MRI DEEP RESOLVE BOOST -SPINE - PART1

[@dlin8910]

Thank you for your insights!! I found tse/seperate is generally better and sharper than integrated. The biggest difference is auto calibration scan (ACS) for GRAPPA is taken during the same TR or separately. What you observed in the CSF is probably due to slight saturation effect with shorter TR when using tse/seperate. Integrated mode usually prolong the TR. So maybe you can compare them again with a fixed TR and revisit the difference?

[@Nguyen_MRI]

Thanks for you comment and thoughts. Good idea. Will do next time ☺️

[@aneetaedwards3775]

thanks bac really appreciate

[@Fletch001]

Outstanding information..!

[@SelmawitDesta]

Hey please do more videos

[@Nguyen_MRI]

More Will come

[@calbang]

If youre getting that smearing appearance I would recommend reducing your acceleration. You really only get that if youre source images are too noisy.

[@Nguyen_MRI]

Thanks for the input ☺️

[@aneetaedwards3775]

hi bac hope your doing good, i am working on sola how can i set up restore mag pulse

[@Nguyen_MRI]

hi its under exam card - CONTRAST - COMMON - Wrap-Up Magn - RESTORE

[@local93]

What is about acceleration factroe PE and Reference Lines PE? Is it worth it to change them or it is better leave them as it is

[@Nguyen_MRI]

You can try and play with those settings. But most of the time it's between choosing integrated vs tse separate

[@scottgloverjohnson]

If your ipat is set really high it may be worth it to add some additional reference lines so that the ref scan, whether separate or integrated, will be a bit more robust. If you don't have enough phase encoded data in some form, you can end up with like only 50 lines in the ky direction if you're not careful. Shouldn't affect time too much to increase it. Like Bac says double check what happens when you change it. Very high ipat and very few ref lines starve the reconstruction of real data, but deep resolve will still try to fill the gaps by making stuff up.

[@DvDciti]

I have noticed that when using deep resolve, for example in the TRA sequence of lumbar spine, the voxel size decreases. Does this affect the final image? Is it recommended to use deepResolve if the voxel drops from 5 to 2? Thank you :)

[@Nguyen_MRI]

@@DvDciti hi thanks for comment. I never seen this before. Sounds strange.

[@DvDciti]

@@Nguyen_MRI Yes, I will try to make a video/photo next week and I can show it to you. But it often happens to me with other sequences, when I activate deepResolve the voxel size changes, for example from 0.5x0.5x0.5 to 0.2x0.2x0.2. (I work with Siemens Vida 3T). Thanks for all your ussefull videos, good job!

[@Nguyen_MRI]

@@DvDcitivery interesting. Yes please do so. Thank you. ☺️👍

[@DvDciti]

@@Nguyen_MRI Hello, today I found out why. Activating DeepResolve also activates interpolation, and this decreases the size from e.g. 0.5x0.5x05 to 0.2x0.2x0.2. I see the image with a higher quality of detail compared to the normal size with the interpolation off, but I don't know what the difference is or if it is right or wrong.

[@Nguyen_MRI]

@ good to hear it solved. Check my other videos regarding the topic interpolation. You will see the benefit of using that ☺️👍

[@iansze2652]

Bac great videos as always, I just have a concept question regarding DRB: Why do DRB sequences tend to wrap so much? I notice that at my previous site that used DRB the Phase Oversample is always set at 200% vs 100% or even 150%, or you would get wrapping artifacts. Thanks!

[@Nguyen_MRI]

Correct. Using tse separste mode as far my experience you need to go beyond with phase oversampling as you say. Especially having small fov. It's sense parallel imaging based. Using integrated you can optimise as from what you are used to without Drb. But then sometimes images can turned out bad. Need to find the balance between those two options.

[@brainthesizeofplanet]

No it is mainly not about wrapping artifacts! - it is because of SNR and how DRB works or what u want. You can only get a faster exams by upping P value or reducing NEX. Most ppl also want a higher in plane resolution or thinner slices. So sometimes you reduce NEX to 1 from 2 and depending on the resolution your SNR is tool low even von DRB high. The other scenario is that u already have only 1 NEX ad can't go an lower here but want better and faster images, so u set grappa to 3 or 4 (4 introduces a sig noise and loss of detail) and u raise the resolution - in those cases u mostly get a 10-20% increase in in-plane Resolution. In both scenarios u can end up with not enough SNR in DRB high and u need oversampling to get 5-10% SNR "back" - that is why oversampling with DRB is mostly always higher set than it was with older protocols. This is also one of the main difference GE vs Siemens as GE offers "half nex" for TSE and 0.1 steps for Blade sequences. With GE u have much more flexibility in terms of speed and SNR with DL / AirRecon DL than u get with the Siemens system - we have a Siemens with DRB and a GE with "full" AirDL. Currently AirRecon DL is hugely superior to DRB - cleaner images, less artifacts (pulsation and gibbs artifact gets removed in k-space). Also the recon time faster, Siemens needs 25-30s per series to compute and display and GE only 5-10.

[@Nguyen_MRI]

@@brainthesizeofplanet thanks for the input. 👍. They all got pros and cons in general. I'm not lucky as you to have those two vendors for comparison 🥲

[@calbang]

​@@brainthesizeofplanetmy images on Siemens are reconstructing mid sequence with Deep Resolve.

[@mihaimoldo]

It's mostly because accelerating to 3 or 4 produces more in folding artifacts due to higher acceleration, because DRB sadly doesn't use the Grappa algorithm but the old Sense one. So with pat 4 you might have in folding even with 200%oversampling .​@@brainthesizeofplanet

[@chrisunguez]

I seem to get far more flow artifact from the abdominal aorta in my DRB sagittal images, particularly the STIRS. Sat bands and additional averages do not seem to fix it. Any suggestions?

[@iansze2652]

Add a NEX and compensate for time/signal increase

[@chrisunguez]

@@iansze2652 As mentioned initially, adding an average does not fix it. I haven’t run more than 2 NEX, but adding a third basically negates the time saving benefits of DRB.

[@iansze2652]

@@chrisunguez Hmmm, I assume also you've tried sat bands superior to the stack as well and this also doesn't work...I think all that's left is swapping your phase. I think sometimes adding even one more average in addition to the other fixes mentioned above may actually minimize the artifact/move it off the AOI. It would also depend on whether you are scanning on a 1.5T or a 3T, as a 3T would natrually pick up more motion/breathing/flow artifacts.

[@chrisunguez]

@@iansze2652 Superior sat bands don’t seem to help, either. I hadn’t considered swapping phase, since I’d assumed the breathing artifact A>P or P>A would be just as bad. But thanks, I might try F>H rather than H>F. Not sure if it’ll help, but worth a shot. Also, we only have DRB on our 1.5T at the moment. I’m curious to see how these various DRB artifacts look at a higher field strength.

[@renzoalejos5965]

@@chrisunguezhola, puedes probar cambiando la dirección de fase en F/H, disminuyendo el flip angle que también ayuda a diminuir los artefactos de movimientos y flujo,adicional a ello puedes colocar compensación de flujo en frecuencia y lo más importante… un posicionamiento correcto del paciente para evitar el artefacto de anefacto el cual se produce por recepción de señal fuera del FOV en zonas donde hay bobinas activadas

[@local93]

Btw, is it applicable for joints with flex coils? Or without dedicated coils it is not great idea? Also with 3T mri I honestly don’t see any reasons to improve quality 😂

[@Nguyen_MRI]

Good questions. Applicable with all coils. And for 3T for better images and scan time purposes.

[@brainthesizeofplanet]

uh.... - The Siemens solution is currently inferior to the GE AirRecon DL and that by quite a a bit. With AirRecon DL u can acquire a much higher resolution and thinner slices than before in less time - i.e a knee Exam in 2.5mm and 0.5/0.4 voxel size in less than 7 minutes - the difference is remarkable

[@calbang]

​@@brainthesizeofplanet I've been able to do 0.2mm voxel size in that same time on a Siemens. I think you might need to revisit your sequences. Also, I heard the GE Recon looks really smooth and plastic. Is this still the case?

[@brainthesizeofplanet]

@@calbang If u have the time u can do it - but better check that u are actually scanning in 0.2, most Siemens scanner show the calculated voxel size and not what u are scanning in. .what scanner are u using it on? The denoising if the GE solution is better and yes in high it looks very smooth/plastic - that's the result if the denoising . I doubt many ppl have a GE and Siemens with DL to compare, what I can tell for now is that the GE solution is more flexible, more advanced (3D sequences, blade and tse) and has better denoising - GE is currently running circles around Siemens with DL and they have a advantage of about 3 years in term of development i.e k-space artifact removal (gibbs, pulsation), complete k-space DL (Siemes is partly in 2d postprocessing), faster denoising calculations, a lot more sequences

[@calbang]

​@@brainthesizeofplanet Acquiring voxel

[@aneetaedwards3775]

bac i am unable to see your reply

[@Nguyen_MRI]

hi its under exam card - CONTRAST - COMMON - Wrap-Up Magn - RESTORE