Neuron Annotation Data

The first set of BigNeuron Annotated neuron reconstructions, resulted from the June 2015 data annotation workshop at Seattle, can be downloaded from

https://www.dropbox.com/sh/c7tttjv0vicgdo5/AADma_T9New3uGcIA6ofEpWZa?dl=0

The total size of these data is about 19GB (of 79 neurons). The users should read the README.txt file in the folder to understand what have been included and how to open each set of image (in the .v3dpbd format to save space) and reconstruction data (.swc file). The same information is enclosed below as well.

Due to the "large" size of these shared data, a user might have to purchase or activate a large Dropbox file sharing account from Dropbox.

###Contents###

• Basic information
• Neuron radius
• Neuron morphology sorting
• Neuron node retyping
• Citations/references
• Questions/help

##Basic information##

This folder includes 79 neuron reconstructions from 78 images (one mouse neuron contains 2 neurons), produced in the BigNeuron (http://bigneuron.org) Neuron Annotation Workshop at Allen Institute, Seattle, June 15-17, 2015.

Each reconstruction entry was manually validated by 6 or 7 annotators working collaboratively. Each such reconstruction, although might not still be be perfectly accurate, does reflect the best possible reconstruction given the limited time and resource that would be realistic for reconstructing the 3D morphology from the respective image.

Each reconstruction entry comes with 3 files: one 3D image stack file (in the Vaa3D’s .v3draw or .v3dpbd format), one neuron reconstruction file (in the .swc format), and one linker file (in the Vaa3D’s .ano format). Each entry can be open conveniently by dragging and dropping the linker file (.ano) into the main window of Vaa3D software (http://vaa3d.org). A 3D viewer will be opened automatically to overlay the 3D morphology of the neuron on top of the 3D display image content. You can then recolor the neuron, or image, and use many other powerful features of Vaa3D or your own methods to observe or manipulate the data.

The metadata of the respective image file(s) are also provided if possible, either in the form of an Excel file or a readme file under each major folder named after “checked…”. When such information is not available, some basic information of datasets can be directly seen from the folder(s)’ name(s). We will also add into these folders as much as possible such metadata when such information has been provided by the respective data contributors.

These datasets have been created with the intention to facilitate the fine-tuning of the ported neuron reconstruction algorithms in the BigNeuron projects. It can also be used for other purposes of neuron reconstruction studies.

The use of these data presented here is constrained by the Rules to Participate the BigNeuron project, as documented at the BigNeuron website http://alleninstitute.org/bigneuron/terms/.

##Neuron radius##

During the manual validation process, the neuron reconstruction nodes’ radius information was not validated or intentionally calculated. In most cases, the radius information has been calculated but occasionally it was not. Therefore, for the datasets released here, the radius (information recorded in the swc files should be used with caution. If the values are too small (e.g. 0 or 1), only the skeletonization information recorded in the swc file should be used to optimize an tracing algorithm.

##Neuron morphology sorting##

For the reconstruction files (.swc files), when they were produced using Vaa3D, they might have been put into a way that a reconstruction node with a smaller indexing number (i.e. the first column in a sac file) might not appear before another node with a greater indexing number. This is completely fine as SWC format essentially can be used to code any directed graph. However, some older neuron analysis software might have additional constraints on the neuron node order, thus might not be able to read the files correctly.

This can be solved by running a simple "swc_sorting" function/plugin in Vaa3D, to produce a sorted swc file. For convenience, we have included here all sorted versions of the reconstructions in a separate folder “sorted_for_othersoftwaretoload” (just unzip the zip-ball file "sorted_for_othersoftwaretoload.zip"), where both the unsorted and sorted versions are kept and the linker files that can be used to open any such pairs of files by Vaa3D immediately are also included in the folder “ano_linkers_to_compare_original_vs_sorted”.

The following is the script to do such sorting:

cd copies_of_final_annotations/ for i in $(ls *.swc); do echo $i; v3d -x sort -f sort_swc -i $i -o sorted_$i; mv sorted_$i ../sorted_for_other_softwarepackages/. echo "SWCFILE=../copies_of_final_annotations/$i" > $i.ano; echo "SWCFILE=../sorted_for_other_softwarepackages/sorted_$i" >> $i.ano; mv $i.ano ../ano_linkers_to_compare_original_vs_sorted/. done;

The new swc files in the folder “sorted_for_other_softwarepackages” can be loaded using software packages other than Vaa3D.

##Neuron node retyping##

In swc files, the neuron type (second column) indicates the basic type of a node. Normally the value should be 1 for root node, 3 for dendrite node and 2 for axon node (http://research.mssm.edu/cnic/swc.html), etc. Different "node type" is also associated with different colors for display. However, during the neuron editing process, this field is often used to produce a convenient coloring scheme to contrast different neuron segments; in such cases, the "node type" does not have a biological meaning.

Since the original reconstructions validated by human annotators may not have consistent node type definitions, to avoid confusion we have also provided the following zipped file "retyped_sorted_for_othersoftwaretoload.zip" to contain fixed node types for all files in the zip file "sorted_for_othersoftwaretoload.zip". For simplicity, we provide only two node types, 1 for root node and 3 for all others.

It should be noted for the skeletonization purpose of the neuron tracing, the node type is not essential. Thus the node retyping should not influence the skeletonization accuracy.

##Citations/references##

*** Citations to the following two BigNeuron papers as well as the website http://bigneuron.org are required if these data have been used in your study.

[1] Peng, H., Hawrylycz, M., Roskams, J., Hill, S., Spruston, N., Meijering, E., Ascoli, G.A, "BigNeuron: large-scale 3D neuron reconstruction from optical microscopy images,” Neuron, 87(2):252-256, 2015. (DOI: http://dx.doi.org/10.1016/j.neuron.2015.06.036)

[2] Peng, H., Meijering, E., Ascoli, G., "From DIADEM to BigNeuron,” NeuroInformatics, Vol. 13(3):259-260, 2015. (DOI: 10.1007/s12021-015-9270-9)

##Questions/help##

For help, send email query to bigneuronproject@gmail.com or post the question to the Vaa3D Discussion Forum at NITRC https://www.nitrc.org/forum/forum.php?forum_id=1553 .