A lightweight system for managing large neuroimaging datasets in a shared computing environment. There are two main components of somsds:
- A backend SQL database, which may be queried directly using
somsds_sql. - A Perl interface giving access to the most common use cases: archiving and retrieving data.
There is no public documentation yet on how to archive data. In any case, only superusers of the shared computing environment can archive data. Thus most users will only want to know how to retrieve data files from somsds.
somsds should work out of the box on most Linux distributions and on Mac OS X. In theory it should work also under Windows with some minor modifications.
clone git://github.com/germangh/somsds
cd sosmds
./somsds_install.pl
somsds_link2rec [recid] [--]
where [recid] is the ID of the recording from which data should be retrieved. A
series of optional arguments will typically follow indicating the subset of data
that we want to gain access to. For instance, one could retrieve all structural
MRI datasets from recording ssmd using:
somsds_link2rec ssmd --condition rs-ec --modality smri
Apart from condition, modality, and subject there are several other tags
that may be used to filter the set of files to be retrieved, e.g. subject, technique,
device, sex, age, etc. See the command line documentation of script
somsds_link2rec for more information.
To produce a list of the available recordings together with a short description:
somsds_rec_list
To list all valid condition IDs (together with a short description)
for recording ssmd:
somsds_rec_get ssmd condition
In general, the accepted values of tag [tagname] for recording [recid] can be
listed using:
somsds_rec_get [recid] [tagname]
Script somsds_link2rec will not create copies of the relevant data files.
Instead it will create symbolic links with to the actual data files. The main
reasons for this being the desired behaviour are:
-
It is unaffordable (and wasteful) to create copies of the raw data files each time they are retrieved. Raw data files should be (and, in fact, somsds enforces them to be) inmutable so there is no need of keeping multiple copies.
-
Raw data files may have arbitrary file names. In some cases, the names of such files are automatically given by the recording device (e.g. an MR scanner), and those names give no clue of the actual contents of the data file (e.g. Philips MR scanners' physiology files). Lacking standard file naming conventions is a great obstacle to scripting analysis pipelines. By using symbolic links, the somsds system is able to name the symbolic links so that they follow (user configurable) naming conventions.
Released under the Creative Commons Attribution-NonCommercial-ShareAlike licence