How to import United Imaging MRS data into Gannet?

MRS Software Gannet
1

Hello Gannet developers and community,

I’m seeking guidance on importing MRS data acquired from United Imaging equipment into Gannet for processing.

Our setup:

  • Scanner: United Imaging uMR890
  • Sequence: MEGA-PRESS
  • Output formats available: .dcm, .raw, and .dat files

Issue: I’ve attempted to load both the .dcm and .raw files into Gannet, but neither format seems to be recognized or successfully imported. The standard loading functions don’t appear to support these file formats from United Imaging systems.

Questions:

  1. Is there currently any support for United Imaging data formats in Gannet?
  2. If not directly supported, are there any recommended conversion tools or workflows to make these data compatible with Gannet?
  3. Has anyone in the community successfully processed United Imaging MRS data with Gannet, and if so, could you share your approach?

Any suggestions or guidance would be greatly appreciated. I’m happy to provide additional information about the data structure or acquisition parameters if that would be helpful.

Thank you in advance for your assistance!

2

Hi @hjt,

Currently, Gannet does not support data from United Imaging (UI) scanners, nor am I aware of any conversion tools compatible with such data. However, I am happy to develop the necessary code to load UI MRS files into Gannet.

I’m a bit surprised that the .dcm files don’t load, as DICOM is a standardized format supported by Gannet. It’s possible that United Imaging uses a non-standard variant of DICOM for their MRS data, which could explain the issue.

We also support .dat files, but those are specific to Siemens TWIX files. UI might use the same extension but for a completely different format.

Could you please share some example datasets exported from UI scanners in all available formats? Phantom data would be fine. You can reach me at mam4041@med.cornell.edu.

Lastly, whichever format you choose, I recommend selecting one that preserves the data in the rawest form possible (i.e., with minimal preprocessing such as coil combination or pre-averaging).

Mark