysis
> E-mail: [log in to unmask]
> 070-1679349
>
> Dag Hammarskjölds v 20
> 751 85 UPPSALA, SWEDEN
> http://biovis.uu.se/
> ===============================================
>
>
>
>
>
>
>
> J. Adler, S.N. Pagakis and I. Parmryd (2008)
> Replicate Based Noise Corrected Correlation for Accurate Measurements of Colocalization
> J. Microscopy 230(1),121-133.
>
> -----Original Message-----
> From: Confocal Microscopy List <[log in to unmask]> On Behalf Of George McNamara
> Sent: Friday, February 2, 2024 2:56 AM
> To: [log in to unmask]
> Subject: Re: Question about proper use of HyD detectors
>
> *****
> To join or leave the confocal microscopy listserv or to change your email address, go to:
> https://lists.umn.edu/cgi-bin/wa?SUBED1=confocalmicroscopy&A=1
> Post images on http://www.imgur.com and include the link in your posting.
> *****
>
> Yes.
>
> On 2/1/2024 12:11 PM, Cammer, Michael wrote:
>> *****
>> To join or leave the confocal microscopy listserv or to change your email address, go to:
>> https://lists.umn.edu/cgi-bin/wa?SUBED1=confocalmicroscopy&A=1
>> Post images on http://www.imgur.com and include the link in your posting.
>> *****
>>
>> George-
>>
>> When you leave the room, do people really take images with 10x line accumulation?
>>
>> Cheers-
>> Michael Cammer
>>
>> -----Original Message-----
>> From: Confocal Microscopy List <[log in to unmask]> On
>> Behalf Of George McNamara
>> Sent: Wednesday, January 31, 2024 7:46 PM
>> To: [log in to unmask]
>> Subject: Re: Question about proper use of HyD detectors
>>
>> [EXTERNAL]
>>
>> *****
>> To join or leave the confocal microscopy listserv or to change your email address, go to:
>> https://lists.umn.edu/cgi-bin/wa?SUBED1=confocalmicroscopy&A=1
>> Post images on http://www.imgur.com and include the link in your posting.
>> *****
>>
>> Hi Michael, and other listserv readers,
>>
>> I train all our Leica SP8 confocal microscope users to use our two HyD's in photon counting mode and 16-bit mode. Around 300 counts the data is really good; at 1000 counts, a waste of time since the data is going to be just a little better than at 300 counts (users can use less line accumulation for that scan track). I never bother wasting time on "gain"
>> (default gain is 10x the photon count). I recommend 10 line
>> accumulation (no frame accumulation), 600 Hz (enables full range of
>> scanning zoom down to what Leica calls 0.75x). For 405nm laser 0.5 or
>> 1% laser power, for 488, 552, 638nm lasers, 1% or 2% or (rarely) 4%.
>> In training, I show new users what frame accumulation does (i.e.
>> usually max 16 line * say 4 frame accumulations) if they bring dim
>> sample(s) to the training. The
>> SP8 has a PMT - used by the field service engineer during installation
>> and preventive maintenance - I essentially never use it ('standard'
>> PMT, not GaAsP or GaAs with somewhat higher quantum efficiency ... by
>> the way, STELLARIS HyD S [SiPM, Hy for this product is marketing, not
>> hybrid] has higher quantum efficiency in the blue-green than Leica's
>> HyD (X and R for STELLARIS) or any PMT on the market. I don't know
>> (off hand, nor if published) the QE of the new Nikons NSPARC (photon
>> coutnign 'available soon'?)or Evident's new SilVIR's (one standard aka
>> blue enhanced, other red enhanced, relative PDE curves at
>> https://www.olympus-lifescience.com/en/downloads/detail/?0[downloads][
>> id]=847254126
>> may want to go to
>> https://www.olympus-lifescience.com/en/landing/precision-imaging/ and scroll down to "Download the Brochure" ), slightly higher than GaAsP PMT.
>>
>> Our SP8 web page is
>> http://confocal.jhu.edu/current-equipment/leica-sp8-confocal-microscop
>> e
>>
>> Line averaging takes EXACTLY as long as the same number of line accumulation for the same number of lines.
>>
>> If you want the photon counts line accumulation image sooner, use 1800 Hz (SP8 [zoom 7.5x or higher; use 3x more accumulations to make up for the scan speed], or fastest setting on STELLARIS [been a while since I hosted my last STELLARIS demo so forget its max Hz).
>>
>> I routinely the users to tap the "T" icon and the icon above it (in that order - I've been pressing those buttons while ignoring the official leica names for a long time) to see the image (tip: for multichannel acquisitions, tapping the two icons on the merge image window pane updates them all).
>>
>> LUTs - I explain to new users that the High-Gray-Low lookup table (icon) enables better contrast than practically any color lookup table (blue especially). I am pretty happy with the PC monitor on the SP8 (and think I have its brightness and contrast settings pretty good).
>>
>> PMT gain, part 1 - On our FV3000RS
>> http://confocal.jhu.edu/current-equipment/fv3000 I normally train with
>> HV 500 V and offset 2% or 3% (intensity values ~80 or ~40, 12-bit data
>> galvo mode). The 500 V was the number our O' salesperson said was
>> optimal (2018) - our current salesperson says 500, 550, 600 V all good
>> (offset 2% or 3% works nicely for all three values). Usually 0.01% to
>> 1% laser power is plenty for well labeled fluorescence specimens.
>> Maximum
>> 16 averaging (I prefer 16 line averaging, sequential scan tracks on the FV3000RS).
>>
>> PMT gain, part 2 - as long as the user is not going to wreck the detector(s), they should be able to use whatever gain they want. Rare but occasional are spots of junk (antibody aggregates?) that are not relevant to their biology - using PMT HV that saturates the junk is ok with me if it optimizes the dynamic range of the biology they are trying to acquire data for (if suspect antibody aggregates, I encourage minifuge-ing the antibodies with a cold rotor [store 4 C when not needed] , say 5 minutes at 10,000g, see Konecny et al2024 citation near
>> top of http://confocal.jhu.edu/mctips   [balanced tubes; best values
>> subject to experimentation, and make sure to not resuspend the pellet of junk]).
>>
>> PMT, part 3 - I explain to FV3000RS users that they should adjust the offset so all pixels are above zero (FV3000RS is 12-bit, so range in galvo mode is 0 .. 4095) so that everything in the imaging field of view is "in the data" (in Leica HyD photon counting mode, sometimes some or many non-cell, non-tissue pixels are zero, that just means the HyD's did not detect any photons from those pixels). They can contrast adjust, subtract background, etc, all they want on the PC back in their lab (both Leica, Evident, and the other major make available their confocal software as free viewers). I point out that they should declare all contrast adjustments in their figure captions and/or methods texts.
>>
>> George
>>
>> p.s. I also mention to users in training that SP8 in 16 line * 16 frame accumulation gets them great data (if well labeled specimens), and that the image core finances would love the revenue ($27/hr work hours; $20/hr nights, weekends, holidays), but that their P.I.s probably would prefer more images not better-than-necessary photon counts. Also, to get more than 16*16 = 256 accumulations, a user could acquire Z-stack with very small steps and then "Stack Arithmetic: Sum" (MetaMorph Process menu - Leica LAS X probably has the equivalent - I have the MetaMorph command memorized). For dim (but good antibodies) specimens I encourage users to investigate Alexa Fluor Plus antibodies (Plus is better linker, 3x brighter, 1.3x more expensive) or tyramide signal amplification (several vendors); likely need to use less primary antibody, saving money on the primary antibody, which helps to partially pay for the TSA kit (the core of course loses money per slide if any user takes my advice - hopefully they will be successful and do more imaging experiments).
>>
>> On 1/31/2024 6:48 PM, Cammer, Michael wrote:
>>> *****
>>> To join or leave the confocal microscopy listserv or to change your email address, go to:
>>> https://lists.umn.edu/cgi-bin/wa?SUBED1=confocalmicroscopy&A=1
>>> Post images on http://www.imgur.com and include the link in your posting.
>>> *****
>>>
>>>
>>>
>>> I am writing to seek advice and knowledge regarding properly using HyD detectors on Leica SP8 and Stellaris systems.
>>>
>>> This week I have tried to train & troubleshoot issues on two SP8s and one Stellaris where I found that people had turned the gain up to 80% to 100% and in one case switched the bit depth to 12 bits to get better images.  This appears to be routine common sense.
>>>
>>> In all cases I tried to explain that when the histogram peaks become discrete, turning up gain (and bit depth) adds no information and probably adds more noise than simply adjusting the LUT.
>>>
>>> People are unwilling to wait for line averaging.
>>>
>>> Accumulation to build an image is both slow and challenges their visceral need to see the image immediately.
>>>
>>> When I train people who have never used a confocal before, I find I can guide people to collect images that don't fill the whole 8 bits with low noise high quality images.  I insist they use glow over display mode and focus on the histogram to define the dynamic range which doesn't need to fill all 8 bits.
>>>
>>> But people who have used confocals in the past are highly resistant
>>> to any of this.  They crank the gain until images have saturated
>>> spots and insist that this is how they were trained in the past.
>>> (And don't get me started on the snapshot mode...)
>>>
>>> Is there really a right way?  (Other than photon counting mode.)
>>>
>>> Does routinely using the detectors at 100% damage them?
>>>
>>> Do you have suggestions how to reeducate or should I just relax about this?
>>>
>>> I'm interested in any an all statements of fact and opinion regarding this.
>>>
>>> Best regards-
>>>
>>>
>>>
>>> Michael Cammer, Sr Research Scientist, DART Microscopy Laboratory
>>>
>>> NYU Langone Health, New York, NY  10016
>>>
>>> [log in to unmask]<mailto:[log in to unmask]>
>>>
>>> http://nyulmc.org/micros  http://microscopynotes.com/
>>>
>>> Cell voice (no text) 1-914-309-3270
>>>
>>> Office 1-646-501-0567
> VARNING: Klicka inte på länkar och öppna inte bilagor om du inte känner igen avsändaren och vet att innehållet är säkert.
> CAUTION: Do not click on links or open attachments unless you recognise the sender and know the content is safe.
>
>
>
>
>
>
>
>
> När du har kontakt med oss på Uppsala universitet med e-post så innebär det att vi behandlar dina personuppgifter. För att läsa mer om hur vi gör det kan du läsa här: http://www.uu.se/om-uu/dataskydd-personuppgifter/
>
> E-mailing Uppsala University means that we will process your personal data. For more information on how this is performed, please read here: http://www.uu.se/en/about-uu/data-protection-policy