fastMRI

Folder/Filename convention for Brain & DICOM dataset

Hi, I have a question on how the folders are organized and the naming convention for the filenames in the Brain and DICOM datasets.

  1. BRAIN
    The filenames look something like “multicoil_val/file_brain_{AXT1, AXT1PRE, AXT1POST, AXFLAIR}{3 digit number}{7 digit number}.h5”.
    a) What do the 3 digit and 7 digit numbers represent? I noticed that the 3 digit numbers are consecutive but the 7 digit numbers aren’t. Is there any significance to this?
    b) Is there a connection between files that have the same 3 digit number or is the 3 digit + 7 digit number a unique identifier?
    c) Is it safe to assume that the filename encodes the pulse sequence (AXT1, etc)?
    d) Could you please confirm that AXT1 is the same as AXT1 pre?
    e) Is each h5 file stand-alone and unconnected with all other files in this folder?

  2. DICOM
    The filenames look something like this: “fastMRI_brain_DICOM/{12 digit number}/{3 digit number}.dcm”.
    a) Are the 12 digit number individual patients or individual studies?
    b) Does it signify anything that the 3 digit numbers are close together but not consecutive?
    c) Is the 3 digit number describing the relative order of dcm files in the MRI sequence? If not, how does one know the correct dcm sequence in order to collate the DCM files into a 3d volume tensor?
    d) Are all the dcm files taken using the same pulse sequence, and what is that pulse sequence?

Thanks!

Hello @RGG, the 3-digit number for the brain data helps identify the particular variant of acquisition sequence. The 7-digit number is merely an anonymized file identifier.

You can access sequence information from the HDF5 attributes.

AXT1 is sometimes pre and sometimes not. This is an artifact of how the data was acquired.

Each h5 file is standalone.

For the DICOM data, the 12 digit number should be a study.

The 3 digit number identifies files in the study.

The dcm files are taken with different pulse sequences.