TrakEM2
What is it?
TrakEM2 is an ImageJ plugin for morphological data mining, three-dimensional modeling and image stitching, registration, editing and annotation.Snapshots
Four movies:
 |
 |
See a lot more snapshots and movies.
TrakEM2 video tutorials for users.
Example data sets
|
|
Latest news
Download latest Fiji with TrakEM2 or go to menu "Help - Update Fiji".Pull source code from the git repos.
Stay up to date with Fiji.
TrakEM2 class diagram for programmers.
TrakEM2 video tutorials for users.
2012-04-05 - 0.9g released
- Mipmaps are now by default generated using area downsampling, by which four adjacent pixels are averaged into one pixel of the next image in the pyramid of powers of 2. This method is many times faster than the old default method (gaussian downsampling) while delivering reasonably good results.
- Removed now obsolete methods to generate mipmaps, in particularly nearest-neighbor, bilinear and scale area averaging. All three were implemented using the AWT Toolkit hardware-accelerated methods, but they are considerably slower, and of equal or worse quality, than the new area downsampling method.
- New default file format for storing mipmaps: .rag, which saves the image data as a raw 8-bit array, and the alpha mask of the mipmap, if any, using a gzipped stream attached to the file. This format is many times faster than the old defaults, .jpg, to the point of making many operations interactive (such as adjusting the brightness and contrast of a bunch of selected images).
- Saving mipmaps as JPEG is bit faster.
- The Coordinate Transform of an image is now always paged, rather than stored in RAM. This means each Patch gets a unique ID for its coordinate transform, from which it can reference the file. When saving to XML, the default is to write the coordinate transform into the XML file just like always. There is now the possibility of saving the XML much faster by choosing "Project - Advanced - Save as ... without coordinate transforms". Upon loading this stripped XML file, each image Patch is checked to confirm that the file corresponding to the ID of its coordinate transform exists.
- New data inspection mode "Display - Inspect image mesh triangles". Click and drag over an image to visualize the triangles used to deform the image at that point. These triangles are only present in images that have coordinate transforms.
- Alpha masks now live in files with unique IDs, with the image patch storing the ID to its corresponding file. This system enables undoing and also closing the project without saving, preserving the correct pointers to the correct alpha masks matching with the mipmaps.
- Both alpha masks and coordinate transform files may now become stale after altering them more than once. The new command "Project - Advanced - Delete stale files..." will inspect the currently opened TrakEM2 project and find out which files are stale, and delete them. Be aware of other XML files created with "Save as" from the same project, which may be using some of the alpha mask files and coordinate transform files that the currently opened XML file considers stale! Use this advanced command only if you understand the consequences.
- Fixed export of tiles for CATMAID; in particular, fix respect the alpha masks when exporting directly from the original images.
- Alpha masks are now finally included in the undo commands.
- The command "Import - Import from text file..." now offers the option, in the dialog, to hide the border of the image with an alpha mask.
- Fixed error when exporting color tiles for CATMAID. Thanks to Bjorn Quast for reporting.
- Fixed error generating mipmaps for LUT images (aka ImagePlus.COLOR_256).
- Numerous small edits to improve performance.
Thanks to Stephan Saalfeld for implementing the new area downsampling method for mipmaps and all other related adjustments, and for pushing for the alpha masks to be undoable and never stale. Thanks to Stephan Preibisch for lots of help with implementing integral images in ImgLib. Thanks to Darren Wong for identifying an error in that prevented the Neurite Identifying Tool plugin from running.
2011-04-12 - Open Connectome Project hosts an ssTEM image data set online with CATMAID
The Open Connectome Project is hosting an ssTEM image data set of 1200 serial sections from Bock et al. (Nature 2011). The data was analyzed with TrakEM2, using the Treeline and Connector data types to reconstruct neuronal arborizations and their synapses of the mouse visual cortex.
The image data set is deployed online with CATMAID, the Collaborative Annotation Toolkit for Massive Amouts of Image Data. TrakEM2 is able to export its image data set to CATMAID with "Export - Flat Image" (with "For web"), and Treeline, AreaTree and Connector instances are currently exported with custom scripts.
Read all news here.
Download and Install
Managed installation:- Download latest Fiji with TrakEM2 or go to menu "Help - Update Fiji".
- Stay up to date with Fiji.
- Compile from source -- has lots of .jar file dependencies.
- Requires ImageJ ij143a or above.
What can you do with it?
- Semantic segmentation editor: order segmentations in tree hierarchies, whose template is exportable for reuse in other, comparable projects.
- Model, visualize and export 3D.
- Work from your laptop on your huge, remote image storage.
- Work with an endless number of images, limited only by the hard drive capacity. Dozens of formats supported thanks to LOCI Bioformats and ImageJ.
- Import stacks and even entire grids (montages) of images, automatically stitch them together and homogenize their histograms for best montaging quality.
- Add layers conveniently. A layer represents, for example, one 50 nm section (for TEM) or a confocal section. Each layer has its own Z coordinate and thickness, and contains images, labels, areas, nodes of 3d skeletons, profiles...
- Insert layer sets into layers: so your electron microscopy serial sections can live inside your optical microscopy sections.
- Run any ImageJ plugin on any image.
- Measure everything: areas, volumes, pixel intensities, etc. using both built-in data structures and segmentation types, and standard ImageJ ROIs. And with double dissectors!
- Visualize RGB color channels changing the opacity of each on the fly, non-destructively.
- Annotate images non-destructively with floating text labels, which you can rotate/scale on the fly and display in any color.
- Montage/register/stitch/blend images manually with transparencies, semiautomatically, or fully automatically within and across sections, with translation, rigid, similarity and affine models with automatically extracted SIFT features.
- Correct the lens distortion present in the images, like those generated in transmission electron microscopy.
- Add alpha masks to images using ROIs, for example to split images in two or more parts, or to remove the borders of an image or collection of images.
- Model neuronal arbors with 3D skeletons (with areas or radiuses), and synapses with connectors.
- Undo all steps.
How does it work?
TrakEM2 has been written in Java as an ImageJ plugin, and contains a virtualization engine for seamlessly working with arbitrarily large datasets, limited only by your file storage capacity.Two independent modalities exist: either XML-based projects, working directly with the file system, or database-based projects, working on top of a local or remote PostgreSQL database.
Authors
TrakEM2 is the design child of Rodney Douglas and Albert Cardona, with the help of the entire Institute of Neuroinformatics, University of Zurich / ETH, and has been implemented by Albert Cardona.Stephan Saalfeld and Stephan Preibisch, from Pavel Tomančák's group, have written the libraries responsible for phase-correlation, cross-correlation, scale invariant feature transform, and associated utilities such as proper, gaussian-exact image resizing and automatically multithreaded processing routines that adapt to the machine's available cores.
Cite TrakEM2
If your work uses TrakEM2 or its associated image registration libraries, cite the following papers:
- Albert Cardona, Stephan Saalfeld, Stephan Preibisch, Benjamin Schmid, Anchi Cheng, Jim Pulokas, Pavel Tomančák, and Volker Hartenstein, "An Integrated Micro- and Macroarchitectural Analysis of the Drosophila Brain by Computer-Assisted Serial Section Electron Microscopy," PLoS Biology, 8(10), e1000502 (2010), doi:10.1371/journal.pbio.1000502.
- Stephan Saalfeld, Albert Cardona, Volker Hartenstein, and Pavel Tomančák, "As-rigid-as-possible mosaicking and serial section registration of large ssTEM datasets," Bioinformatics, 26(12), i57-i63 (2010), doi:10.1093/bioinformatics/btq219.
- Schmid B, Schindelin J, Cardona A, Longair M, Heinsenberg M. "A high-level 3D visualization API for Java and ImageJ," BMC Bioinformatics 11:274 (2010). doi:10.1186/1471-2105-11-274.
Who uses TrakEM2
See the list of publications that cite TrakEM2.
Acknowledgements
TrakEM2 would not have been possible without the continuous help from ImageJ's author, Wayne Rasband, and the economic support to Albert Cardona from both Rodney Douglas at INI and Volker Hartenstein (NIH Grant NS054814) at the University of California Los Angeles.We are also particularly grateful to German Koestinger for testing and extensive feedback, David Lawrence for his assistance in PostgreSQL, and also to Nuno da Costa, Rita Bopp, Lauriston Kellaway, John Anderson, Wayne Pereanu and Davi Bock for their input.
3D visualization and mesh-making by marching cubes have been possible thanks to Bene Schmid and Johannes Schindelin from Würzburg.
Image stitching has been made possible thanks to Stephan Preibisch, and the web viewer thanks to Stephan Saalfeld, both from Pavel Tomančák's in the Max Plank Institute for Cell Biology and Genetics, Dresden.