北师大fMRI Experimental Design_2011.ppt

Periodic Single Trial Designs Advantages ? Clean signal ? Can extract behaviorally ? Good estimation efficiency Disadvantages ? Unnatural design for subjects ? Few stimuli can be gathered ? Infrequently used ? More sensitive to scanner Rapid event-related fMRI. Fixed interval + short ISI = no variance from trial to trial Short ISI’s only work if the intervals are Randomized (“jittered”) Rapid ER-fMRI trials (or events in this case) are truly randomised as you would in a behavioural study Advantages Designs more similar to traditional “lab experiments” (flexibility) Ability to extract responses to behaviorally meaningful events (trial sorting) Images (face vs objects) and orientation (side up vs upside down) one type of event; two types of events; four types of events Rapid ER-fMRI Advantages Can present sufficient stimuli within a run to achieve good estimation efficiency and reasonable detection ability Distinguish different brain process associated with specific parts of a task, based on their relative timing. Randomized ER-FMRI Caveats Typically lower detection power than blocked design Requires optimized randomized sequence to achieve optimal estimation and detection efficiency Can be easy or somewhat difficult to analyze If modeling known HRF response, relatively easy Multiple regression methods If estimating HRF from data (“deconvolution”) then somewhat more challenging Require initial impulse response function estimation Randomized ER-FMRI Optimal Rapid ITI Rapid Mixed Trial Designs Short ITIs (~2 sec) are best (Dale and Buckner, 1997) Probability of Event for the optimal efficiency (Friston, 1999) If only interested in the differences among the N trial type, the probability of each event type should be 1/N. When null event and N trial type were included, the probability of each event type should be 1/(N+1). ISI vs. Stimulus Duration Bandettini and Cox 2000 make a brief mention of a 50/50 ratio of ITI to SD in rapid ER-fMRI So the mean ITI should