[3dem] [ccpem] Adequate particle normalization prior to 2D classification?

[Marin van Heel]

Dear Nancy,

What Sjors is suggesting is a simplified version of what we suggested in:

"A posteriori correction of camera characteristics from large image data 
sets", P. Afanasyev et al, Nature, Scientific Reports.

The camera correction should preferably include the subtraction of the 
average background and the normalization of the standard deviations of 
the pixel vectors, especially if you are operating your DE camera in 
integrating mode. The movie alignments can improve significantly after 
the a posteriori camera correction.The other aspect is the FRC 
validation of the camera correction. The FRC is then not used as a ‘gold 
standard’ 2D resolution criterion but rather as an indicator of the 
independence of different images collected with the same sensor. The FRC 
should not cross the 3-sigma threshold indicating the expected random 
noise fluctuations around FRC=0. For details see the paper.

If you want to test the a posteriori camera correction programs, contact 
michael at imagescience.de <mailto:michael at imagescience.de>!

Hope this helps!

Marin

=====================================

On 28/01/2016 00:48, Nancy Meyer wrote:
> Again, thank you for the reassuring response! Maybe I've lucked out with similar image intensities in previous datasets - I should probably look at a few min/max, mean, and stdev values to convince myself.
>
> I was pretty careful to remove particles on that horiz artefact and if restoring their inclusion is the biggest gain from redoing gain correction, I can maybe let that slide for now. I did just see "A posteriori correction of camera
> characteristics from large image data sets", P. Afanasyev et al, Nature, Scientific Reports, after looking up your recommendation - looks like they've implemented it in Imagic  - wonder if you could do the same correction in just ImageJ...
>
> Thank you for taking the time to look at the images... slowly getting the hang of Relion, but the input is so invaluable!
>
> Best,
> Nancy
>
> ______________________________________
> From: Sjors Scheres [scheres at mrc-lmb.cam.ac.uk]
> Sent: Wednesday, January 27, 2016 1:10 PM
> To: Nancy Meyer
> Cc: ccpem at jiscmail.ac.uk
> Subject: Re: [ccpem] Adequate particle normalization prior to 2D classification?
>
> Hi Nancy,
> The particles look perfectly fine to me. The overall differences in
> greyscale you see is because each of the particles is scaled from white to
> black with their own min and max values, which are susceptible to
> outliers. The mean will probably be perfectly fine. You can get rid of
> this with removing white and black dust, but it will not be necessary to
> do so.
>
> The classes also look quite OK. Just select all the nice ones and proceed
> with another 2D classification or go into 3D already. Some details are
> already coming up, so the 3D structure will look nice.
>
> The micrographs show a sign of a bad gain correction. If you have many of
> them (> several hundreds) you could correct the gain a-posteriori by
> calculating the average of all micrographs and then divide each micrograph
> by that average. This should get rid of the horizontal artefact (as long
> as it is present in all micrographs). That may save some particles that
> lie on that line.
>
> HTH,
> Sjors
>
>
>
>

-- 
================================================================

     Prof Dr Ir Marin van Heel

     Professor of Cryo-EM Data Processing

     Leiden University
     NeCEN Building Room 05.27
     Einsteinweg 55
     2333 CC Leiden
     The Netherlands
      
     Tel. NL: +31(0)715271424 // Mobile NL: +31(0)652736618
     Skype:    Marin.van.Heel
     email:  marin.vanheel(A_T)gmail.com
     and:    mvh.office(A_T)gmail.com

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ncmir.ucsd.edu/pipermail/3dem/attachments/20160130/5674d9b3/attachment.html>