MRIReco.jl Developers/Users at ISMRM

[alexjaffray]

Hi all,

I think the upcoming ISMRM might be a good chance for some of the people who develop and/or use this project to meet in-person and discuss ideas. @mrikasper and I would be happy to share our experiences using the package in practice, and learn from what other people have done with it as well!

Thoughts?

[tknopp]

I will unfortunately not be at ISMRM but would of course encourage you to meet and discuss ideas.

[aTrotier]

I will be at ISMRM, we can setup a meeting.

I don't have a lot of things to show because I am still mostly using BART for the CS processing but I'll be happy to discuss about the way to finish that transition.

By the way we can also have a discussion about the creation of an online meeting on a regular basis like the one in gadgetron ?

[alexjaffray]

Let's do that. I am also interested in CS recons in Julia so would be very happy to ask some questions about your experience and see what is needed to move forward.

I think a regular meeting is a good idea, and a starting point for a more structured way forward for the package. Let's definitely discuss it when we meet at the ISMRM.

[mrikasper]

Dear @aTrotier and @alexjaffray,

I will also be at ISMRM and would be interested to meet with MRIReco.jl users. Not sure whether @JakobAsslaender and @JeffFessler have seen this discussion, they might also be around and interested to join?

Just as a first iteration for finding a time and place, @alexjaffray's poster uses MRIReco, so maybe we could meet there (Wed, 17:45-18:45 UTC) or before/after for a beverage and chat?

I also tried to find other Julia/MRIReco abstracts/educationals at ISMRM that would be obvious gathering points, but didn't find anything meaningful. Maybe the keyword "Julia" is too ambiguous :-)

All the best,
Lars

[JeffFessler]

I will plan to come to that poster, and I may bring some group members - thanks for the invite!

[aTrotier]

Seems good to me :)

[alexjaffray]

Fantastic. I am looking forward to it.

[alexjaffray]

It was great to see everyone at our little informal meeting at the ISMRM, it was exciting to hear about what people are working on in Julia!

I think that the big takeaway from the gathering was that it would be useful to set up a recurring meeting for the developers/users of the package. This way we could, as a community, define a roadmap forward for the package and other tools in the broader Julia MR ecosystem.

@tknopp, @migrosser, @mrikasper, @JeffFessler, @JakobAsslaender, @aTrotier: let me know what you think, and how we should possibly schedule such a meeting.

[alexjaffray]

@aTrotier I am also available in July for a meeting, perhaps we can start with something like the last week of July or so, or September as you suggest

[mrikasper]

I would also be available for end of July or early September at that time.

[monreal93]

Hi all,
I would also like to join the meetings, I am available at the end of July and September as well!

[aTrotier]

First thursday of each month -> 2pm CET
Let us know if it is a problem for one of you ?

If it is ok, we schedule a virtual meeting September 1st.

[cncastillo]

Hi! I would love to be in the meetings as well, 2 pm CET on the 1st of September is ok with me 😄. I am the developer of KomaMRI.jl an MRI simulator fully written in Julia (besides the UI). Currently using MRIReco.jl for the reconstructions (as BART would do in JEMRIS).

[aTrotier]

The first meeting will be held :
September 1st -> 2pm CET

Do you have some topics you want to discuss ?

A list with some ideas (some of them might not be relevant):

About MRIReco :

• How we use it
• Splitting package : feedback
• enhancement / future plans ?
• Benchmark reconstruction time (example : reconstruction functions vs building the operator by hand)
• Comparison of CS results with BART : https://atrotier.github.io/EDUC_JULIA_CS_MP2RAGE/

MRI Community :

• EPG package ? @jondeuce
• Dedicated MRI viewer (missing features from ImageViewer.jl )

[aTrotier]

That's the idea of the meeting, If you have real needs for the high level tools we can decide to implement them (and build examples / tests).

For my research, I am mostly using the conversion to RAW / acquistionData and then i'm using the low-level function of MRIReco to perform sense estimation, FT or nufft operator, density compensation.

[mrikasper]

That's great to hear, thank you for the pointers. I also have a lot of blind spots with respect to the Julia ecosystem around MRIReco.

And yes, I would be keen to invest more into high level workflows, tests, and documentation. SMS or 3D stack-of-spirals (tilted hexagonal spirals, to be precise :-)) would be areas of application we have in mind.

[aTrotier]

Additionnal discussion : it would be great to have some kind of consensus about our different packages for the dimension position.

I think MRIReco is based on a 5D array with : x,y,z,echo,channel,NR but some of the function like espirit returns array with other dimensions.

For information BART uses the convention :

enum mri_dims {
	READ_DIM,
	PHS1_DIM,
	PHS2_DIM,
	COIL_DIM,
	MAPS_DIM,
	TE_DIM,
	COEFF_DIM,
	COEFF2_DIM,
	ITER_DIM,
	CSHIFT_DIM,
	TIME_DIM,
	TIME2_DIM,
	LEVEL_DIM,
	SLICE_DIM,
	AVG_DIM,
	BATCH_DIM,
};

[mrikasper]

In principle, I see two ways to go about this: flexible dimensionality with agreed-upon labels (and a label vector, AxisArrays does this, right?) or a fixed high-dimensional data array that all interfaces respect. I find the former solution more elegant (and would assume it's more performant), but the established packages and data formats (e.g., ISMRMRD) I know of in Matlab or C seem to prefer the latter...I guess because a fixed convention is easier to maintain than flexible interfacing. But I don't know whether this is also true for Julia.

[JakobAsslaender]

My guess is that C and Matlab do not have this functionality build in, do they? I would vote for using AxisArrays instead of a high-dimensional array as the latter are very unintuitive. To me, the more important question is where to include this. For the more basic code/packages, I do not think that we need it and I would refrain from making the code more MRI specific than necessary. I would rather include such things only in high-level wrapper packages like MRIReco.jl.

[JeffFessler]

Do we need a zoom URL for the meeting next week?

[aTrotier]

Yes. I can create one but it would be better if someone else is the admin.
(1st day of school for my daughter, she might be around during the meeting...)

[migrosser]

I took the freedom to create a zoom-link for our meeting. Looking forward to seeing you tomorrow.
https://tuhh.zoom.us/j/82596367645?pwd=Zld0NDRXbitPMCt5QzBsSS9zYVNoUT09

[JakobAsslaender]

Sorry, I won't make it to the meeting tomorrow.

[mrikasper]

Hi everyone,

as discussed today, I started our MRIReco Community Meetings discussion in a new thread: #99

Please have a look, I also put some minutes to review and amend.

[aTrotier]

Topics for next meeting

Thu, Oct, 6th at 12:00 UTC

• Benchmark results
• Accuracy of fista (example in benchmark or https://atrotier.github.io/EDUC_JULIA_CS_MP2RAGE/)
• Reconstruction of repetitions : PR reco of repetitions along dims=6 #105
• Conversion from acq to BartIO and BartIO to acq ? (also part of PR reco of repetitions along dims=6 #105)
• Be able to use different trajectories for all dimensions (only implemented for echoes)
• ISMRM abstracts ?

[mrikasper]

I just saw you added topics here, @aTrotier, was wondering whether we should consolidate it in #99

[aTrotier]

Let's do that