tag:blogger.com,1999:blog-4402160631955197288.post7431411675508823894..comments2024-03-12T19:57:17.818-07:00Comments on practiCal fMRI: the nuts & bolts: Use of split slice GRAPPA (aka Leak Block) for SMS-EPI reconstructionpractiCal fMRIhttp://www.blogger.com/profile/07387300671699742416noreply@blogger.comBlogger14125tag:blogger.com,1999:blog-4402160631955197288.post-39902014127705458232017-07-17T08:37:52.972-07:002017-07-17T08:37:52.972-07:00I'm not a recon expert either, but my understa...I'm not a recon expert either, but my understanding (talking with those who are recon experts) is that the recon itself will perform worse, which makes intuitive sense to me. It is inherently harder to unalias something that is homogeneous with no internal structure. -Mike H.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-53620650159025638542017-07-16T10:34:14.312-07:002017-07-16T10:34:14.312-07:00Hi Mike, do you mean that the recon will perform w...Hi Mike, do you mean that the recon will perform worse for a spherical homogeneous phantom, or that artifacts are easier to see in this case? I'm not completely up-to-date on the literature, and I don't have extensive knowledge on the recon itself, but it's been my assumption that the recon would work independent of the object being imaged (but very dependent on the hardware and parameters). practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-58360836505097091322017-07-13T07:06:49.444-07:002017-07-13T07:06:49.444-07:00Hi.
You hint at this above, but I think it is wort...Hi.<br />You hint at this above, but I think it is worth explicitly making the point that it is inherently more difficult/challenging to unalias the data without artifact in a spherical homogeneous AGAR phantom than in a human. Which is to say that a set of parameters that appear unreasonable when testing in the AGAR phantom, may actually be fine to use in a human.<br /><br />cheers,<br />-Mike HarmsAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-24456026328519770352017-07-13T07:01:24.069-07:002017-07-13T07:01:24.069-07:00Hi. Just a couple commments. For the Aging and D...Hi. Just a couple commments. For the Aging and Development HCP projects we have turned on Leak Block for both fMRI and dMRI. Neither use any in-plane acceleration (no iPAT). [You mentioned above that "For diffusion-weighted imaging, on the other hand, the use of iPAT is all but required in order to keep TE reasonable." That may be true on a Trio, but not on a Prisma!] Also, for HCP-D/A, the rf pulse duration for the fMRI is now 6600 us.<br /><br />cheers,<br />-Mike H.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-56182593452807027952017-01-09T14:08:50.342-08:002017-01-09T14:08:50.342-08:00Hi John, I think you have a choice. From Cauley, u...Hi John, I think you have a choice. From Cauley, use of Leak Block "...can be viewed as a specificity and sensitivity trade-off." Presumably in your stats comparison you found lower false negatives (or better discrimination). What you don't know is the effect of false positives. If you are worried then you could do a correlation using a seed in FFA, say, and use the method of Todd et al. to determine where the false positives will project to. practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-60967828757721581292017-01-09T13:23:37.244-08:002017-01-09T13:23:37.244-08:00Extremely helpful as always Ben!
Question: After...Extremely helpful as always Ben! <br /><br />Question: After reading through this I'm not sure where this leaves those of us using lower MB factors. We're currently scanning at MB3 with 6/8 partial Fourier. Originally we were using Leak Block, but based on your earlier MB posts I tried comparing stats with it turned off in a simple localizer (face/scene), and got slightly better results with no Leak Block. So I have been leaving it off recently. Any thoughts? Thanks!<br /><br />Johnnoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-83655497976754437522017-01-06T16:48:44.353-08:002017-01-06T16:48:44.353-08:00Ah, ok. Yes, MB4 is about as far as we push for DW...Ah, ok. Yes, MB4 is about as far as we push for DWI. You should be able to use RF pulse lengths around 4/8 ms for the 90/180. Actually, we usually ask for flip angles more like 78/160 at 3T when using the 32-channel head coil to improve the B1 distribution across the whole brain (crude but empirically effective--the automatic transmitter adjustment is pretty consistent but biased).Eddie Auerbachhttps://www.blogger.com/profile/11339165302279011568noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-33332341527504475182017-01-06T09:30:48.165-08:002017-01-06T09:30:48.165-08:00Oops, I interpreted your pars as being for DW-MB, ...Oops, I interpreted your pars as being for DW-MB, not BOLD. Sorry! So, as Eddie's comment shows, your pars nearly match the HCP pars. practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-76325063501895315732017-01-06T09:26:50.448-08:002017-01-06T09:26:50.448-08:00Thanks Eddie. Sorry, I confused the discussion by ...Thanks Eddie. Sorry, I confused the discussion by shifting to DW apps. Just to clarify, I also don't now use longer than 8200 us (the precision in the parameter doesn't seem to allow 8192 us specifically) for BOLD excitation, and I've always advocated a low FA so that peak power and SAR are essentially non-issues. The 10240 us refocusing pulse I mentioned in my comment refers to a nominal 180 in DW-MB. That seems to limit the MB to around 5 for DW. Not that I am proposing using an acceleration that high, since 2 or 3 makes the scan duration acceptable. Right now I'm just trying to establish the practical limits before more detailed testing. (I've barely used DW-MB yet.)<br />Cheers!practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-26314907160817457722017-01-06T09:08:58.779-08:002017-01-06T09:08:58.779-08:00Used 8192 us pulse duration, TE 38 ms, partial Fou...Used 8192 us pulse duration, TE 38 ms, partial Fourier 7/8. I am using all coil elements on the 64 head/neck (head elements alone should be 48).<br /><br />Best regardsAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-31395650227764773062017-01-06T07:56:33.201-08:002017-01-06T07:56:33.201-08:00FYI, if you want to manually control the FoV shift...FYI, if you want to manually control the FoV shift factor in the CMRR sequences, you can unlock additional UI options by following the instructions here: https://github.com/CMRR-C2P/MB/wiki. If you can think of any other options you would like to be able to change please ask.<br /><br />Also I would note the standard HCP FMRI protocol is MB8, no iPAT, no LeakBlock, and this has historically performed very well. Practically there shouldn't be a peak power limitation since the TR will be short and you don't want or need a high flip angle. I would recommend selecting the phase scrambling option in any case if peak power is a problem...I don't like to see people using pulses longer than 8192 us due to the sampling constraints of the RF waveform generator (and reduced RF bandwidth in general). HCP uses 7120 us.Eddie Auerbachhttps://www.blogger.com/profile/11339165302279011568noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-50956942514357364712017-01-05T16:38:51.056-08:002017-01-05T16:38:51.056-08:00Hi Peter, what duration refocusing pulse are you u...Hi Peter, what duration refocusing pulse are you using? I just tried MB5 and MB6 and required a duration of 10240 us to avoid exceeding the voltage limit on Tx. 2 mm slices also. So on my Trio the practical limit is already not more than MB6. Also, what is your typical TE without GRAPPA? Do you use partial Fourier?<br /><br />Cheers!<br />practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-40209324287711807762017-01-05T08:50:15.630-08:002017-01-05T08:50:15.630-08:00Thanks for all the important work. On our Skyra wi...Thanks for all the important work. On our Skyra with 64ch Head/Neck MB8 2mm voxel size 72 slices no Grappa with elimination of head motion through special fixation the artifact reduction by enabling Leak Block also seems to work very well. MB10 and 12 though failed in reconstruction, but this was just testing the limits. Really looking forward to data with respect to DWI.<br />Peter<br />Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-36771697365034759322017-01-05T02:19:23.330-08:002017-01-05T02:19:23.330-08:00Nice stuff Ben, plenty to think about as we look t...Nice stuff Ben, plenty to think about as we look to take a step towards multislice EPI here.Paulhttps://www.blogger.com/profile/15493949606449696440noreply@blogger.com