Post on 28-Mar-2015
When zero is not zero: The problem of ambiguous baseline conditions in fMRI
Stark & Squire (2001)
By Mike Toulis
November 12, 2002
OR
Does ‘rest’ really mean rest?
Objective:
Determine if condition of ‘rest’ is always an appropriate baseline in fMRI studies
Determine if alternative baseline conditions could be more appropriate
Design
Used both block and rapid event-related studies to examine 6 memory-encoding tasks along with a ‘rest’ condition
Participants
Participants were the same for both studies: 3 men, 5 women; mean age of 27.6 yrs (range 24 – 31 yrs)
Tasks
Novel pictures Arrows
Familiar pictures Moving fixation
Noise detection ‘Rest’
Odd/even digits
Experiment 1 - Block Design
Each task was presented three times in each run (21 blocks per run) in a ‘fixed pseudorandom’ order
Each block lasted 21 sec (1 run approx. 7.5 min)
Four runs were completed for a total of 12 blocks of each task (data for each participant averaged over the four runs)
Brief instruction appeared at top of screen for each task (e.g. “Odd or Even” or “Rest”)
Imaging Parameters
Siemens 1.5T Vision scanner
Whole-brain T2* images obtained using echoplanar single-shot pulse sequence (matrix size 64 x 64)
TE = 43msec
Flip angle = 90º
In-plane resolution of 4 x 4 mm
Functional Images
32 4-mm-thick slices aligned with the principle axis of the hippocampus
153 images obtained for each slice
TR = 3sec
Stimulus presentation began on the 5th image and ended on the 148th image (allowed for stabilization and return to baseline)
After last image, high resolution (1x1x1mm) structural scan was obtained
Functional Images (cont’d)
Images transformed into Talairach space
Transformation resulted in voxel size of
2.5 x 2.5 x 2.5 mm
Ten ROIs identified for analysis (perirhinal cortex, temporopolar cortex, entorhinal cortex, parahippocampal cortex, hippocampal region, each bilaterally analyzed)
Results
Four ROIs displayed significantly greater activity during the ‘rest’ task than during several other tasks:
Left/right parahippocampal cortex
Left/right hippocampal region
Results (cont’d)
Authors’ Conclusions
With ‘rest’ as baseline, familiar picture task shows no activity in medial temporal lobe and activity associated with encoding novel pictures is limited to right parahippocampal cortex
When either odd/even digit or noise detection task used as baseline, substantial medial temporal activity was observed in both familiar and novel pictures tasks
Authors’ Conclusions (cont’d)
Appears that presence of activity during the ‘rest’ condition masked activity throughout the medial temporal lobe during both novel and familiar pictures tasks
2 problems:
‘rest’ was not rest – instruction displayed
fMRI is difference in activity NOT presence of activity
Experiment 2 – Rapid Event-Related Design
Two rapid event-related designs used (trials all 3sec each)
1) Trial using novel picture task, familiar picture task, and ‘rest’ (“conventional design”)
2) Trial using novel picture task, familiar picture task, and odd/even digit task
No instructions displayed on screens for Exp. 2 (changed tasks from Exp. 1)
Imaging parameters, functional images, ROIs in Exp. 2 were all identical to those used in Exp. 1
Same 8 participants as in Exp. 1
Results
Magnitude of response to both novel picture and familiar picture tasks greater when odd/even digit task used as baseline than when ‘rest’ used as baseline
Results (cont’d)
When ‘rest’ used as baseline, significant response to novel pictures found in both left and right parahippocampal regions; but, no significant response to novel pictures found in hippocampal region
When ‘rest’ used as baseline, no response to familiar pictures found in any of the ten ROIs
Results (cont’d)
When odd/even digit task used as baseline, significant response to novel pictures found in eight of the ten ROIs
When odd/even digit task used as baseline, significant response to familiar pictures found in seven of the ten ROIs
Results (cont’d)
Authors’ Conclusions
Effect of activity during ‘rest’ trials reduced fMRI response; often to the point of apparently eliminating the effect of presenting novel or familiar stimuli
Other Findings
Activity during baseline condition can reverse sign (direction) of hemodynamic response (Exp. 2)
Left Parahippocamal Region Left Motor Cortex
Other Findings (cont’d)
Activity in occipital regions decreased in noise detection task (a below) and
odd/even digit task (b below) relative to rest (Exp. 1)
Other Findings (cont’d)
Authors’ potential reasons for decreased activity relative to‘rest’ in occipital areas:
Possibility of scanning room during rest
Suppression of visual cortical activity outside central focus of task
Presence of visual imagery during rest
Another possibility? – Presentation of word “Rest” (i.e. 4 characters) contains more visual stimuli than presentation of a digit (1 char.) or presentation of white field (0 char.)
Discussion
Authors’ claim that results of studies demonstrate that even short periods of ‘rest’ do not provide an optimal baseline for fMRI research
Significant cognitive activity found during ‘rest’ condition
Study highlights key limitation of BOLD fMRI studies – it is a contrastive methodology with no ‘true’ baseline
Takeaways
Baseline is what you choose it to be
Results only contrast blood oxygenation levels between conditions
Can only infer differences in activity not whether or not activity exists
Know ahead of time what you want to compare and ensure it’s appropriate for your study