MRI Repository/Current Directories

This page is adapted from the file: /data/graphics/mri/README

Diffusion MRI datasets in $GROOT

The directory, $GROOT/data/diffusion, is a single shared location for storing diffusion MRI datasets as they are given to us by collaborators, datasets in processed forms, and models derived from them.

There is a standard format for paths within this directory, so for future accessibility by your colleagues, please try as much as possible to conform to it. Briefly, here are the path formats for a few different things we store here:

For a one-off patient, the scan's directory is:

 $GROOT/data/diffusion/<site>/<patient>.<scan_date>

For a group of scans, the group directory is:

 $GROOT/data/diffusion/<site>/<group>.<earliest_scan_date>

and each scan's directory within that is:

 <patient>.<scan_date>

Within a scan's directory, there are the following directories:

 as_downloaded --- the master copy of the data as they came
                     from collaborators
 dwis ------------ softlinks to the diffusion-weighted images
                     within "as_downloaded"
 <name>.<date> --- intermediate and processed forms and models
                     generated by the pipeline

• Here's the long-winded explanation:

The top level directories are all named for data collection sites; for example, "edinburgh" and "south_africa". They're in lowercase and use underscores for spaces. Every dataset we get comes from a specific place, and the master copy of each dataset should therefore live in an appropriate place-name directory.

The next level of directories are named for either

A) individual scans (e.g., for pilot data)
B) groups of scans (e.g., a multi-patient study or longitudinal data)

A) Individual Scans Each directory should correspond to one and only one diffusion MRI acquisition, and its name should concisely indicate the type of patient and the date of acquisition. Examples include "hiv.2008.05.09" and "tumor.2007.11.06". Coded patient identifiers are also acceptable but not recommended.

B) Groups of Scans Each directory should correspond to one group, and its name should concisely and descriptively indicate the identity of this group and the FIRST DATE OF ACQUISITION for the group. Note that coded identifiers are NOT APPROPRIATE for this type of directory; each name should be descriptive and in plain English. One example is "cohen_cohen_hiv_study.2007.02.07".

Within each group directory there should be one directory per patient scan. This directory does not necessarily have to be date-stamped and should probably use a coded patient identifier.

Within each individual scan directory should live at least three subdirectories. One, named "as_downloaded", contains exactly the files that the collaborator gave to us via FTP or a CD. Another, named "dwis", contains softlinks to the diffusion-weighted images within the "as_downloaded" directory. All other directories contain data that we generate by processing the contents of the "dwis" directory. The name of each such directory should be concise but should describe the processing that's going on, and should also include the processing date.

Each file that applies to multiple datasets, like gradient tables or patient information databases, should live in the deepest-level directory that contains all of the relevant datasets. Usually this will be a project directory. Deeper directories that like to have such files accessible at their own level should contain softlinks to the generic file.