Is it possible that we need to pass only single flip angle here?

[chaithyagr]

I am wondering if this could be a potential bug. On my side I am trying to simulate an MPRAGE sequence, and I see bizzare results and I wonder if this is the reason, where I see the entire range of flip angles is passed as input, while maube we need to only send flip[n] right?

https://github.com/INFN-MRI/mrsim/blob/ceefc3c9a8b781f0384ffcb885d163ba6a76ce28/src/mrsim/epg/model/mprage.py#L315

[mcencini]

Hi! I have now completed the refactor I started before the workshop, and the syntax has changed a little bit. You can try the new mprage simulator torchsim.mprage_sim() - it is quite simplified, as it only simulates signal from the inversion to the kz = 0 point and neglect T2* (assuming perfect spoiling). It could easily be extended with the transverse component and to account for multiple acquisition segment - let me know if this could be useful!

[mcencini]

P.S: I had to change the name of the package due to conflicts with existing projects in PyPI

[chaithyagr]

Thats awesome. I was internally simulate MPRAGE TR also. Its great you have it!
I see an arg in docs, but its not there in source, do you want me to add it?

torchsim/src/torchsim/models/mprage.py

Lines 73 to 79 in 8de085e

	def set_sequence(
	self,
	TI: float,
	flip: float,
	TRspgr: float,
	nshots: int | npt.ArrayLike,
	):

[mcencini]

That'd be great, thanks! To be meaningful, we should probably add an nreps argument (see torchsim.models.MRFModel), modify nshots to also accept a tuple nshots=(npre, npost), default to (nshots/2, nshots/2) if a scalar is provided (see torchsim.models.MP2RAGEModel) and also account for relaxation after the ADC. Does that make sense?

[chaithyagr]

I just added it, will update my PR