The Data Sets

This documentation provides instructions on how to use qualifier phase datasets for DIADEM contestants. Once you have registered your group, dataset download information will be sent to you via E-mail.

There are 5 datasets, all of which have to be reconstructed for the qualifier phase. The 5 datasets are:

1. Cerebellar Climbing Fibers

2. Hippocampal CA3 Interneuron

3. Neocortical Layer 6 Axons

4. Neuromuscular Projection Fibers

5. Olfactory Projection Fibers


General Download Instructions:
(Dataset - specific instructions can be found in the readme files within each dataset description)

1. Download RAR files.

2. Extract RAR files using software of choice. As an example, PeaZip (http://peazip.sourceforge.net/) is free and can be downloaded for Windows and Linux platforms.

3. Each RAR file contains 1 folder with 1 or 2 subfolders: 'Image Stacks' and/or 'Manual Reconstructions.

4. 'Image Stacks' folder will contain 2 subfolders: 'Training Image Stacks' and 'Qualifier Image Stacks', except for datasets where the same image stack(s) are used for both training and qualifier data (e.g. neuromuscular and neocortical datasets).

5. Depending on the dataset, the Training and Qualifier Image Stacks folders will contain 1 or more subfolders each containing an image stack.

6. All images are in TIFF format.

7. The 'Manual Reconstructions' folder contains example manually traced reconstructions to use as templates for the training phase of the corresponding dataset.

8. All manual reconstructions are in non-proprietary SWC format (see FAQ for explanation of SWC format).

9. Freely available ImageJ ( http://rsbweb.nih.gov/ij/) or Neuromantic ( http://www.rdg.ac.uk/neuromantic/) software can be downloaded to open image stacks. Neuromantic can also load SWC files to overlay on corresponding image stacks. Training reconstructions will load in correct XYZ alignment with corresponding image stacks unless otherwise specified in the dataset-specific READMEs (see neuromuscular and neocortical dataset READMEs in particular).

10. SWC file Z values correspond to image sequence number (top image Z = 0, second image = 1, etc) within their associated image stacks. In contrast, X and Y values correspond to pixel values.

11. Neuromantic automatically converts image stacks to grayscale. ImageJ will load image stacks in color but may require changing default computer memory restrictions (see http://imagejdocu.tudor.lu/doku.php?id=faq:technical:how_do_i_increase_the_memory_in_imagej for instructions). Numerous plugins have been created for ImageJ, some of which may aid in visualization and the reconstruction process (see http://rsb.info.nih.gov/ij/plugins).

12. TIFF files contain no inherent information regarding the physical size of the captured region of interest. Therefore, ImageJ and Neuromantic both use pixel-format X and Y coordinates for loaded image stacks and image sequence number within an image stack for Z coordinates (top image Z = 0, next image Z = 1, and so on).

13. Below each dataset-specific README 'Experimental Procedures' section, the Z distance between images when converted to pixels is shown. This information can be used by contestants to compare physical Z distance relative to X and Y values. Like the provided manual reconstructions, final automated reconstruction X and Y values should be in pixels and Z values should be in image sequence number format.

14. The training material, including SWC files and corresponding image stacks, is provided for contestants to compare automation attempts against in order to aid development of accurate entries.

15. These SWC files have been traced manually by the given data provider and serve as the standard for comparing contestants' automatic reconstructions.

16. As subjectivity is an inherent aspect of neuronal reconstruction, subjective reconstruction decisions may be present in these manual traces. These traces serve as gold standards not because they are known to be perfect but because they reach the limits of subjective accuracy.

17. Each SWC within a dataset represents one tree. All trees within a dataset are grouped together to create one score per dataset.

18. Final entries will be scored only on specified neuronal trees for which no manual reconstructions were provided (See the "Manual Reconstruction Starting Coordinates" section in each dataset-specific README for the specific image stacks and structures). These structures have also been manually traced by data providers, but the manual reconstructions are not provided to contestants so that they can be used to score contestant entries.

19. Finished automated traces should have the same name as is found in the 'Manual Reconstruction Coordinates' section in each dataset-specific README. Uploaded traces should also be placed in as many dataset subfolders as required to ensure that dataset files will not get mixed up. For example, an automated trace using the starting coordinates provided for 'NC_13' should be named 'NC_13.swc' and can be placed in a folder named 'Neocortical Layer 6 Axons'. '01' from 'Section 02' of the Hippocampal CA3 Interneuron dataset should be '01.swc' and can be placed in a subfolder named 'Section 02', which in turn is placed in the folder 'Hippocampal CA3 Interneuron'.

20. All uploaded material should be contained in a parent folder that uniquely identifies the contestant. Please be descriptive enough with the parent folder name to ensure that contestants will not be mixed up.

21. All registered contestants will be informed of the specific steps to upload their results via email to the team leader several weeks prior to the close of the contest.

22. Contact information is provided below for any questions not answered in the provided READMEs or on the DIADEM website FAQ.