Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs...
Transcript of Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs...
![Page 1: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/1.jpg)
Experimental Design for fMRI
Petra Schweinhardt
Nov 4, 2013
![Page 2: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/2.jpg)
FMRI Experiment
Aim: to manipulate the subject’s mental state/behavior/experience in some way to produce a neurovascular response
Define process to examine Define task to manipulate process Measure activation during task with fMRI Compare fMRI signal between tasks
![Page 3: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/3.jpg)
Types of Experimental Designs
1. Categorical – comparing the effect of one task on the dependent variable to the effect of another task (or rest)
2. Factorial - combining two or more factors within a task and looking at the effect of one factor on the response to other factor(s)
3. Parametric – varying ‘doses’ of task
![Page 4: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/4.jpg)
Comparing activation in response to one task to another task/rest
Control/comparison condition of most importance
Categorical design
![Page 5: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/5.jpg)
Comparing activation in response to one task to another task
Control/comparison condition of most importance
Categorical design
+ + +
![Page 6: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/6.jpg)
Comparing activation in response to one task to another task
Control/comparison condition of most importance
Categorical design
Try to isolate the effect of interest as much as possible!
![Page 7: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/7.jpg)
Factorial design Combining two or more factors within a task and looking at
the effect of one factor upon the other(s)
LOW
LOW HIGH
Distraction
A
C
B
D Pain
HIGH
![Page 8: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/8.jpg)
Factorial Design
13*17?
14*19? 2*2?
1*3?
LOW
Pain
HIGH
LOW HIGH
Distraction
A
C
B
D
![Page 9: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/9.jpg)
• Main effect of pain: A+B+C+D
• Simple contrasts, e.g.:
High Pain > Low Pain (C+D)-(A+B)
• Influence of one factor on the
other, e.g: Which voxels show most activation with high distraction in high pain condition? (D-B)-(C-A)
13*17?
14*19? 2*2?
1*3?
A C B D
LOW
Pain
HIGH
LOW HIGH
Distraction
A
C
B
D
![Page 10: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/10.jpg)
Parametric Design
• Linear • Nonlinear
Hypothesis: as task dose changes, BOLD response changes
Exploring systematic changes in the brain responses according to various ‘doses’ of task (difficulty levels, increasing sensory input, drug
doses, etc)
Contrast -3 -1 1 3
![Page 11: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/11.jpg)
Independent of design …
• Subject instructions
• Check for task performance
• Record additional explanatory variables / confounding factors (reaction times, accuracy, subjective ratings, autonomic measures, etc)
![Page 12: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/12.jpg)
FMRI Experiment
47*9=?
Define process to examine Define task to manipulate process Measure activation during task with fMRI Compare fMRI signal between tasks
- =
![Page 13: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/13.jpg)
47*9=?
-
![Page 14: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/14.jpg)
47*9=?
-
Signal is very noisy No absolute signal in
BOLD fMRI: BOLD is not quantitative
![Page 15: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/15.jpg)
47*9=?
46*8=?
45*7=?
51*9=?
53*7=?
47*8=?
![Page 16: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/16.jpg)
47*9=?
46*8=?
45*7=?
51*9=?
53*7=?
47*8=?
![Page 17: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/17.jpg)
47*9=?
46*8=?
45*7=?
51*9=?
53*7=?
47*8=?
![Page 18: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/18.jpg)
47*9=?
46*8=?
45*7=?
51*9=?
53*7=?
47*8=?
![Page 19: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/19.jpg)
ß1 =
47*9=?
46*8=?
45*7=?
51*9=?
53*7=?
47*8=?
![Page 20: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/20.jpg)
ß1 =
47*9=?
46*8=?
45*7=?
51*9=?
53*7=?
47*8=?
+ ɛ
![Page 21: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/21.jpg)
Block designs
Either one stimulus is presented throughout the block or several stimuli of the same type are presented at fixed intervals
rest A rest A rest A rest A rest A
![Page 22: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/22.jpg)
Block designs
Optimal block length: ~ 15 to 20 sec (blocks should be equal)
Total cycle time: << 100 sec
Statistically the most efficient design
![Page 23: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/23.jpg)
But…
47*7=? 47*9=? 43*9=?
15 sec 15 sec 15 sec 15 sec 15 sec 15 sec
![Page 24: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/24.jpg)
But…
47*7=? 47*9=? 43*9=?
15 sec 15 sec 15 sec 15 sec 15 sec 15 sec
Little control over what subject is doing (e.g. time on task) Stimulus predictability → anticipation Habituation within blocks and between blocks Subject might get bored Parallel comparison (A B A B) is problematic A rest A in Scan 1 versus B rest B in Scan 2 is problematic A rest B rest A rest B is problematic
![Page 25: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/25.jpg)
Event-related Designs
Image BOLD signal in response to very brief stimuli
![Page 26: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/26.jpg)
Events, ISI=32 s
![Page 27: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/27.jpg)
Events, ISI=32 s
BOLD response
Allows estimation of Hemodynamic Response Function (HRF)
![Page 28: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/28.jpg)
=
Fixed ISI = 16s
Not particularly efficient…
Stimulus (“Neural”) HRF Predicted Data
![Page 29: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/29.jpg)
=
Fixed ISI = 4s
Very Inefficient…
Stimulus (“Neural”) HRF Predicted Data
![Page 30: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/30.jpg)
=
Randomised ISImin= 4s
Much more efficient
Stimulus (“Neural”) HRF Predicted Data
![Page 31: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/31.jpg)
Flexibility of event-related designs
Ploner et al., PNAS 2010;107(1):355-60
![Page 32: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/32.jpg)
Ploner et al., PNAS 2010;107(1):355-60
Pain?
No pain? Pain?
No
pain? Pain?
No
pain? Pain?
No pain?
Pain?
No pain? Pain?
No
pain? Pain?
No pain?
Pain?
No pain?
Pain?
No pain?
Pain?
No pain?
Flexibility of event-related designs
![Page 33: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/33.jpg)
Ploner et al., PNAS 2010;107(1):355-60
Pain?
No pain? Pain?
No
pain? Pain?
No
pain? Pain?
No pain?
Pain?
No pain? Pain?
No
pain? Pain?
No pain?
Pain?
No pain?
Pain?
No pain?
Pain?
No pain?
Pain > No Pain
![Page 34: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/34.jpg)
Ploner et al., PNAS 2010;107(1):355-60
Pain?
No pain? Pain?
No
pain? Pain?
No
pain? Pain?
No pain?
Pain?
No pain? Pain?
No
pain? Pain?
No pain?
Pain?
No pain?
Pain?
No pain?
Pain?
No pain?
No Pain > Pain
![Page 35: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/35.jpg)
• Why does it work to have BOLD responses close together in time?
![Page 36: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/36.jpg)
• Why does it work to have BOLD responses close together in time?
Two key properties of a linear system: Scaling Superposition
![Page 37: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/37.jpg)
Huettel and McCarthy, NeuroImage 2000, 11; pp. 547–553
![Page 38: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/38.jpg)
Dale and Buckner (1997), Human Brain Mapping 5:329–340
ISI=2 sec
“Time-invariant
response”
![Page 39: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/39.jpg)
Savoy et al. (1995), Proc. Soc. Magn. Reson. Med. Third Sci. Meeting Exhib. 2, 450
![Page 40: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/40.jpg)
What is the minimum ISI for a valid linear approximation?
Linearity assumed
Friston et al., NeuroImage 10, 607–619 (1999)
![Page 41: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/41.jpg)
Event Related Designs
• Trials (events) are presented briefly in (pseudo-) random order
• ISI should be randomized (“jitter”)
• ISI should be short but not shorter than 2 to 5 sec
![Page 42: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/42.jpg)
Event-related Designs
Flexibility in design
Not every experiment can be turned into block design
Randomised trial order
Post-hoc classification of trials (correct responses, reaction times, etc)
Some events can only be indicated by participant
But
Assumptions about shape of HRF
Lower statistical power
![Page 43: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/43.jpg)
Mixed designs
Tonic effects
Phasic effects
Allows investigation of phasic and tonic effects Perhaps different processes of interest Exploit advantages of block and event-related
designs
![Page 44: Experimental Design for fMRI - bic.mni.mcgill.ca · 04/11/2013 · Types of Experimental Designs 1. Categorical – comparing the effect of one task on the dependent variable to](https://reader035.fdocuments.in/reader035/viewer/2022062918/5eddd91dad6a402d66690f84/html5/thumbnails/44.jpg)
Good Practices of fMRI Experimental Design
• Choose stimulus and timing to create maximal change in (cognitive) process of interest
• Get task-related measure of subject behavior in the scanner (task verification, additional regressors)
• Scan as many subjects as possible; scan as long as you can,
considering psychological effects (fatigue, habituation)
• Use block design if you care about detecting differences, and event-related designs if you want to link activity to particular events
• Limit the number of conditions; pairwise comparisons far apart in time decrease power and overlap with low-frequency noise
• In event-related designs, randomize (or pseudo-randomize) the ordering of events
• Randomize (‘jitter’) intervals between events that need to be distinguished (decorrelate their predicted signals after HRF convolution)