Open whitecraine opened 3 weeks ago
mharms
commented
3 weeks ago FYI: A copy of CMRR_MB_Slice_Order.pdf can be found at the re-located HCP Wiki: https://wiki.humanconnectome.org/docs/HCP%20fMRI%20slice-timing%20acquisition%20parameters.html
Can you please clarify what you mean by "temporal instabilities"? (Spin-history, steady-state issues leading to a visible banding artifact are not "temporal instabilities".)
whitecraine
commented
3 weeks ago Hi Michael!
I'll see if I can make a gif or video upload. But in short, if you acquire a time series of b=0 images, you will see low-intensity slice banding that cycles through all the slices periodically across the time series.
Joseph
On Thu, Oct 10, 2024 at 9:42 AM Michael Harms @.***> wrote:
FYI: A copy of CMRR_MB_Slice_Order.pdf can be found at the re-located HCP Wiki: https://wiki.humanconnectome.org/docs/HCP%20fMRI%20slice-timing%20acquisition%20parameters.html
Can you please clarify what you mean by "temporal instabilities"? (Spin-history, steady-state issues leading to a visible banding artifact are not "temporal instabilities".)
— Reply to this email directly, view it on GitHub https://github.com/CMRR-C2P/MB/issues/380#issuecomment-2405590436, or unsubscribe https://github.com/notifications/unsubscribe-auth/AATFYMCZYKEJJNJI2CXU2JDZ22VATAVCNFSM6AAAAABPWC66RKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIMBVGU4TANBTGY . You are receiving this because you authored the thread.Message ID: @.***>
I noticed tSNR in the cmrr_ep2d_diff (and presumably cmrr_ep2d_se) depends markedly on the exact number of slices specified (as measured by a timeseries of b=0 images). I determined this to be a spin-history, steady-state problem relating to the slice order implemented by the sequence, which depends on the number of shots (i.e., number slices divided by MB factor). There used to be a CMRR_MB_Slice_Order.pdf online which detailed how the number of shots changed the slice order but not sure where one can find the information now.
With an odd number of shots (and MB > 1), you get high tSNR with the standard interleaved order. With an even number of shots, the "increment-n-slice interleave pattern" results in temporal instabilities and significantly reduced tSNR. From the screenshot below, the case with 78 slices shows the adjacent slices being excited much more closely in time than with the 76 slice case, the former resulting in temporal instabilities and low tSNR. Something should be done either to prevent the use of an even number of shots or at least provide a warning describing the issue and slice order change.
Importantly, the CMRR_MB_Slice_Order.pdf doesn't advise on what happens when using the Slice Multiplier 2 feature combined with MB > 1. It turns out that the behavior is different from the case where Slice Multiplier is not used (e.g., slice order for 128 slices SM1 is different for 64 slices SM2). However, the same tSNR problem exists - when the slice order is not standard interleaved order, tSNR is low. Initial testing indicates that standard interleaved order is used when the number of slices specified in the UI divided by MB factor is an even number (e.g., in one of the below screenshots, number of slices in the UI is set to 76 but with SM=2, actual number of slices acquired is 152).