View source for 3T Multiband

= Multiband =

== Overview sequences ==
All h 10% gap, fat suppression SPIR offset 175Hz
* 3.0mm, TR=0550ms, MB4 S1.5, FOV 240x240x118mm, WFS 14.2pix (42.6mm), PNS moderate. Gradient Mode Maximum, Dyn stabilization Regular, Max dyns 1820
* 2.7mm, TR=0700ms, MB4 S1.5, FOV 240x240x130mm, WFS 14.4pix (38.9mm), PNS moderate, Gradient Mode Maximum, Dyn stabilization Regular, Max dyns 1489 
* 2.0mm, TR=1600ms, MB4 S1.5, FOV=224x224x125mm, WFS 23.7pix (47.4mm), PNS moderate, Gradient Mode Maximum, Dyn stabilization Regular, Max dyns 1170 

== Sequence-parameter considerations ==
* Shimming
** Shimbox: inside brain? Is bottomline shimbox wrt OFC/air transition most important?
** What is a good shim? Check LINEWIDTH before start EPI? Log values
** Acquire extra B0-scan as check to avoid large B0-gradient?
** Use Maarten's patch for image based shimming: this patch applies shim to B0-prescan
* TR: 
** for physiological noise correction, aim for 750ms or less
** minimal TR: to be adjusted on-scanner (not realistic in VM)
* Gradient mode: Set to Enhanced if duty cycle is limiting factor (enormous difference on AMC-wide bore scanner)
* Water Fat Shift: minimal. Increase to reduce TR
* TE not phase: not important as fat signal is already dispersed at this TE. TE30-32, 27 to short.
* Slice gap optional for voxel sizes <3mm, set to sefault (10%) 
* Acquired/Reconstructed Matrix size should match
* Multiband factor * SENSE ~= 6
* Half scan: no. Unless minimal TR deviates more than 10% from shortest. But: see Gradient mode
* Dynamic Stabilization: regular (enhanced optional for longer runs, ~100ms penalty. May give distortion)
* Offc/Ang: free rotatable Yes
* PNS moderate or low
* Flip angle --> [http://www.mritoolbox.com/ErnstAngle.html Ernst angle] A typical T1 value for gray matter @ 3T: 1330 ms (Kruger, et al, 2001).
* Planning: angulate in parallel to OFC. Change read-out direction? (AP-->PA?)
* Equidistant dynamic sampling

== distortion correction and registration==
* Create second GE-EPI with opposite fatshift direction, set preparation to auto (so it won't perform a second B0 shim) and group the sequences 
* Alternatively create two SE-EPIs for distortion correction (better unwarping around regions with dropout) 
* Perhaps not include dummies and use one of the pre-saturation scans for registration as it has better grey/white matter contrast (see [https://fsl.fmrib.ox.ac.uk/fslcourse/lectures/practicals/registration/index.html#multiband Multiband data registration])

== Data-export ==
* enhanced dicom does not work for >16k images
* export nifti
* export PAR/REC or XML/REC to store sequence parameters

== Slice timing in multiband ==

One way to figure this out is using Sequence Development Mode/Editor (SDM). During the simulation, under Parameters tab for your particular sequence look under UGN8_DEF>UGN8_DEF_loc_order (make sure to skip the sense reference scan and other pre-scans.) :good news: On the scanner, this will be under VAL08_DEF>VAL08_DEF_loc_order. This tells you the actual order in which the “locations” are acquired simultaneously (note that the index starts from 0).

What are the locations? They are the simultaneous slices or as you call them slice pairs (total number of locations = divide the total number of slices by your multiband factor).

How are the locations ordered? It mainly depends on the Slice Scan Order set in the UI under the Geometry tab (e.g. interleaved, default, etc) and on the scanner’s Help (F1) menu you can search for them.

How do you translate location orders into scan pairs that are simultaneously excited? Here’s an example: say you had 12 slices and MB-factor = 2, and set the Slice Scan Order to “Default” and we assume the number of packages is 1. You can calculate the total number of locations to be 6 (12 slices / MB factor 2). On the scanner’s SDM, the VAL08_DEF_loc_order would show: 0, 2, 5, 3, 1, 4. Grab the first location (i.e. 0 or slice1) and add the total number of locations (i.e. 6) to it. So in this case location 0 (slice 1) and location 6 (slice 7) are simultaneously excited/paired first. Repeat for all locations in the order provided by VAL08_DEF_loc_order and you will get:
(loc 0 and loc 6 = slice 1 and slice 7),
(loc 2 and loc 8 = slice 3 and slice 9),
(loc 5 and loc 11= slice 6 and slice 12),
(loc 3 and loc 9 = slice 4 and slice 10),
(loc 1 and loc 7 = slice 2 and slice 8),
(loc 4 and loc 10 = slice
>> 
 
For MB>2 ( our MB=3 scan) you then have to keep adding the number of locations (Nslices/Mbfactor)
so for ascending order, MB=3, and 30 slices (10 locations)
we would have
(loc 0 and loc 10 and loc 20 = slice 1 and slice 11 and slice 21),
(loc 1 and loc 11 and loc 21 = slice 2 and slice 12 and slice 22),
(loc 2 and loc 12 and loc 22 = slice 3 and slice 13 and slice 23),
(loc 3 and loc 13 and loc 23 = slice 4 and slice 14 and slice 24),
etc, which all makes sense in order to maximize the distance between slices, but always good to be 100% sure 😊

Jeroen Siero is the one with expertise in this matter.