Reconstruction outputs multiples images?

[Jody-Todd]

I am very new to Koma, and I am having trouble with my reconstructions. I am using the UI to simulate and reconstruct Pulseq files. I can successfully simulate and reconstruct the preloaded epi.seq file. I have been trying to simulate and reconstruct some .seq files from the Pulseq tutorials, specifically DEMO_gre0.seq. I can simulate this sequence fine, and the output seems reasonable, but for some reason it reconstructs 2 images (neither of great quality, lots of streak artifacts). I expect the output to be one 256x256 image, but instead, I am getting two 256x256 images. I am using the default brain2D_axial.h5 phantom which may not be not dense enough for my desired matrix size? I would like some guidance about how densely sampled the phantom should be relative to the matrix size, but I am not sure of my core problem. Any help would be appreciated.

[cncastillo]

Hi welcome! Thanks for trying Koma out 😄

(neither of great quality, lots of streak artifacts)

Can you attach the images?

Did you write the file using Pulseq 1.5? We are still working on updating the reader #614 for that. If you could use 1.4.x that could fix it. As I don't really understand why it would generate two images. Could you attach the Pulseq file?

I am using the default brain2D_axial.h5 phantom which may not be not dense enough for my desired matrix size?

This could be a problem. You can increase its density with:

phantom = brain_phantom2D(;axis="axial", ss=1) # 1 mm spin separation

The required density depends on your image resolution and therefore the max kspace. You need at least 1 spin per grid point, that example uses a sequence with dx = 1mm, so the code I provided would get you sufficient density.

Not sure if @aTrotier has experience with this example and could have an idea why it is generating two images?

[aTrotier]

I have this issue sometimes but I can't remember how I solved it.

After loading the sequence can you check the value of Nz ?

using KomaMRI
seq = read_seq("DEMO_gre0.seq")
seq.DEF["Nz"]

It should be equal to 1.

if it is the case let's check after the conversion the value of acqData.encodingSize

# Define scanner, object and sequence
sys = Scanner()
obj = brain_phantom2D()

# Define simulation parameters and perform simulation
sim_params = KomaMRICore.default_sim_params() 
raw = simulate(obj, seq, sys; sim_params)

# Auxiliary function for reconstruction
acqData = AcquisitionData(raw)
acqData.encodingSize

Can you share the .seq file ?

Edit :

One thing that really help is to define all the matrix parameters as definition in pulseq file and not let KomaMRI trying to figure out while reading the pulseq file with :

seq.DEF["Nx"] = get(seq.DEF, "Nx", maximum(adc.N for adc = seq.ADC))
seq.DEF["Nz"] = get(seq.DEF, "Nz", length(unique(seq.RF.Δf)))
seq.DEF["Ny"] = get(seq.DEF, "Ny", sum(map(is_ADC_on, seq)) ÷ seq.DEF["Nz"])

[Jody-Todd]

Thank you for your reply. I should be using the latest Pulseq version as I recently downloaded the code from Github.

using KomaMRI
seq = read_seq("DEMO_gre0.seq")
seq.DEF["Nz"] 
seq.DEF["Ny"]
seq.DEF["Nx"]

The last three lines returned 2, 128, & 256 respectively. I set these values to be Nx=256, Ny=256, and Nz=1. This did result in a single image reconstruction, although I still don't know about the image quality. I attached the sequence file I used for the reconsruction (gre.zip) and the resulting image. Thanks for the help :)

[cncastillo]

Yeah, you can modify the phantom to have a higher density in the REPL:

obj_ui[] = brain_phantom2D(;axis="axial", ss=1) # 1 mm spin separation

and that should fix that part. Let me know if the density is not enough.

The correct Nx, Ny and Nz can be included in the [DEFINITIONS] section of the Pulseq file so you don't need to modify them manually.

[Jody-Todd]

This worked!

Thank you for the help!

[cncastillo]

I still see some lines, so you can further increase the density with us (up-sample):

https://juliahealth.org/KomaMRI.jl/dev/reference/2-koma-base/#KomaMRIBase.brain_phantom2D

obj_ui[] = brain_phantom2D(;axis="axial", ss=1, us=2)

But it is looking good 😃 !