1 SMS-NLINV Simultaneous Imaging of Multiple Slices Sebastian Rosenzweig Diagnostic and Interventional Radiology University Medical Center Göttingen May 29, 2020
1
SMS-NLINVSimultaneous Imaging of Multiple Slices
Sebastian RosenzweigDiagnostic and Interventional Radiology
University Medical Center Göttingen
May 29, 2020
SMS-NLINV
2
Motivation
[1] Uecker et al., Magn. Reson. Med. (2008)[2] Uecker et al., NMR Biomed (2010)
real-time imaging
NLINV
Regularized Nonlinear Inversion1,2
limited to single slice
SMS-NLINV
3
Motivation
[1] Uecker et al., Magn. Reson. Med. (2008)[2] Uecker et al., NMR Biomed (2010)[3] Larkman et al., Magn. Reson. Med. ( 2001)
real-time imaginglimited to single slice
SMS
Simultaneous Multi-Slice3
NLINV
Regularized Nonlinear Inversion1,2
less data demandtime-consistent slices
SMS-NLINV
4
Motivation
SMS-NLINV[1] Uecker et al., Magn. Reson. Med. (2008)[2] Uecker et al., NMR Biomed (2010)[3] Larkman et al., Magn. Reson. Med. ( 2001)
less data demandtime-consistent slices
real-time imaging
SMS
Simultaneous Multi-Slice3
NLINV
Regularized Nonlinear Inversion1,2
limited to single slice
SMS-NLINV
5
Conventional Multi-Slice Imaging
[Figures] adapted from Rosenzweig, MA Thesis 2016
SMS-NLINV
6
Conventional Multi-Slice Imaging
k-space reconstruction
Figures adapted from Rosenzweig, MA Thesis 2016
SMS-NLINV
7
Conventional Multi-Slice Imaging
k-space reconstruction
k-space reconstruction
Figures adapted from Rosenzweig, MA Thesis 2016
SMS-NLINV
8
Simultaneous Multi-Slice Imaging
k-space reconstruction
+
SMS-NLINV
9
Simultaneous Multi-Slice Imaging
-
+
k-space reconstruction
SMS-NLINV
10
Simultaneous Multi-Slice Imaging
-
+
+
-
k-space reconstruction
SMS-NLINV
11
Simultaneous Multi-Slice Imaging
-
+
k-space
SMS-NLINV
12
Simultaneous Multi-Slice Imaging
-
+
SMS-NLINV
k-space reconstruction
SMS-NLINV
13
Optimization Problem
SMS-NLINV
14
Optimization Problem
SMS-NLINV
15
Optimization Problem
SMS-NLINV
16
Optimization Problem
SMS-NLINV
17
Optimization Problem
● solve for update in each Newton-step
● iterative update
● Tikhonov regularization
SMS-NLINV
18
● non-linear formulation
● complementary k-space samples
Benefits
?
SMS-NLINV
19
Cartesian | 2 slices
SMS-NLINVnon-linear reconstruction
(calibration-less)
difference toreference
difference toreference
N=12
[1] Uecker et al., Magn. Reson. Med. (2014)
kspace
ESPIRiT1linear reconstruction
(additional coil calibration)
SMS-NLINV
20
Cartesian | 2 slices
difference toreference
difference toreference
[1] Uecker et al., Magn. Reson. Med. (2014)
kspace
N=4
SMS-NLINVnon-linear reconstruction
(calibration-less)
ESPIRiT1linear reconstruction
(additional coil calibration)
SMS-NLINV
21
Radial | 3 slices
NLINV
undersampled (10x)fully sampled
... ...
slice
SMS-NLINV
22
Radial | 3 slices
SMS-NLINVNLINV
aligned (10x)undersampled (10x)fully sampled
... ... ...
slice kz
SMS-NLINV
23
Radial | 3 slices
SMS-NLINVNLINV
complementary (10x)aligned (10x)undersampled (10x)fully sampled
... ... ......
slice kz
SMS-NLINV
24
Radial | 3 slices
SMS-NLINVNLINV
undersampled (10x)fully sampled
... ... ......
ky
kx
ky
kx
complementary (10x)aligned (10x)
kzslice
SMS-NLINV
25
Temporal Regularization
SMS-NLINV
26
Temporal Regularization
SMS-NLINV
27
Temporal Regularization
penalize the difference to the previous frame
SMS-NLINV
28
SMS real-time MRI
5 spokes per partition & frame
3 slices & 29 framesper second
SMS-NLINV
29
Wrap Up
SUMMARY● joint estimation of
images and coils
LIMITATIONS● ~ 2 - 5 slices● problem size● SMS-FLASH study
OUTLOOK● SMS-bSSFP sequence● T1 mapping1,2
[1] Wang et al., submitted for publication to Magn. Reson. Med. (2020). [2] Wang et al., ISMRM 2020
➔ improved image quality➔ time-consistency➔ multi-slice real-time MRI
SMS-NLINV T1 mapping1,2
https://github.com/mrirecon/sms-nlinv
https://github.com/mrirecon/sms-nlinv
Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 25Slide 26Slide 27Slide 28Slide 29