Older adult scanning methodology II: How to decide between conflicting literature?

1
Older adult scanning methodology IIHow to
decide between conflicting literature?

• Andy James
• fMRI Journal Club
• October 26, 2004

2
Conflicting articles who should we believe?

• How much do the articles conflict each other?
• Assess methodology for differences in
• subject samples
• experimental paradigm
• data collection and processing
• statistical analyses
• Do results support claims?
• Note Should be doing this already!

3
Example D BOLD HDR Function with Age
Delayed return to baseline in older (red) vs
younger (black) adults in visual (A) and motor
(B) cortices. Aizenstein et al., 2004. The BOLD
Hemodynamic response to aging. Journal of
Cognitive Neuroscience, 16, 789-793.
No difference in HDR between older and younger
adults for visual cortex. Huettel et al., 2001.
The effects of aging upon the hemodynamic
response measured by functional fMRI.
Neuroimage, 13, 161-175.
Fusiform
Calcarine
4
Methodology issue?
Aizenstein 2004
Huettel 2001

• Subjects
• Younger 8 (3 male) 24.2 (4.37) years old no
  medication
• Older 10 (6 male) 67.2 (4.83) years old no
  psychotropic medicines
• Paradigm
• Keypress with both index fingers whenever
  stimulus tap appeared. (1 s stimulus with 12 s
  ITI)

• Subjects
• Younger 11 (7 male) 23 (range 8-32) years old
• Older 11 (7 male) 66 (range 57-76) years old
• Medications not addressed
• Paradigm
• Passive viewing of flashing checkerboard (500 ms
  stimulus with 16.5s ITI)

5
Methodology issue?
Aizenstein 2004
Huettel 2001

• Data Collection/ Processing
• TR 2s, 1.5T GE, 3.8x3.8 mm2
• Statistical Analysis
• HDRs generated from positive sig. voxels only

• Data Collection/Processing
• TR 1s, 1.5T GE, 3.8x3.8 mm2
• Statistical Analysis
• HDRs generated from positive sig. voxels (?) or
  most sig. voxel

6
Do results support claim?
Delayed return to baseline in older (red) vs
younger (black) adults in visual (A) and motor
(B) cortices.
But both Huettel and Aizenstein excluded negative
voxels! (Aizenstein compares Huettel and Buckner,
but never addresses this studys discrepancy.)
7
How much do the articles conflict each other?
Alternate explanation older adults have more
negative voxels that affect HRF when averaged in.
8
Alternate explanation older adults have more
negative voxels that affect HRF when averaged in.
9
Conclusions

• Claim 2 SNR decreases with age
• Older brains exhibit greater HRF variability
• Older brains are activated to a lesser spatial
  extent (smaller ROI areas) and to a lesser
  magnitude (t-value thresholds)
• SNR improves with the square root of trials
  performed
• Possibly due to attenuated return to baseline?
  (Aizenstein)
• 1.5 SNR between groups means 2.25x as many
  trials for older adults
• How feasible is this for paradigms?

Discussion Your experiences with geriatric fMRI
research.
10
Relevance of Aging Research
Data from the US Bureau of the Census, 2000
11
Statistics for Participant Selection Criteria
Older Americans 2000 Key Indicators of
Well-Being Federal Interagency Forum on
Aging-Related Statistics (Forum)
12
Conditions Affecting Participant Selection
Criteria

• Neurological conditions
• depression
• strokes / infarcts
• memory impairment

• Physical conditions
• cardiac pacemaker
• artificial joints
• dental fixtures
• aneurysm clips
• arthritis
• spine curvature
• tattoos

DEsposito MD, Deouell L, and Gazzaley A. (2003).
Nature Reviews, 4, 1-11
13
Participants Ability to Perform Functional Task

• Performance influenced by
• Eyesight
• Hearing
• Arthritis
• Memory
• Attention and working memory

• Example Serial Reaction Time task
• Participants make motor responses to viewed
  stimuli
• Young RT m (sd) 323 (17) ms
• Older RT m (sd) 524 (88) ms
• Howard JH and Howard DV. (1997) Psychology and
  Aging, 12, 634-656

300-600 ms
300-600 ms
Rest of trial
Rest of trial
response
response
Total trial time 1500 ms
Total trial time 1500 ms
Introducing a sequence to stimulus location
results in decreased RTs (learning). Should
paradigm be adjusted to accommodate longer
RTs? Is a 100 ms learning gain in RT equivalent
across groups?
14
Ability to compare functional data
How do rigid / nonrigid transformations used to
convert brains to Talairach or MNI space account
for age-related morphology? (i.e. cortical
shrinkage, ventricular enlargement) How can we
compare sizes/shapes of ROIs across age
groups? Head motion stroke age mean 58
(range 22-78) nonstroke age mean 59 (range
25-71) young age mean 28 (range 25-38)
Seto E, Sela G, McIlroy WE et al.. 2001.
Neuroimage, 14, 284-297
15
Functional signal detection
Huettel SA, Singerman JD and McCarthy G. (2001).
The effects of aging upon the hemodynamic
response measured by functional MRI. Neuroimage,
13, 161-175.
Claim 1 The hemodynamic response function (HRF)
changes with age
Calcarine
Fusiform
16
Functional signal detection
Claim 1 The hemodynamic response function (HRF)
changes with age
17
Functional signal detection
Claim 1 The hemodynamic response function (HRF)
changes with age
Nonparametric comparison of relative standard
deviation across all epoch time points revealed
that elderly subjects had a higher standard
deviation than had the young in 15 of 19 time
points (plt.01).
18
Functional signal detection
Claim 2 Older participants have greater signal
to noise ratios (SNRs) in activated voxels
than younger participants
Calcarine SD (ROI)
Calcarine SD (voxel)
Intersubject group variability
19
Functional signal detection
Claim 2 Older participants have greater signal
to noise ratios (SNRs) in activated voxels
than younger participants
SNR not due to head motion SNR differences
largest when considering only single best voxel
from ROI
20
Functional signal detection
Claim 2 Older participants have greater SNRs in
activated voxels than young
Younger participants have significantly more
active voxels (plt.001, both ROIs)
Difference above is not an artifact from selected
t-value (3.5) (note divergence at t2.5)