FLASH sequence in Jemris

[Nikola Milosevic]

Dear Jemris-Team,

I am trying to simulate a whole-brain image acquisition using the FLASH sequence. To simplify everything related to spoiling, I want to null the transverse magnetization components at the end of each repetition.

Is there some hook in the framework that I can use? Or, do you know if someone has extended it with something like this before?

I would be happy to add the functionality if you think it would be useful.

Thanks,

Nikola

[tony.s...@gmail.com]

Hi Nikola,

as far as I remember, this functionality is not implemented. I typically just set T2<<TR to achieve "perfect spoiling". It might be worth having that functionality in some cases (although results will not be realistic).
Best,

Tony

[Nikola Milosevic]

Hi Tony,

thanks for the reply. I want a contrast that is as close as possible to the real FLASH contrast on human brain tissue. So, I don't think that setting T2<<TR will work for me. 

I'll try to implement the mentioned feature myself, but are you interested in something like this as a contribution to the main project?

I'd like to contribute in general, maybe add something to the documentation (e.g. for MT simulations). Is JEMRIS open to contribution from the outside? Are there any guidelines?

Best wishes,

Nikola

[Lars G. Hanson]

Dear Nikola,
Nice topic, and I hope it is OK to join the discussion. The "perfect spoiling" functionality that you describe would be useful (e.g. to explore effects of imperfect spoiling), but if you want "contrast that is as close as possible to the real FLASH contrast" then you should not use it, since real spoiling is not perfect. Instead, you should apply a spoiler in JEMRIS with an area of the actual spoiler in the scanner sequence that you want to simulate, and add the applied RF spoiling too. And then you need to make sure that you simulate with a sufficiently dense array or isochromats to get trustworthy results.

You may want to reconsider whether actually to simulate whole brain acquisition, since it will be computationally demanding. Typical aspects of interest can be explored with simulations in fewer dimensions, which is normally much preferred.

Kind regards,
Lars

--
You received this message because you are subscribed to the Google Groups "JEMRIS" group.
To unsubscribe from this group and stop receiving emails from it, send an email to jemris+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/jemris/104b7ec2-4186-49e3-8f10-d7e0b1eebc9an%40googlegroups.com.

[Nikola Milosevic]

Dear Lars,

Thank you very much for the feedback. I'm happy if more people join the discussion. 

My supervisor (I'm a student helper) wants to try simulated whole-brain scans (or at least individual slices) for a machine learning application. I can only simulate single slices at the moment. One slice already takes a couple of hours on an HPC cluster.

I have tried pretty much all of the things you have described. The current simulations run with 512 Isochromats "per voxel" (this seems to be a good number from what I see in the results), and I programmed the sequence with gradient and rf spoiling. The problem is that -- as far as I understand it -- rf spoiling relies on diffusion effects. So, without diffusion I get a bad result in terms of spoiling. Simulation of diffusion effects, however, doesn't seem to scale well. Thus, our idea was to try the perfect spoiling approach at least to see how good the final images can get. 

I thought that, if it doesn't produce good enough results, the feature would still be nice to have.

I hope I was able to provide a better perspective on the whole issue :)

Thanks again,

Nikola

[tony.s...@gmail.com]

Hi Nikola, 

Sorry for the late reply. I agree with Lars, would be good to have that functionality to study imperfect spoiling. RF spoiling is independent of diffusion,  it only depends on phase history.  512 spins per voxel is quite a lot - is this really necessary? You could find out by reducing the number until the signal changes significantly. It is however true that quite a number of spins are needed for realistic simulation of intra-voxel dephasing, and this is especially the case for RF spoiling - at least in my experience. If I remember correctly, 64 spins per voxel were sufficient in most cases of RF-spoiled sequences. (Diffusion, as it is implemented in JEMRIS, needs even more spins per voxel for realistic simulations. But this is another topic.)

Best wishes,

Tony

[Nikola Milosevic]

Dear Tony,

I think I confused diffusion with intra-voxel dephasing. 

Thanks for the feedback!

Cheers,

Nikola

[Yonghao Gao]

Dear Tony,

 

I'm working on a project to use gradient spoiling, but I don't how to simulate the real spins in a voxel. I see that you have simulated 512 Isochromats "per voxel" , do you have any open-source code that I can use for reference. Thank you very much.

Best,

Yonghao