Interpretation of '.seq' files is not matching with sequence designed on PyPulseq

[ZemaTimoteo]

What happened?

First, thank you for your nice repository, I found it very useful.

However, I am struggling with an issue regarding the interpretation of '.seq' files. Apparently, KomaMRI interpretates each block of gradient built in PyPulseq, with the need to start and finish at 0.

Here are the full concatenated blocks I want to perform in the simulations. However, as you can see they have unexpected behavior of going to zero between each block of gradients design in the Pypulseq.

Here are an example of block 1 and 2 for Gz:
Block 1 - Koma

Block 1 - Pulseq

Block 2 - Koma

Block 1:2 - Koma

Block 1:2 - Pulseq

I feel like this shouldn't be the case, as I need to simulate the '.seq' and don't want to edit it in the Koma environment as it would make it a different sequence from the one I will design for further scanning.

Should I do something different, or is this something that needs to be update?

Best
ZemaTimoteo

Environment

OS x86_64-w64-mingw32
Julia 1.9.4
KomaMRIPlots 0.7.8
KomaMRI 0.7.5
KomaMRICore 0.7.6

[cncastillo]

Hi!

Thanks for using Koma 😄!

This issue is related to #205. While it looks weird, it is more of a plotting bug, as it has no impact on the simulations. The worst thing that it could do is to add a sample with amplitud 0 and duration 0 to the simulation, which has no effect.

Having said that, I don't think it is a hard one to solve. Could you provide this pulseq file to do some tests? 

In the meantime, we have tools to check what is actually being simulated. Take a look at the results of  plot_seqd(seq) (see it in the docs). It would be good to update Koma anyway, as we recently fixed a bug #315 which could have an impact on the simulation.

Cheers!

[cncastillo]

I believe we found a case where this could affect the simulation, opened #321 to fix it.

[ZemaTimoteo]

Having said that, I don't think it is a hard one to solve. Could you provide this pulseq file to do some tests?

• I attatch the seq file that is give me this trouble. It is a simple MSE sequence.

MSE_test_KomaMRI.zip

• I tried to run the code plot_seqd(seq) but i got the following error message:

`ERROR: MethodError: no method matching discretize(::Sequence; sim_params::Dict{String, Any})

Closest candidates are:
discretize(::Sequence; simParams) got unsupported keyword argument "sim_params"
@ KomaMRICore C:\Users\filia.julia\packages\KomaMRICore\wY77R\src\datatypes\simulation\DiscreteSequence.jl:53`

[beorostica]

Hi @ZemaTimoteo!
Thank you for your MSE sequence. We are going to perform some tests with it 😃.

Can you please update your KomaMRI version? That is why you see ERROR: MethodError: no method matching discretize .... To do this, just type ] up in the REPL.

[cncastillo]

Hi, to give an update on this. What we were doing differently. We hadn't added the heuristic that Pulseq uses for the first and last samples (more info here #321 (comment)). It is an easy fix and should be solved this week.

[cncastillo]

This has been solved in the current dev version of Koma (master). In some cases, the first sample is not exactly the same, but the difference is very small and only happens when the sample duration dt=0, so it had no effect in the simulation. We added tests to confirm this in

KomaMRI.jl/KomaMRIFiles/test/runtests.jl

Lines 53 to 82 in 3d7e92d

	@testitem "Pulseq compat" tags=[:files, :pulseq] begin
	using MAT, KomaMRIBase, Suppressor
	# Aux functions
	inside(x) = x[2:end-1]
	namedtuple(x) = x[:]
	namedtuple(d::Dict) = (; (Symbol(k == "df" ? "Δf" : k) => namedtuple(v) for (k,v) in d)...)
	not_empty = ((ek, ep),) -> !isempty(ep.t)
	# Reading files
	path = joinpath(@__DIR__, "test_files/pulseq_read_comparison")
	pulseq_files = filter(endswith(".seq"), readdir(path)) .|> x -> splitext(x)[1]
	for pulseq_file in pulseq_files
	#@show pulseq_file
	seq_koma = @suppress read_seq("$path/$pulseq_file.seq")
	seq_pulseq = matread("$path/$pulseq_file.mat")["sequence"] .|> namedtuple
	@testset "$pulseq_file" begin
	for i in 1:length(seq_koma)
	blk_koma = get_samples(seq_koma, i)
	blk_pulseq = NamedTuple{keys(blk_koma)}(seq_pulseq[i]) # Reorder keys
	for (ev_koma, ev_pulseq) in Iterators.filter(not_empty, zip(blk_koma, blk_pulseq))
	@test ev_koma.t ≈ ev_pulseq.t
	@test inside(ev_koma.A) ≈ inside(ev_pulseq.A)
	@test first(ev_koma.A) ≈ first(ev_pulseq.A) || ev_koma.t[2] ≈ ev_koma.t[1]
	@test last(ev_koma.A) ≈ last(ev_pulseq.A)
	end
	end
	end
	end
	end

plot_seq

plot_seqd

Please let us know if you encounter any other issues!