FIRE support

[cncastillo]

Hi, I recently got the FIRE WIP for Siemens. I was wondering if anyone has any experience using MRIReco.jl with FIRE (or Gadgetron, I believe that they should be similar). My understanding is that it uses ISMRMRD, so it theory it should work. I would be nice to make FIRE and MRIReco work ! :) any pointers are appreciated.

[aTrotier]

Hi, I am still using Gadgetron on my scanner.

I did some test with Julia and was able to call it (offline) with the external language interface. No reason it should not works online (I only have to change the external language from matlab to julia in https://github.com/CRMSB/PAPER_MP2RAGE_CS/blob/785f029c8d6d1a7c11b258ba1a3e8a7874c03599/gadgetron-gadgets-MP2RAGE/gadgets_MP2RAGE/config/MP2RAGE_CS_Bucket.xml#L64

You can find a discussion here : https://github.com/orgs/MagneticResonanceImaging/discussions/165

Gadgetron.jl uses differents structures which are sent to the external language :

• Acquisition ~ one profile
• AcquisitionBucket ~ a vector of profile (close to MRD)
• ReconData ~ a k-space
• Waveform
• Image

In order to simplify the work we should wrote some function to convert the first 3 into either a RawAcquisitionData or AcquisitionData. It should be pretty simple.
Not as efficient as having exactly the same structures between Gadgetron and MRIBase but it is a starting point.

You can find more example about the matlab external interface here that I wrote for the Gadgetron Online Course.

I will be to collaborate on that.

[cncastillo]

Hi Thanks for the pointers, I am definitely interested to make this work with Pulseq in Julia. Right now, I need to study a bit how Gadgetron and FIRE work.

Do you recommend watching these? https://www.youtube.com/playlist?list=PL3dITfhovhLW-ySr_KPBS8KKI9tdLjfdR.

Happy to collaborate after I understand the problem a bit better.

[aTrotier]

If it is the first time using gadgetron / fire for sure take a look at the Gadgetron Online Course.

I am not invested as much as I was in the Gadgetron community (our setup works and I don't want to change it) but for what I am seeing most of the modification are linked to the way gadgetron is built (conda environment etc) and a new way to wrote gadgets but you don't need to dive into this most of cases especially if we perform most of the reconstruction steps with MRIReco.

If I have the time, I would like to reimplement this work with pulseq and an online julia reconstruction (rather than matlab + bart). (I already have the Pulseq sequence)

Of note, the default parameter maps used to convert pulseq to MRD need to be changed in order to have access to the signature and sequence name as well as to fix some issues

[alexjaffray]

@nbwuzhe has some experience with working with FIRE, and experience (separately) with Julia, so he may have some thoughts on this