|
|
|
A tool for automated crystal alignment
on high throughput macromolecular X-ray beamlines
Author:
Bernard Lavault
Page
Updated: July. 2008
C3D is a standalone program doing image processing and calculation of positions necessary for searching, centring and moving backlit cryo-cooled crystal onto a beam. It is designed to be used by a software client which controls specific goniometer hardware.
8. Requirements
How to use the program
The program takes
as input
some
command line arguments and a series of images. C3D
can find the loop/crystal 2D position on one image or the loop/crystal
3D
position from a series of images. The program generates a
result
file including the requested position.
This program does not
include any centring
strategy. Organizing the search for the loop or moving the motors is a
higher-level
job done by the client program.
Background
The first algorithm for C3D: J. Appl. Cryst. (2004). 37, 265-269C3D and visible lightening: Acta Cryst. (2006) D62, 1348-1357
C3D and UV lightening: Acta Cryst. (2006). D62, 253-261
C3D at ESRF: Acta Cryst. (2006). D62, 1162-1169 and last Poster
Last presentation for TID (Bioxhit) training at the ESRF: autocentring_in_pipeline
How to get C3D ?
C3D is distributed through EMBL Enterprise managment (EMBLEM). You can download it from there http://www.embl-em.de/software/index.php. The version 3.8.x is free for academic use. It is provided "AS IS WITHOUT WARRANTY OF ANY KIND", with the hope that it will be useful for other beamlines. For special questions and requests, you can ask Bernard Lavault.C3D Licensing
You must get a license to use C3D. EMBLEM is providing the agreements on his web site according to your intented use (industrial, academic) and your status (Bioxhit/SPINE) member.C3D Technical support
We appreciate your feedback. Send your bug reports, requests and questions to Bernard Lavault.
We provide technical support according to our possibilities with no commitment in terms of delay. Users with free licenses will be handled with less priority.
History of realeases
| version | Content | Comment |
| C3D 3.9.7 | Loop detection is more robust Better removal of light/background artifacts Bug fixes |
Download the full package with new libraries when upgrading from a version < 3.9 |
| C3D 3.9.3 | Centring Playback. Speed improvments. Bug fixes. | New libraries comming with the packet |
| C3D 3.8.x | Loop shape detection. LithoLoop and MicroMount detection. Shadow & ligthening artifacts detection AutoScore between 1-100% compatible XREC. Debug version available "c3d_debug" and compact version (faster): c3d_search. |
WARNING: Introduction of C3D licensing. |
| C3D 3.7 | Loop shape detection prototype. | |
| C3D 3.6 | Introduction of detection using UV. Region of Interest. |
|
| C3D 3.0 | Improvement on crystal detection. Image rotation to adapt to different setups. Server version to improve speed. |
First user Manual. |
| C3D 1.x -> 2.x |
Finds loop tip or crystal based on scoring of objects. | Funded by SPINE. |
| crystaltrace 1.0 | Finds crystal based on correlation of crystal trace between images. | Funded by SPINE. Collaboration with INRA. |
Technologies
The program is a standalone executable. The program has been tested on Windows 2000, XP, Linux Suse 8.2 and RedHat 9.0. All Linux with kernel 2.2.x or 2.4.x should work with glibc (libc6) 2.1.2 or 2.2.5 or with emulation packages.Requirements: hardware and software setup
A setup sample is shown below. You can also find information about this setup at EMBL-Grenoble Instrumentation web Site , especially in the MD2 brochure
|
You need: - a camera perpendicular to the rotation axis- a motorized Centring Table - a stable backlight (below 10% of variations is recommanded) - no front lightening - no overlay on the image - a video server providing lossless compressed images (PNG, PPM, TIFF, BMP) - a zooming device (optical or numerical) - a loop pre-centring script using C3D results - a loop pre-orientation script using C3D results |
 |
Requirements: assumptions concerning the sample
- The lightening must be behind the sample. Front or side lightening is absolutely not supported and leads to mis-detection of loop contour. Front lightening often brings light reflection artifacts. Therefore we think it is not worth to support it. - The sample can be in a nylon loop (any location) or a LithoLoop (Mesh are forbidden !)- Micromount are not supported. However, if the samples are in the positionning hole, One can select loop detection only to do the job, or use UV illumination.
- There must be a single crystal
Hints
The good screening conditions are:- proper flash-freezing (vitrification)
- enough cryo protectant
- liquid not marked by the beam (false crystal edge signature)
- no other artifacts (precipitate...etc...)
- no visual effect (crytal image reflection on the drop surface)
- minimize lens effect with smaller drops and a good pre-orientation script: use at least 6 pictures of the crystal, with 2 showing loop edge and 4 or more (6 is good) near the phi position with maximum loop face visible.
Results images
More simulations...
Speed and timing
C3D needs less than 0.5s to analyze a loop shape and less than 1s to detect a crystal and its shape. The overall centring speed depends mainly on the Centring Server and the speed of the motors. The role of the Server is to mask as much as possible calculation time during motor movements. C3D takes about 100 ms to start.C3D has been mainly tested on MD2(M) diffractometers, including several iterations due to the necessary zooming. The table below gives our average numbers for the MD2, on a Dell Precision 490 with 4 Xeon 1.9 GHz., C3D using only one CPU.
| Step | Duration seen by user (seconds) |
Real (seconds) |
| Search loop by translation and rotation* | 6 | 6 |
| Align tip on optic centre, adjsut contrast* | 1 | 1 |
| Take10 snapshots at different angles | 8 | 8 |
| Detect loop in 2D | 0 | 2 |
| Calculate loop in 3D | 0.1 | 0.1 |
| centre loop | 1 | 1 |
| Auto-zoom* | 4 | 4 |
| Prepare axis orientation for crystal centring * | 2 | 2 |
| Take 8 snapshots at different angles | 6 | 6 |
| Detect crystal in 2D | 2 | 5 |
| Calculate crystal in 3D | 0.1 | 0.1 |
| Crystal centring | 1 | 1 |
| Centring server scripting, miscellaneous | 4 | 4 |
| TOTAL (rounded) | 35 | 40 |
On minidiff (MD2M), the total time seen by the user is typically 50s, due to slower hardware and control software. One can see speed improvement margins. It should be possible to reduce the time by a factor of two with optimization of key elements: taking snapshots (on the fly), zomming (digital), field of view (wide) to avoid to search for the loop .
Statistics and Conclusion
The results with version 3.8 against the test database show 68% of good centrings. 80 % of calculated centre positions are within 30 microns from user clicks. If the freezing is good and the crystal is visible (removal of bad conditions in the test database), C3D statistics raise to 10% of failure and 90% of good and average centrings, with of course a good centring script (see requirements paragraph). If the beam is large (let say above 30 microns square), the average centring can be used and then it is possible to do full automated crystal screening with a good success rate.Acknowledgements
This project is supported by EU programs Bioxhit and initially by SPINE. We also collaborated with the AMIB group from INRA-Jouy-en-Josas (France) to elaborate the first prototype. |
|
 |
