Steady-state free precession and other 3D methods for

1
Steady-state free precession and other 3D methods
for high-resolution FMRI
Karla L. Miller FMRIB Centre, Oxford University
2
Why is high-resolution FMRI so difficult?

• Signal-to-noise ratio
• For example, 2x2x2 mm has 8x SNR of 1x1x1 mm
  (would require 64 times longer scan)
• For single-shot, distortion increases with matrix
  size
• Isotropic resolution (thin slices) is hard in 2D

3
2D High-resolution FMRI
Acquire EPI in multiple shots (segmented or
interleaved) Allows increased resolution
without increased distortion High-resolution
in-plane, but limit on slice thickness!
4
2D Multi-slice MRI
excited slice
Each slice excited acquired separately TR time
between repeated excitation of same slice
(typically 13 seconds) Slices no thinner than 1
mm
5
True 3D imaging
excited volume
Excite entire slab, readout in 3D k-space TR
time between repeated slab excitations (5-50
ms) Can achieve thin slices (isotropic
resolution, like structurals)!
6
SNR Benefit of 3D Trajectories
SNR is higher for 3D since same magnetization is
sampled more frequently Calculated for 3D
stack-of-spirals
Yanle Hu and Gary Glover, Stanford
7
3D Functional MRI

• Advantages
• SNR benefits, provided short TR can be used
• Can achieve thinner slices (e.g., for isotropic
  voxels)
• 3D multi-shot ? low distortion

• Disadvantages
• Can require long volume scan times (may be
  fixable!)
• Acquisition time (e.g., slice timing) is
  difficult to define
• Slices must be contiguous (no inter-slice gap)

8
Adapting echo planar imaging (EPI) to 3D
2D segmented EPI
9
3D EPI GRE at 3T
0.8 x 0.8 x 0.8 mm3 0.5 mm3 TR69 ms, 7 s/vol,
24 minutes scan time
10
3D EPI GRE at 3T (0.8 x 0.8 x 0.8 mm3 )
Single image 7 s scan time
Mean timecourse image 4 min scan time
11
Adapting spiral to 3D
2D interleaved spiral
3D stack-of-spiral
Yang et al, MRM 1996
12
Comparison of 2D vs 3D spiral FMRI

• 20 higher functional SNR in 3D compared to 2D
• Significantly more activated voxels (2x at chosen
  threshold)

Hu and Glover, MRM 2006
13
3D spiral GRE with partial k-space

• Faster imaging 64 slices in 6.4 s (full) vs. 4.0
  s (partial)

• Higher statistical power due to reduced
  physiological noise

Hu and Glover, MRM 2006
14
High-resolution retinotopy at 7T
2D single-shot EPI
3D segmented EPI

• 1x1x1 mm3 resolution
• Identification of retinotopically-distinct
  regions
• Reduced distortion in 3D segmented EPI

Itamar Kahn and Randy Buckner, MGH
15
3D GRE BOLD at 7T
0.67 x 0.67 x 0.67 mm3 0.3 mm3 12 minutes scan
time
Karla Miller and Chris Wiggins, MGH
16
3D GRE BOLD at 7T
0.58 x 0.58 x 0.58 mm3 0.2 mm3 18 minutes scan
time
Karla Miller and Chris Wiggins, MGH
17
3D Imaging GRE vs. SSFP

• 3D imaging generally requires short TR
• SSFP tends to out-perform GRE in this regime

18
Balanced Steady-state Free Precession (SSFP)

• SSFP signal dependence on off-resonance

• Transition band SSFP image in signal transitions
• Contrast deoxyHb frequency shift

• Passband SSFP image in flat part of signal
  profile
• Contrast T2 at short TR

19
Transition-band SSFP

• Functional contrast occurs in bands
• Changing center frequency shifts region of high
  signal (and functional contrast)
• Multi-frequency experiments
• Repeat stimulus at multiple center frequencies to
  extend coverage
• Combine data into single activation map

20
3D Spiral transition-band SSFP at 1.5T
1 x 1 x 2 mm3, 3D spiral, standard head coil
Courtesy Jongho Lee, Stanford University
21
3D EPI tbSSFP at 3T
0.8 x 0.8 x 0.8 mm3 0.5 mm3 TR35 ms, 8.3
s/vol, 24 minutes scan time
22
3D EPI tbSSFP FMRI at 7T
0.75 x 0.75 x 0.75 mm3 0.4 mm3 22 minutes scan
time
Collaboration with Chris Wiggins, MGH
23
Physiological noise transition-band SSFP
Compared to GRE, higher physiological noise in
tbSSFP Poor fit with standard physiological noise
model
24
Reducing physiological noise in SSFP
Respiration modulates frequency shift in SSFP
bands Real-time feedback to compensate for
frequency drift
Jongho Lee et al, MRM 2006
25
Dynamic frequency tracking
compensation off
compensation on
Jongho Lee et al, MRM 2006
26
Passband SSFP vs. GRE (3T)
GRE
pbSSFP
TE 3 ms
TE 25 ms
27
Physiological noise passband SSFP
Short TR (6-12 ms)
Compared to GRE, lower physiological noise in
pbSSFP
28
Conclusions

• Why 3D for high-resolution FMRI?
• High-res ? multi-shot ? short TR ? 3D
• Lower distortion with short, 3D readouts
• Can achieve isotropic resolution (thin slices)
• Challenges and advances
• Efficient 3D versions of both EPI and spiral
  trajectories
• Volume acquisition times Speed up with partial
  k-space (or parallel imaging)
• SSFP FMRI
• New method for FMRI contrast
• Highly suitable to 3D due to short TR

29
Acknowledgements

• Martinos Centre, MGH
• Christopher Wiggins
• Graham Wiggins
• Itamar Kahn

Funding NIH, GlaxoSmithKline, EPSRC, Royal
Academy of Engineering
Related work 357 SSFP analysis (Th-AM), 272
SSFP modeling (Th-PM)