To Jordan and other hcp-users, Here is a response to some questions from Jordan Poppenk that came to [email protected]. Because the responses may be of general interest I am responding via hcp-users. Now that hcp-users is operational, we encourage others to use this as a primary communication mode on issues of this type.
> -----Original Message----- > Sent: Friday, November 09, 2012 12:45 PM > To: [email protected] > Subject: [HCP-Info] Feature Request Submission > From: Jordan Poppenk ([email protected]), Princeton University > Suggestion: Dear HCP, > > First of all, many thanks for undertaking this initiative. Your database will > be a tremendous resource. Having unpacked a sample subject from your initial > release, I have a few questions / requests. Please forgive me if I somehow > missed this information while browsing your documentation. You're welcome, and no problem! Everything you ask about is pertinent, and much of it hasn't been publicly addressed until now. > > 1) I am excited to preview your 7T scans, but I couldn't find information on > your website about when they are expected. I assume they are not represented > in the current releases? Data acquisition for the 200 subjects to be scanned at 7T is currently scheduled to commence in the fall of 2013. The first 7T data release will likely be in the winter of 2014. Please bear in mind that our consortium worked intensively during the first two years of the grant to optimize 3T scanning on our customized Skyra platform, in order that data acquisition could begin this past summer - on schedule! A comparably intensive effort is now underway to optimize 7T scanning, and there are many technical challenges being addressed by our consortium, especially our colleagues at CMRR (UMinn). > > 2) Would it be possible to include Freesurfer output in your db, perhaps as a > separate download (it is already described as a preprocessing step in the > user guide)? There are files that appear FS-like in the current packages, but > the normal folder structure has been eradicated and files are provided Caret > format. Caret is great, but my scripts require the standard Freesurfer files > and folder organization (also, I am already expert with FS but not Caret). > Users can always re-run FS on their local machines to obtain this, but it > seems like a lot of wasted effort and CPU given that you have already run it. Our pipelines generate data in standardized NIFTI (*.nii) and GIFTI (*.gii) format as much as possible. NIFTI and GIFTI are readable across many platforms, including Freesurfer. We also have modified the directory structure of necessity, owing to various non-Freesurfer processing steps that provide clear benefit (e.g., introducing correspondence between left and right hemisphere surfaces; downsampling to coarser meshes that are matched to the spatial resolution of fMRI and connectivity data). However, we understand the potential utility of platform-specific directory structures and file formats, for compatibility with existing pipelines like yours. We will take your request under advisement, and will be interested to hear from others in the neuroimaging community as to what particular Freesurfer data types may be of interest that we don't currently provide. We are not fundamentally averse to this request, but would want to make sure there aren't technical gotcha's lurking under the hood. Realistically, there is little chance that we will be able to act on this prior to the forthcoming February data release, simply because our dedicated informatics team is heavily burdened already with numerous high priority action items. However, we do want to hear from the community as to what issues are considered important, and why, so that we can prioritize based on external feedback as well as internal considerations. CIFTI data format. On a related note, please bear in mind that some of the HCP-generated datasets are in a new 'CIFTI' format, which can include surface vertices and a partial set of subcortical voxels (e.g., only gray matter) in a single file (http://www.nitrc.org/projects/cifti/). This has inherent advantages in visualizing 'dense connectome' datasets and also various other data types (e.g., task-fMRI) on surfaces and volumes concurrently. The CIFTI format was designed in consultation with other major software platforms, with the intent of facilitating cross-platform data compatibility. To date, we are not aware of other platforms besides Connectome Workbench that support the CIFTI format (though it is in the queue for FSL). If others in the neuroimaging community find the CIFTI format useful and would like to be able to make use of it on other platforms, it may be appropriate to suggest this to the relevant software developers. > > 3) If this is not possible, could you provide scripts that can be used to run > preprocessing pipelines A and B based on the files from your database? It's > not clear which inputs are best to use to FS, or how to perform steps as > "Downsampling of 0.7mm T1w to 1mm using splines" described in your > documentation. The preprocessing pipelines will be described in a manuscript that is currently in preparation (Glasser et al.), targeted for submission in December. We intend to release the pipelines once the paper is published. Until then, the pipeline source will not be available (and it may undergo additional refinement - see below), but questions and issues can be discussed on the hcp-users mailing list. > > 4) It is not clear from your documentation *why* it is not recommended to > publish using the data from your first release - all that is mentioned is > that the sample is not complete. If 12 subjects are adequate, is the data > suspect for some reason? There are two reasons why the initial data release is not intended for analyses that will lead immediately to scientific publications. (1) There is a caveat emptor to the minimally preprocessed datasets, because our pipelines have not been completely finalized. Indeed, a recent (Nov 9) email to hcp-users reported a glitch in how the fMRI timeseries data were processed; in the meantime we also made several improvements to the pipelines. Version 2 datasets are currently being generated and will be released in about a week. (2) In addition, many of the subjects are related to one another (twins or non-twin siblings), which can bias the results of some analyses. I hope these responses are helpful to hcp-users in general. David VE _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
