Reconstructions from 40+ implementations of neuron tracing algorithms on the gold166 datasets. The reconstructions were obtained using different computers at different time for reproducibility-test and optimization of all these algorithms. The data analyses are still ongoing (as of Oct 2016).

• "optimization_no__reconstructions_for_gold166_titan_edison_identical_2407.tar.gz" -- 2407 exactly-matching reconstructions before optimizing the "parameters" of different neuron tracing algorithms (the matching set was obtaining by comparing two sets independently generated on TITAN @ Oak Ridge National Lab and EDISON @ Lawrence Berkeley National Lab supercomputers about the same time)
• "optimization_yes__20161101_reconstruction_for_gold166_rhea_5571.tar.gz" -- 5571 reconstructions after optimization and merging of the results of different "parameter-optimized" neuron tracing algorithms

For faster download, these two tar balls may also be downloaded from Oak Ridge National Lab's BigNeuron site ( http://bigneuron.ornl.gov/data/optimization_no__reconstructions_for_gold166_titan_edison_identical_2407.tar.gz and http://bigneuron.ornl.gov/data/optimization_yes__20161101_reconstruction_for_gold166_rhea_5571.tar.gz )

Use of this release and datasets in it requires that you cite this work appropriately. Please cite the BigNeuron project (http://bigneuron.org) and the following publication:

Peng, H., Hawrylycz, M., Roskams, J., Hill, S., Spruston, N., Meijering, E., Ascoli, G.A (2015) "BigNeuron: large-scale 3D neuron reconstruction from optical microscopy images," Neuron, DOI: http://dx.doi.org/10.1016/j.neuron.2015.06.036, Vol. 87, No. 2, pp. 252-256, 2015. ( http://www.cell.com/neuron/abstract/S0896-6273%2815%2900599-1 )

The respective image datasets of these neurons can be downloaded from the previous release ( https://github.com/BigNeuron/Data/releases/tag/Gold166_v1 )

This release contains 166 BigNeuron neuron datasets with "gold" standard reconstructions and respective raw image stacks. These image datasets were originally contributed by a number of labs around the world and then standardized and reconstructions by 6~7 annotators during the BigNeuron Annotation workshop held at Allen Institute for Brain Science, Seattle, in June 2015.

Distributed download links are shown below. Please click the one that is geographically close to your physical location to download:

• Asia/Singapore (A*Star): http://web.bii.a-star.edu.sg/bigneuron/gold166.zip
• Europe (Blue Brain project): https://zenodo.org/record/168168/files/gold166.zip
  Please, note that Windows users who expect a standard zip file will see the message "zip ... is invalid". As it is zipped from Linux, it can be unzipped from a unix shell like WSL or Linux itself with the 'zip' program.

Use of this release and datasets in it requires that you cite this work appropriately. Please cite the BigNeuron project (http://bigneuron.org) and the following publication:

Peng, H., Hawrylycz, M., Roskams, J., Hill, S., Spruston, N., Meijering, E., Ascoli, G.A (2015) "BigNeuron: large-scale 3D neuron reconstruction from optical microscopy images," Neuron, DOI: http://dx.doi.org/10.1016/j.neuron.2015.06.036, Vol. 87, No. 2, pp. 252-256, 2015. ( http://www.cell.com/neuron/abstract/S0896-6273%2815%2900599-1 )

The bench testing reconstructions can be downloaded from the data release here ( https://github.com/BigNeuron/Data/releases/tag/gold166_bt_v1.0 )

These datasets were originally contributed by http://www.flycircuit.tw/ in their form of Zeiss LSM files, and had been then converted to a unified and simpler format (8bit Vaa3D's raw file, see http://vaa3d.org) to data IO and other uses.

To use the data, simply unzip it and there will be a folder containing 2000 files. Each such file can be further unzipped to be used.

Note that these image datasets are relatively sparse, thus most of the image content is black. Also there has been substantial gaps between portions of neurites, therefore for successful tracing, one should consider preprocessing using a simple Gaussian filter (such as with the kernel 7x7x5 voxels), or use the anisotropic diffusion filter. Check the BigNeuron data preprocessing page (https://github.com/BigNeuron/BigNeuron-Wiki/wiki/Image-Preprocessing) for more details.

The metadata for this dataset should be referred to this paper (http://www.sciencedirect.com/science/article/pii/S0960982210015228), which stated that "The following settings were used in image acquisition: scanning speed 7, resolution 1024 × 1024, line average four times, zoom 0.7, and optical slice 2 μm for 20× objectives and 1 μm for 40× objectives, making the image stack composed of about 60 to 70 serial images under 20× objectives and 120 to 140 serial images under 40× objectives. The corrected voxel size of x:y:z is 0.32 × 0.32 × 1 μm.".

To use the data, you will have to agree with the license term at http://alleninstitute.org/bigneuron/participate/ .

Use of this release and datasets in it requires that you cite this work appropriately. Please cite the BigNeuron project (http://bigneuron.org) and the following publication:

Peng, H., Hawrylycz, M., Roskams, J., Hill, S., Spruston, N., Meijering, E., Ascoli, G.A (2015) "BigNeuron: large-scale 3D neuron reconstruction from optical microscopy images," Neuron, DOI: http://dx.doi.org/10.1016/j.neuron.2015.06.036, Vol. 87, No. 2, pp. 252-256, 2015. ( http://www.cell.com/neuron/abstract/S0896-6273%2815%2900599-1 )