Reconstruction for 3D sequences #323
Comments
|
Great that you are using KomaMRI, and thanks for reporting the issue! An immediate work around might be to use:
if using KomaUI:
This is a 3D phantom that is very thin along Z, will take a bit longer to simulate, but less so than the default brain_phantom3D(). ss defaults to 4 for both 2D and 3D. Hope this helps while the issue is investigated! ===================== This unfortunately does not fix the issue. |
cncastillo
changed the title
|
I am trying the seq now. Nice job getting it through github! |
|
This is related to #308, but instead of the "echo dimension," the problem is in the "slice dimension." Right now, we are guessing how to reconstruct stuff (the dimensions Nx, Ny, Nz); a proper way would be to specify the label counters in the "[EXTENSION]" section of the pulseq file. I believe previously, there was a warning that told the user we were doing this. We should put it back (@beorostica). For now, for the echoes, we could add a In the meantime, you could specify the correct sequence dimensions in the REPL before reconstructing. seq_ui[].DEF["Nx"] = 32
seq_ui[].DEF["Ny"] = 32
seq_ui[].DEF["Nz"] = 32I can not test it right now, but I think it should work. For convenience, this also can be specified in the sequence file |
|
This seems to be the issue, the current seq_ui[].DEF below has Nz=1: The Nz, Nx, Ny parts of DEF are generated by KomaMRI(?) and are not in the current seq file: I saw a similar issue with: https://github.com/pulseq/pulseq/blob/master/matlab/demoSeq/writeGradientEcho3D.m but had not gotten around to resolving yet. Will try the above fix, thanks! |
|
Changing the Nx-z parameters in seq.DEF: And running the simulation, results in: which seems to be simulating everything:
however: with Trying to manually set the recon size fails with: Next step is to fill in the encoded size correctly in raw. |
Ok, this should work. So it is an unrelated bug (@beorostica can you create an issue?).
Last time I tried using 3D k-spaces (nodes) with MRIReco it always gave an error, so we are just storing 2D k-spaces in the raw data (or at least that is the default |
|
Manually editing raw.params: gives mixed results: and unfortunately still does not reconstruct: probably because of acq.traj and other incorrect parameters or dimensions of trajectory and other related data? |
|
The data is getting there but in the wrong dimensionality/indexing: |
|
Line 448 suggests it may work if the DEF items are placed in the seq file. Line 456 may be the problem when not present in 3D sequence? I will try the Nx-z DEF items in the seq file.
|
|
This is a test where the Nx-z parameters are in the [DEFINITIONS] block of the seq file: |
|
Here is some more testing: It looks like fixing the parameters is not enough, the kdata is still packed up wrong? |
|
Testing 3D non-Cartesian, apparently has the same issues as 3D Cartesian. |
KomaMRICore/src/rawdata/ISMRMRD.jl definitely is only doing 2D. Is the idea to have it do 3D again, or fool it at least? |
|
Functions so far that handle 2D only at the moment.
|
Draft
Draft
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
What happened?
When using a 3D pulse sequence (MPRAGE) with brain_phantom3D, the recostruction is not performed propperly. Looks like a 3D k-space is considered as a set of 2D k-spaces or something like this. I attach the .seq file as an example (actually, I couldn't attach it as it is, so you will need to remove "txt" after you save it) MPRAGE_brain_25cmfov.seq.txt
Environment
The text was updated successfully, but these errors were encountered: