Thank you for your interest in the Collaboratory! We are currently looking to expand the project dataset. If you would like to contribute a dataset, please take a look at the project’s data sharing policy below, and reach out to affectiveneuroimaging@gmail.com for further communication. Please note that you can indicate the level of sharing permission of the dataset you contribute as follows:
- To share publicly (will be listed on the website)
- To share publicly with permission from the contributor(s) (will be listed on the website)
- To share with the consortium members
- To share with the consortium members with permission from the contributor(s)
- Not to share until further notice.
SHARING IMAGES WITH ANiC:
WHAT TO SHARE:
- A good first option is usually to share individual subject-level images from one or more test conditions, so that applying the signature pattern results in one “pattern expression” or “signature response” value per individual per condition. In SPM, these are “beta_**.nii/.img” images from first-level (single-subject) analyses, and in FSL, these are parameter estimate images. If there are 3 conditions and 20 subjects, you would share 3 x 20 = 60 beta images. Once we calculate the signature response for each condition, it is easy to perform contrasts.
- It is also possible to share contrast images (e.g., Condition 1 – Condition 2). We can’t calculate the response for each condition separately in this case, but we can still calculate the difference scores. These images are con_* images in SPM or COPE images in FSL.
WHAT WE WILL DO WITH THE DATA:
We’ll create a script that runs image diagnostics, and visualizes the results of the diagnostics and signature responses (if applies). We will share the script, results, and intermediate files with you, so you can see everything. Then, we can discuss it together and talk about the most sensible ways to proceed. We won’t publish anything without your approval, of course, and see this mainly as fast way of getting an idea of what we’re working with.
HOW TO SHARE, NAMING CONVENTIONS, AND META-DATA:
A number of groups have shared images with us over Dropbox or Google Drive. This is easy if sharing first-level maps as described above. We don’t want to share images with any identifying information (see IRB/ethics below), but these statistic images do not include such identifying information by default. It is easiest if each subject’s data is shared in its own folder (e.g., “subj01” “subj02” etc.). In each subject folder, images can either be named with descriptive names (“pain_high.img”, “pain_low.img”) or with other names (“beta_0037.img”) as long as there is a file with a key.
We can also share files with rsync in Unix, or Globus.
The amount and type of meta-data needed depends on what we want to test. One thing we need is to know which images belong to which individual subjects. The folder structure described above takes care of that. If there are multiple conditions, we need to know what they are. Often, there is person-level meta-data that is useful, too — this includes things like age, overall symptom scores, etc. Sometimes, image-level meta-data may also be useful, including things like pain ratings or emotion ratings corresponding to each condition (within each individual). There are multiple ways of sharing these, but an Excel or text file is easy and convenient. It is also convenient if the ordering of the entries in the meta-dataset (the rows) is the same as what Unix/Mac/Linux would produce if listing the subjects from the command line. E.g., “subj01” “subj02”, “subj10”, “subj11” will appear in that order, so it’s nice to have the entries in the Excel sheet in that order as well. Using names like “subj1” “subj2” is a bit more tricky, as they will list in this order: “subj1” “subj10” “subj2”.
REQUIREMENTS AND FILE FORMATS:
We can easily work with Analyze (.img) and Nifti (.nii) files. AFNI .brik images and BrainVoyager images can be converted, but this requires an extra step.
One basic requirement is that the images are normalized/warped to the same space as the signatures. All our signatures are registered to MNI space. Nonlinear warping to MNI space is recommended; some programs use a linear affine registration, which is less accurate and may be problematic.
IRB/HUMAN SUBJECTS CONSIDERATIONS:
The images described above are usually considered de-identified (there is no way to recover the person’s identity), and are commonly shared in open repositories on the web once papers are published. For example, repositories include OpenFMRI, NeuroVault, and OpenPain, and some journals require that data be published to such a repository upon publication.
De-identified data is not technically considered “human subjects data” in the U.S., so it’s usually given “exempt status” and is not an IRB concern. However, it is always your call on how to proceed, and your IRB’s call on how to handle sharing. If you’re unsure, it’s usually a good idea to check with your local IRB and ask whether it is exempt, and whether they want to review it. In some cases, data may not be considered fully “de-identified” until the master records linking subject IDs (“subj01”, etc) to personal information are destroyed, in which case the data might be considered “coded,” and still shareable with approval, but still classified as “human subjects research.” This is an IRB call, but it’s helpful to know a bit about what the rules are, as IRB members are often themselves uncertain.
This page is under construction. Some information is subject to change. Please reach out to affectiveneuroimaging@gmail.com if you are interested in contributing data.