Review of ISMRMRD
The ISMRM Raw Data (
ISMRMRD) format originates from a
subcommittee formed at the 2013 ISMRM Sedona, Arizona workshop on image
reconstruction. ISMRMRD aims to meet the critical need for a vendor-neutral
(non-proprietary) file format for MR research and development, especially in
light of the widespread push for reproducible research across all disciplines of
biomedical science. Potential users of the domain-specific ISMRMRD format may be
skeptical of the underlying domain-independent Hierarchical Data Format (
HDF5) format, but that should not be a
concern. HDF5 comes from the National Center for Supercomputing Applications
(NCSA). If you use a recent version of MATLAB, you are already using HDF5 when
saving .MAT files from MATLAB. Support for loading HDF5, and hence ISMRMRD, is
widespread with functions/modules in Python, MATLAB, C/C++, Java, Julia, etc.
ISMRMRD-specific interfaces are available for
MATLAB
and
Python. It is
important to note that ISMRMRD files include creation dates for the file and
objects within a file (as do all HDF5 files by default). This causes two ISMRMRD
files containing the exact same data, but created at different times, to have a
different checksum value. To accurately compare two ISMRMRD files, a utility
such as
h5diff should be used.
ISMRMRD can also store reconstructed
spatial domain images. More details of the ISMRM raw data format appear in a
2016 peer reviewed publication in the journal
Magnetic Resonance in
Medicine (
http://dx.doi.org/10.1002/mrm.26089). The
manuscript describes the design strategy, general specifications and some
implementation details of the file format. Table 1 in the paper is particularly
useful because it shows how dimension labels from different vendors are mapped
to the ISMRMRD labels. For example, the “echo” number label is mapped to the
ISMRMRD label named "Contrast". The published paper includes examples from
multiple vendors using Cartesian, spiral and EPI data sets. ISMRMRD supports
storing arbitrary k-space coordinates making it very well suited for
non-Cartesian applications.
In summary, the ISMRMRD
file format has high potential to make a positive impact on the MR research
community. The success of the format will undoubtedly be determined by how
widespread its use becomes among MR researchers. If you believe in open and
reproducible research, then you should strongly consider using ISMRMRD.