ClusterAlign Files

ClusterAlign 

Citation:
Shahar Seifer, Michael Elbaum, "ClusterAlign: A fiducial tracking and tilt series alignment tool for thick sample tomography", Biological Imaging, 05 August 2022 (DOI: 10.1017/S2633903X22000071)
link to the article:
https://www.cambridge.org/core/journals/biological-imaging/article/clusteralign-a-fiducial-tracking-and-tilt-series-alignment-tool-for-thick-sample-tomography/69AF0DED063635F5FBFA126E7CF94FB6

1. Purpose of the software

Transmission electron microscopes are often utilized for tomography based on a series of projection images (tilt series), followed by alignment and 3D reconstruction. Numerous software tools are available for these post-acquisition processing steps, including IMOD, bsoft, EMAN2, and others. The mechanism of rotation always introduces some jitter that must be corrected by alignment of the images before (or during) 3D reconstruction. Additionally, electron-optical distortions as well as geometric distortions due to beam exposure may need to be corrected for a successful reconstruction. In order to improve the alignment, nanoparticles are often distributed over the sample to serve as fiducial markers. Despite significant advances in software, precise tracking of fiducials can still be a difficult process to automate. This is especially so in thick samples suitable for STEM tomography, where individual nanoparticles may be shadowed or hidden by the specimen itself and therefore their visibility may differ drastically from one image to the next.  ClusterAlign has been developed based on tests with such difficult datasets. The principle of operation is based on analysis of coherent clusters of fiducial markers rather than individual particles. Moreover, the structure of fiducial clusters is compared between the zero-tilt projection and all other projections without any assumption on correlation between adjacent tilts. Alignment error is mainly translational and thus expected to be corrected by solving for ?x,?y shifts of the image. Given a sufficiently large number of fiducial markers there will be points that share a similar height in z; in this case their relative locations in x,y are predictable according to a rigid body rotation and it is sufficient to track some of the markers uniquely in order to use them as anchors. Note that clusters are defined by a common height with respect to the tilt axis and need not be composed of apparently adjacent fiducials in any given projection; however, the vicinities in z and in x,y are often correlated so a cluster size limit may be imposed. The aim of the software is to optimize translational alignments under conditions of problematic fiducial visibility without manual intervention. Further optimizations such as focused local alignment or global distortion corrections or fixing additional degrees of freedom may be performed in other dedicated software such as tomoalign, EMClarity, or IMOD.
The input of the ClusterAlign software is a tilt series in mrc file format (or tif format) and a text file containing the rotation angles. The output includes the realigned tilt series in mrc format and text files with the fiducials locations and filled-in (“no-gaps”) locations, according to the standard format used in IMOD software. Gaps are filled by fitting to a model of a fixed rotation axis with minimum error, and the fitting errors are reported to a file. Optionally the 3D reconstruction is generated as well.      
 
2. Installation

Windows
In windows, download the zip file found in:
https://sourceforge.net/projects/clusteralign/files/.
Extract the folder to your location of choice and run the setup.exe file. Then continue to install the optional Matlab toolboxes.
The installation makes use of 
Nuget packages of EMGU.CV (https://www.emgu.com/wiki/index.php/Main_Page) wrapper of OpenCV (https://opencv.org/).
Avalonia (https://avaloniaui.net/) for user interface.
DotNet cross-platform framework by Microsoft (https://dotnet.microsoft.com/en-us/).
GPL3 license.
Source code can be found in https://github.com/Pr4Et/ClusterAlign.

See the rest of the user manual in a pdf file within the installtion folder.