[3dem] Image processing from the gold grids

Robert McLeod robert.mcleod at unibas.ch
Fri Jun 10 01:42:03 PDT 2016

Hi All,

I've only taken a brief look at the Ultrafoil grids but one problem I did notice was the appearance of diffraction contrast from the Au in the middle of empty holes.  Is that what's being discussed here? Basically due to the large amount of defocus used in the cryo-TEM you have a detector plane that is in-between the image and diffraction planes in the instrument.  If the edges of the beam touch the Au support (as they probably should), you can see dark-field images of the crystallites in the center of the hole. 

At 300 keV, the [111] Au spacing has 2*theta_Bragg = 8.4 mrad, and then the next ring at [211] is 11.9 mrad. I.e.

2*theta_bragg( Au[111] ) = 8.4 mrad = 2.35 Angstrom
2*theta_bragg( Au[211] ) = 11.9 mrad = 1.67 Angstrom

To compare the objective apertures on our Titan correspond to

OA(30 um) = 4.3 mrad = 4.6 Angstrom (too small)
OA(70 um) = 9.7 mrad = 2.03 Angstrom (about right)
OA(100 um) = 13.6 mrad = 1.45 Angstrom (too big, let's in higher order spots)

So the 70 um aperture is the best choice but it still lets the first Au spacing through.  A 58 um aperture would be about perfect to completely cut-off all the crystalline gold signal (and limit your structure resolution to a maximum of 2.5 Angstrom).    

I don't know nor want to derive the exact formula for how camera length scales with defocus but for parallel illumination it should scale linearly with defocus (and also be a function of wavelength).  I.e. 

R = (2*theta_bragg) * L( defocus, wavelength )

It would be easy to measure empirically.  Other things that would help: less defocus, good objective aperture centering, bigger Ultrafoil holes (although this may reduce the ability of the Au to reduce the charge), and a condenser aperture and illumination fit to the hole size.  


Robert McLeod, Ph.D.
Center for Cellular Imaging and Nano Analytics (C-CINA)
Biozentrum der Universität Basel
Mattenstrasse 26, 4058 Basel
Office: +41.061.387.3225
robert.mcleod at unibas.ch<mailto:robert.mcleod at unibas.ch>
robert.mcleod at bsse.ethz.ch<mailto:robert.mcleod at ethz.ch>
robbmcleod at gmail.com<mailto:robbmcleod at gmail.com>
From: 3dem [3dem-bounces at ncmir.ucsd.edu] on behalf of Alushin, Gregory (NIH/NHLBI) [E] [gregory.alushin at nih.gov]
Sent: Thursday, June 09, 2016 8:59 PM
To: Reza Khayat; Xiaodi Yu; 3dem at ncmir.ucsd.edu
Subject: Re: [3dem] Image processing from the gold grids

Hi Xiaodi and Sharon-

Here is the script:


It uses the EMAN2 image handling library, scipy, and numpy (hopefully you have these installed), and should run very fast even for fairly large images.  Let me know if you have any problems.

Also, I did not try CTFFIND4 at the time.  If it can handle gold-containing images as mentioned, it is almost ceratinly a better option than using this program.  It could nevertheless be helpful just for looking at the images without funky contrast.


Gregory M. Alushin, Ph.D.
Early Independent Scientist
Laboratory of Macromolecular Interactions
Cell Biology and Physiology Center, NHLBI
Building 50, Room 4140
T: (301)-496-6530
E: alushing at mail.nih.gov

From: Reza Khayat [rkhayat at ccny.cuny.edu]
Sent: Thursday, June 09, 2016 2:54 PM
To: Xiaodi Yu; Alushin, Gregory (NIH/NHLBI) [E]; 3dem at ncmir.ucsd.edu
Subject: RE: [3dem] Image processing from the gold grids


The latest version of CTFFIND 4 can deal with the mentioned CTF problems –at least in my hands it does.

Best wishes,

Reza Khayat, PhD
Assistant Professor
Department of Chemistry
City College of New York
85 Saint Nicholas Terrace, CDI 2.318
New York, NY 10031

From: 3dem [mailto:3dem-bounces at ncmir.ucsd.edu] On Behalf Of Xiaodi Yu
Sent: Thursday, June 09, 2016 2:48 PM
To: Alushin, Gregory (NIH/NHLBI) [E] <gregory.alushin at nih.gov>; 3dem at ncmir.ucsd.edu
Subject: Re: [3dem] Image processing from the gold grids

Hi Greg Alushin:

That would be very useful for us. Do you mind to share the link of the program? I would like to have a test for my images.

Thanks a lot.


From: 3dem <3dem-bounces at ncmir.ucsd.edu<UrlBlockedError.aspx>> on behalf of Alushin, Gregory (NIH/NHLBI) [E] <gregory.alushin at nih.gov<UrlBlockedError.aspx>>
Sent: Thursday, June 9, 2016 11:25:17 AM
To: 3dem at ncmir.ucsd.edu<UrlBlockedError.aspx>
Subject: Re: [3dem] Image processing from the gold grids

Hi Xiaodi-

Following up on Bob's comment, if you do have gold occupying a significant portion of your image (e.g. the edge of the hole), it can give you problems with image normalization and CTF estimation with standard practices (programs like CTFFIND).  We once collected a dataset where we had enough images of this type that I wrote a small program to try and salvage them.  It finds the mean and stdev of the pixel values in a central 1024 region of the image (which is assumed to be free of gold), then "removes" the gold by replacing pixels more than a certain stdev threshhold with random noise from a distribution that matches the statistics of the central 1024 region.  We were able to get OK CTF estimates from these images and they were good for picking particles.  Obviously you would want to actually extract particles from the original, unmodified images.

If you (or anyone else on this list) would find this useful, I'll put up a download link.

Good luck!
-Greg Alushin

On Thu, Jun 9, 2016 at 9:08 AM, Grassucci, Robert <rg2502 at cumc.columbia.edu<UrlBlockedError.aspx>> wrote:
Dear Xiaodi,
What you may be seeing is normalization/display  issue if the gold is a significant portion of your image.  Of course I am assuming that your ice is the same thickness in both types of grids.  In our hands the holey gold grids do seem to require a slightly longer blot time so that could also be the problem.  We have also noticed some gold particles on both the commercial and home-made holey gold grids but as long as it does not significantly alter the overall standard deviation it is not a problem .  I hope that helps you.

From: 3dem [mailto:3dem-bounces at ncmir.ucsd.edu<UrlBlockedError.aspx>] On Behalf Of Xiaodi Yu
Sent: Wednesday, June 08, 2016 11:27 PM
To: 3dem at ncmir.ucsd.edu<UrlBlockedError.aspx>
Subject: [3dem] Image processing from the gold grids

Dear All:

We have one dataset collected from the gold grids. We feel that the particle contrast from the gold grids is weaker comparing with the one from the regular carbon grids. We do find some gold particles in the images. So we wonder is it possible that these gold particles affect the image normalization/contrast? And is any prefilter or re-adjust the normalization of the images needed before we process the dataset? The last question is comparing with the carbon grids, do the gold grids prefer longer blotting time in order to get the thin ice?

This is the first time we collected the data from the gold grids. So any suggestion is welcome. Thank you in advanced.

Best ,


3dem mailing list
3dem at ncmir.ucsd.edu<UrlBlockedError.aspx>

Gregory M. Alushin, Ph.D.
Early Independent Scientist
Laboratory of Macromolecular Interactions
Cell Biology and Physiology Center, NHLBI
Building 50, Room 4140
T: (301)-496-6530
E: alushing at mail.nih.gov<UrlBlockedError.aspx>

More information about the 3dem mailing list