fMRI Journal Club August 31st, 2004

1
fMRI Journal ClubAugust 31st, 2004

• Experimental Design for Brain fMRI
• Chapter 30 of Functional MRI
• By G.K. Aguirre M.DEsposito
• Block Design Nelson Klahr
• Event Related Paul Wright

2
Goals of an fMRI Experiment

1. Be inferentially capable of rejecting a
   hypothesis of interest.
2. Maximize sensitivity for a predicted effect.

3
Hemodynamic Transfer FunctionA Low Pass System
4
Unaltered Frequency
5
High Pass Filter
6
Low Pass Filter
7
Temporal Autocorrelation

• Distribution of power is characterized by 1/f
  (sampling frequency)
• High enough to avoid the elevated noise range
  due to physiological fluctuations

8
Limitations of BOLD Output

• BOLD fMRI signal has no absolute interpretation
• weighted by its concentrations and is not a
  direct measurement of deoxyhemoglobin
• Generally tests for differences in the magnitude
  of the signal between different conditions within
  a scan.
• Can vary between subjects, scans, and voxels.

9
Cognitive Subtraction

• Differences in neural activity between the two
  conditions can be attributed to the cognitive
  process of interest.
• Can maximize sensitivity by minimizing
  inter-trial interval and by evoking cognitive
  processes in a homogenous manner.

10
Assumptions

• Pure Insertion adding to a pre-existing set of
  cognitive processes without affect.
• Ex. Button response
• Linearity transformation of neural activity into
  fMRI signal is linear. The response to the
  temporally adjacent stimuli is equal to the total
  response of the separated stimuli.

11
The Cons of Block Designs

• Constrained to grouping trials of the same type
  with each other in time. Does not allow for
  randomization of the order of stimuli.
• Imaging signal may be the result of anticipatory
  behaviors in which the subject engages before the
  presentation of each stimulus.
• Fixed A-B-C-A-B-C-A-B-C
• Random A-C-C-B-A-B-C-B-A

12
The Pros of Block Designs

• Superior Statistical Power
• Easy to position fundamental frequency
• Optimal block length for a boxcar design is 30
  sec but if the subjects hemodynamic transfer
  function is known then 14-20 sec is more desired.

13
Multi-Conditioned, Caterogical, Blocked
Experiments

• Can compromise previously mentioned problems with
  randomization by fixing the order of blocks
  within a subject but varying (counterbalance)
  this order across subjects.