SLR pulse

[aTrotier]

What happened?

Hi,

I am playing with the SLR implementation for a course but I get some weird effect when using different SLR filter.

With :ms it seems to work correctly

but with :pm / :min / :max I have some peaks at the beginning and end of the waveform that generates weird results ->

If I remove the first point for :max I get something better (still a lot of ripple but I don't really know what I am suppose to get :

and for :pm without the first and last point it is correct :

The code used to generate that :

function SINC_vs_SLR(;alpha=pi/2,filter_type = :ms)
	sys = Scanner() #hide
	sys.Smax = 50; #hide
		
	Trf = 3.2e-3
	zmax = 2e-2
	fmax = 5e3
	thickness = 0.05

	z = range(4*-zmax, 4*zmax, 400)	

	# Build sinc sequence
	rf_sinc, gz_sinc, gzr_sinc = KomaMRI.PulseDesigner.make_sinc_pulse(alpha; 
						  duration=Trf, 
						  slice_thickness=thickness, 
						  apodization=0.5, 
						  time_bw_product=8, sys)
	seq_sinc = Sequence(sys)
	addblock!(seq_sinc, rf_sinc; z=gz_sinc)
    seq_sinc.DUR[end] = ceil_to_raster(dur(seq_sinc[end], sys), sys.DUR_Δt)
    addblock!(seq_sinc; z=gzr_sinc)
		
	sim_params = Dict{String, Any}("Δt_rf" => Trf / length(seq.RF.A[1]))
	M_sinc =  simulate_slice_profile(seq_sinc; z=z, sim_params)

		## SLR parts

	rf_slr, gz_slr, gzr_slr = KomaMRI.PulseDesigner.make_slr_pulse(alpha; 
											 duration=Trf, 
											 slice_thickness=thickness, 
											 time_bw_product=8,
											 passband_ripple=0.01,
    										 stopband_ripple=0.01, 
											 filter_type=filter_type, sys) #:ms,:pm, :min, :max, :ls
	rf_slr.A=rf_slr.A[2:end-1]
	seq_slr = Sequence(sys)
	addblock!(seq_slr, rf_slr; z=gz_slr)
    seq_slr.DUR[end] = ceil_to_raster(dur(seq_slr[end], sys), sys.DUR_Δt)
    addblock!(seq_slr; z=gzr_slr)

	sim_params = Dict{String, Any}("Δt_rf" => Trf / length(seq.RF.A[1]))
	M_slr =  simulate_slice_profile(seq_slr; z=z, sim_params)

	############# Figure
	f = Figure()
	ax=Axis(f[1,1])
	lines!(ax,abs.(rf_sinc.A),label = "SINC")
	lines!(ax,abs.(rf_slr.A),label = "SLR filter = :$filter_type")
	axislegend(ax)
		
	ax=Axis(f[2,1],xlabel = "z position [mm]")
	lines!(ax,z*1000,abs.(M_sinc.xy),label = "SINC")
	lines!(ax,z*1000,abs.(M_slr.xy),label = "SLR filter = :$filter_type")
	axislegend(ax)
	f
        return f
end

SINC_vs_SLR(alpha = pi/2,filter_type = :max)

Environment

OS arm64-apple-darwin24.0.0
Julia 1.12.5
KomaMRIPlots 0.13.0
KomaMRIFiles 0.11.5
KomaMRI 0.13.1
KomaMRICore 0.12.5
KomaMRIBase 0.13.4

[cncastillo]

Hi! Thanks for trying this out. The funny thing is that the MATLAB version of Pulseq experiences the same behavior (that is expected as we re-implemented it).

I am also not really sure what should happen exactly for each variant. I think it's reasonable to remove the first/last points, we might need to adjust the area of the pulse so the flip angle is preserved.

The problem is that this breaks Pulseq parity, should this be moved as an issue to Pulseq instead?

[aTrotier]

Yes and we might get more answer on pulseq side. (I don't know if the sigpy implementation also give thus kind of results.

For now I will stick to :LS which is working correctly and, according to Pauly courses, is recommended.

[aTrotier]

Hum,

for :min, the rephasing gradient area is smaller (as expected) :

But for max, the value is not a lot different to the sinc gradient :