-
Notifications
You must be signed in to change notification settings - Fork 5
Home
This is an open source tool for cellular-scale registartion for multiplexed 3D microscopy experiments as seen in "Simple, scalable proteomic imaging for high-dimensional profiling of intact systems" Murray at al 2015 in Cell.
This is a feature-based method, implementing a combination of 3D Harris Corner Detectors for keypoints and a 3D-modification of SIFT by Scovanner et al, 2012. The final warping is done with a thin plate spline, allowing for subtle corrections in a tissue's morphological change across multiple experiments.
In comparison to other 3D registration methods, this method is computationally intensive and requires involvement for the user (ie, there is no GUI and you will have to format your data before using this system). For fMRI datasets, ANTs by Brian Avants is a mature and well-reputed project. Steven Preibisch has also created a descriptor-based [2D/3D Registration FIJI Plugin] (http://fiji.sc/Descriptor-based_registration_(2d/3d)). We chose a feature-based (rather than information-based) approach to simplify any debugging, and used the non-affine Thin Plate Spline to best model the countless subtle deformities of our tissue throughout the multiple fluorescent staining experiments. For our work published on December 3, 2015, which involved about two dozen teravoxel sized 3D light microscopy datasets that had to be registered at cellular resolution, we custom-designed this software pipeline.
The algorithm was developed through collaboration between [Chung lab] (http://www.chunglab.org) at MIT, Simons Center for Data Analysis and the Seung lab at Princeton University. Detailed experimental protocols and resources regarding the method are available through (www.chunglabresources.org)
Because the size of these 3D datasets can be rather large, this software is presented as a set of scripts that create intermediate files. The default folder structure for this code is:
/Registration/Code/
/Registration/Data/input/
/Registration/Data/output/
These folders can be specified for your needs by the loadExperimentParams.m file.
Each input file is a Data format is also important. This module was developed using 3D 16-bit, monochrome stacks.
The pipeline has six primary components that are encapsulated in two scripts. Each step is covered in the Parts 1-3 pages in the Wiki.
- It is up the user to give a best effort of cropping, zooming and rotating the various experiments by hand, then placing the data into the folder in the right format
- The images volume is partitioned into subvolumes for parallelization and robustness checking in later steps.
- Keypoints and feature descriptors are calculated per subvolume
- Correspondence calculations per subvolume
- Affine transformation are calculated on random subsets of the correspondences
- Non-linear transform applied across the entire sample using the validated correspondences
["Simple, scalable proteomic imaging for high-dimensional profiling of intact systems"] (http://www.sciencedirect.com/science/article/pii/S0092867415015056)
Evan Murray, Jae Hun Cho, Daniel Goodwin, Taeyun Ku, Justin Swaney, Sung-Yon Kim, Heejin Choi Jeong-Yoon Park, Austin Hubbert, Meg McCue, Sara Vassallo Naveed Bakh, Matthew P. Frosch, Van J. Wedeen, Sebastian Seung and Kwanghun Chung
Published in Cell, December 2015
- Ignacio Arganda-Carreras for sharing his wisdom and experience for evaluating tool
- Director Leslie Greengard and the colleagues at [Simons Center for Data Analysis] (https://www.simonsfoundation.org/simons-center-for-data-analysis/)
- Paul Scovanner for sharing his [3D SIFT code] (http://www.cs.ucf.edu/~pscovann/)
- Yang Yang for sharing his [Thin Plate Spline 3D code] (http://www.mathworks.com/matlabcentral/fileexchange/47409-glmdtps-registration-method/content/GLMD_Demo/src/TPS3D.m)
- Fitzgerald J Archibald for sharing his [3D interpolation function] (http://www.mathworks.com/matlabcentral/fileexchange/24315-warping-using-thin-plate-splines/content/tpsWarp/nearestInterp.m)