Arterial spin labeling

1
Arterial spin labeling

• Edith Liemburg
• Ze Wang et al., 2008

2
Introduction to ASL

• PET-like for direct CBF measurement
• Measurement of slow neural changes
• Absolute quantification of blood flow

http//www.umich.edu/fmri/asl.html
3
Principle of ASL
1.  Tag inflowing arterial blood by magnetic
inversion 2.  Acquire the tag image
3. Repeat experiment without tag 4.  Acquire the
control image
http//www.umich.edu/fmri/asl.html
4
Image subtraction
-
-
-
-
5

• Terms
• CASL continious arterial spin labeling
• PASL pulsed arterial spin labeling

http//www.umich.edu/fmri/asl.html
6
ASL vs BOLD

• ASL
• Less slow drifts
• Changes localized in cappilaries
• Reduced inter-subject variability
• Better functional localization

• BOLD
• Short TR
• Many thin slices
• Higher intrinsic SNR

http//www.umich.edu/fmri/asl.html
7
Introduction ASL analysis

• Low intrinsic SNR signal only 1 5 of mean MR
  signal intensity
• CBF calculation ? intesity difference
• Critical preprocessing steps
• (Bit resolution)
• Motion correction
• Spatial smoothing normalization
• Global spike elimination
• Measure of global signal as covariate

8
ASL-toolbox

• Ze Wang et al. http//www.cfn.upenn.edu/perfusion
• Raw image to CBF conversion
• 3 methods of calculation
• Simple subtraction
• Zinc subtraction
• Surround subtraction
• Also unscaled (perf) and pseudo-BOLD images

9
Material and Methods

• Visual stimuli and finger tapping task in
  blockdesign (n 10)
• CASL 64 x 64 x 12 voxels (slice 6 mm), TR 3,
  labelling 1.6 s, delay 800 ms
• Pseudo-CASL sequence
• CBF calibration

f CBF, ?M control label signal, R1a
longitudinal relaxation rate of blood (R1a0.67
s-1), t labeling time (1.6 s), ? delay time
(0.8 s), a labeling efficiency (0.68), ?
blood/tissue water partition coefficient (0.9
g/mL) M0 control image intensity
10
Motion correction

• Signal intensity difference ? motion
• Seperate realignment vs combined PCASL or
  background suppression

11
Smoothing normalization

• Smoothing before CBF calcutation ? increased SNR
  outlier reduction
• Normalization after CBF calculation first level
  ? before minor signal changes incorrect
• time lag estimation
• (reslicing)

12
Spike elimination

• Combined realignment ? average substraction of
  motion paramaters of adjecent control label
  images
• Criteria averaged translation or rotation gt 3
  mm/o, subtracted translation or rotation gt 0.8
  mm/o, global CBF 3 standard deviations

13
Statistical analysis

• Simple subtraction increases peak t-value in
  visual cortex
• Including the global signal increases peak
  t-value cluster size

14
Proposed preprocessing steps

• Realignment seperate for control label
• Coregistration labelled to control
• Smoothing
• Exclude extracranial voxels (brain mask)
• CBF calculation simple subtraction
• First level analysis global CBF as covariate (no
  explicit masking, no high-pass filter, no HRF)
• Masking from mean CBF image
• Global spike elimination according to criteria
• Coregistration mean control to anatomy, con as
  other
• Normalization anatomy to template
• Second level