March 25 2012 Jason Su Questions ASSET vs ARC How do parallel imaging artifacts propagate through to the maps Is mcDESPOT w BS B1 correction a repeatable experiment at 3T What is the effect of inaccurate B1 on each of the maps ID: 776075
Download Presentation The PPT/PDF document " mcDESPOT Repeatability at 3T" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
mcDESPOT Repeatability at 3T
March 25, 2012
Jason Su
Slide2Questions
ASSET vs. ARC: How do parallel imaging artifacts propagate through to the maps?
Is
mcDESPOT
w/ B-S B1 correction a repeatable experiment at 3T?
What is the effect of inaccurate B1 on each of the maps?
How do the maps compare to 1.5T?
Does MWF remain the same? We would hope it would but does not for UBC group’s MWF
Slide3Scanning
1 volunteer
2 scans, baseline and follow up exactly 24 hours later
Acceleration: 70% partial k-space, ASSET 2 or ARC 2x1.5
B-S: TR 500ms, BS FA 270 deg
.
, 32x32 22cm FOV, 6mm slice
Day 0 – no HOS
8ch head coil: ASSET, ARC
32ch head coil: ASSET
Day 1 – some HOS
8ch head coil: ASSET (SSFP HOS), ARC (HOS)
32ch head coil: ASSET (no HOS), ARC (no HOS)
6:30 for one complete
mcDESPOT
protocol at 2mm isotropic
Slide4Post-Processing
B-S B1 maps via updated Orchestra recon
Converted to flip angle then “kappa” maps
xmtAddScan
and
xmtAddFTG
are stored in header for efgre3d22M4v04
Need to add ia_rf1 to header, assuming now that = 32767 when
xmtAddScan
< 200, otherwise unknown (thankfully not the case)
Linear interpolation from 27 to 80 slices
3T2 protocol expiry? Interest in getting it running at Sherman and 3T3.
Registration to intermediate “halfway” space
Distributes interpolation errors/blur equally between two volumes
Allows for much better
voxelwise
comparison
White matter segmentation is performed with FSL FAST on the registration target
The mask is then eroded with a spherical
kernal
of radius 2mm to stay away from possible GM edges
Slide5ASSET vs. ARC
Images and maps are presented without B1 correction
Would have added an additional variable and not needed for this comparison
Artifacts are more prominent in ASSET
Clear skull aliasing artifact running through center of images
How do parallel imaging artifacts propagate through to the maps?
Slide6Sample Images: 32ch ASSET SPGR
Slide7Sample Images: 32ch ARC SPGR
Slide8Sample Images: 8ch ASSET SPGR
Slide9Sample Images: 8ch ARC SPGR
Slide1032ch ASSET: Single T1
Slide1132ch ARC: Single T1
Slide1232ch: Single T1
Slide1332ch: Single
T1 in WM
Slide148ch: Single T1
Slide158ch: Single
T1 in WM
Slide1632ch ASSET: Single T2
Slide1732ch ARC: Single T2
Slide1832ch: Single T2
Slide1932ch: Single
T2 in WM
Slide208
ch: Single T2
Slide218
ch: Single T2 in WM
Slide2232ch ASSET: MWF
Slide2332ch ARC: MWF
Slide2432ch: MWF
Slide2532ch:
MWF in WM
Slide268ch: MWF
Slide278ch:
MWF in WM
Slide28Conclusions
32ch coil seems to give less consistent results than 8ch for ARC vs. ASSET
Artifacts are more apparent with 32ch (surprising?) and are responsible for higher percent difference than 8ch
8ch SPGR images appear better with ASSET
MWF seems less affected by the artifacts than T1 and T2, or at least harder to see
Slide29Repeatability
Is
mcDESPOT
w/ B-S B1 correction a repeatable experiment at 3T?
Will show 8ch ASSET, 8ch ARC, 32ch ASSET
32ch ARC was not collected at baseline
Slide30Sample Image: Kappa Map
Slide31Sample Image: Kappa Map
Slide328ch ASSET 0d: Single T1
Slide338ch ASSET 1d: Single T1
Slide348ch ASSET: Single T1
Slide358ch ASSET: Single
T1 in WM
Slide368ch ARC: Single T1
Slide378ch ARC: Single
T1 in WM
Slide3832ch ASSET: Single T1
Slide3932ch ASSET: Single
T1 in WM
Slide408ch ASSET 0d: Single T1
Slide418ch ASSET 1d: Single T2
Slide428ch ASSET: Single T2
Slide438ch ASSET: Single
T2 in WM
Slide448ch ARC: Single T2
Slide458ch ARC: Single T2 in WM
Slide4632ch ASSET: Single T2
Slide4732ch ASSET: Single
T2 in WM
Slide488ch ASSET 0d: MWF
Slide498ch ASSET 1d: MWF
Slide508ch ASSET: MWF
Slide518ch ASSET:
MWF in WM
Slide528ch ARC: MWF
Slide538ch ARC:
MWF in WM
Slide5432ch ASSET: MWF
Slide5532ch ASSET:
MWF in WM
Slide568ch ASSET vs. ARC 1d: MWF
Slide57Results Table
0d Mean1d Mean0d SD1d SDMean ChangeSD ChangeMean % ChangeSD % ChangeT18ch ASSET1704.91773.6708.9732.368.7023.324.033.298ch ARC1511.01749.3654.0731.3238.3077.2815.7711.8232ch ASSET1681.81690.9707.4702.89.12-4.610.54-0.65T18ch ASSET103.5104.490.889.70.90-1.160.87-1.288ch ARC100.0105.596.693.35.53-3.385.53-3.5032ch ASSET111.8113.693.994.41.840.421.650.45MWF8ch ASSET0.0780.0760.0890.087-0.002-0.002-3.132-2.4818ch ARC0.1030.0760.1020.090-0.027-0.012-26.567-12.12132ch ASSET0.0680.0680.0810.0820.0000.0000.3730.449
Mean and SD are measured over whole brain.
Slide58Conclusions
Need to collect data with better HOS control for 8ch
Does HOS have similar effect of dropping T1 and T2 for 32ch (it probably should)?
If HOS indeed has a disruptive effect, test in phantom with known T1 to see whether we should use or not
32ch ASSET repeatability is admirable
0d vs. 1d seems about equivalent to difference we saw in ASSET vs. ARC (ignoring artifacts): within noise
There may be a bias/dependence on coil choice (see MWF 8ch vs. 32ch ASSET)
mcDESPOT
is a repeatable experiment at 3T with B-S B1 correction and 32ch coil
Most likely 8ch too but being cautious, single T1 shift perplexing
Slide59B1 Inhomogeneity
What is the effect of inaccurate B1 on each of the maps?
Compare the results after processing with and without B1 map correction
Slide60Sample Image: Kappa Map
Slide618ch ASSET w/o B1: Single T1
Slide628ch ASSET w/ B1: Single T1
Slide638
ch: Single T1
Slide648ch ASSET w/o B1: Single T2
Slide658ch ASSET w/ B1: Single T2
Slide668
ch: Single T2
Slide678ch ASSET w/o B1: MWF
Slide688ch ASSET w/ B1: MWF
Slide698
ch: MWF
Slide70Conclusions
Overflipping
-> higher T1
Overflipping
-> same T2?
SSFP curve is broad, less sensitive to 10% FA error
Overflipping
-> lower MWF
Tissue looks like the longer component is taking up bigger fraction
Slide711.5T vs. 3T
How do the maps compare to 1.5T?
Does MWF remain the same?
UBC group’s MWF increases at 3T (in WM, 0.08 -> 0.11, Oh 2005)
Comparison of 29 vs. 34 year old males respectively
1.5T scanning from
MSmcDESPOT
, done in London, Ontario
No parallel imaging is done at 1.5T
Slide728ch 1.5T: Single T1
Slide738ch ASSET 3T w/ B1: Single T1
Slide741.5T vs. 3T: Single T1
Slide751.5T vs. 3T: Single T1 in WM
Slide768ch 1.5T: Single T2
Slide778ch ASSET 3T w/ B1: Single T2
Slide781.5T vs. 3T : Single T2
Slide791.5T vs. 3T : Single T2 in WM
Slide808ch 1.5T: MWF
Slide818ch ASSET 3T w/ B1: MWF
Slide821.5T vs. 3T: MWF
Slide831.5T vs. 3T: MWF in WM
Slide84Conclusions
T1s get longer
at 3T as expected
T1s are in
the
expected range from literature (
Santisz
2005):
1.5T WM-GM ~780-1100ms
3T WM-GM ~1100-1500ms, paper measures
GM at 1800ms
Smaller
shift in T2 downward as
expected
T2s seem
underestimated in WM
1.5T
WM-GM ~80-95ms
3T WM-GM ~55-70ms
Should extract same ROIs as paper for better comparison
MWF is comparable between 1.5T and 3T
Similar
distribution especially in WM
Certainly much less difference than the 30%+ shift for UBC MWF
Slide85Future Work
efgre3d PSD modifications to streamline post-processing pipeline
Determine the effect of HOS
Evaluate bias due to coil choice
ARC 32ch
repeatability
Subject controlled 1.5T and 3T comparison
N++
Slide86Thank You
Lex
Mitchell as the willing volunteer
Mehdi
Khalighi
as the B1 mapping expert
Michael
Zeineh
as the catalyst
Brian
Rutt
as the drive