FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI...
Transcript of FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI...
![Page 1: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/1.jpg)
FMRI Experimental Design
Lila DavachiDepartment of Brain & Cognitive Sciences, M.I.T.
![Page 2: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/2.jpg)
Because fMRI BOLD data is not anBecause fMRI BOLD data is not anabsolute measure of neuronal activity,absolute measure of neuronal activity,
all study designs must provide theall study designs must provide theopportunity to statistically contrast theopportunity to statistically contrast the
neuronal activity of interest with aneuronal activity of interest with asuitable rest or background conditionsuitable rest or background condition..
Thus, study design is Thus, study design is of paramount importance.of paramount importance.
![Page 3: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/3.jpg)
Why?Why?HypothesisHypothesis
How?How?fMRI Study DesignfMRI Study Design
Where?Where?NeuroanatomyNeuroanatomy
What?What?BehaviorBehavior
Thanks to Chantal Stern
![Page 4: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/4.jpg)
43 * 7 = ?
![Page 5: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/5.jpg)
![Page 6: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/6.jpg)
Key Points
• What can fMRI tell you?
• Always comparing across conditions
• Characteristics of the hemodynamic response(HRF) and how this affected the sequentialdevelopment of fMRI paradigms and influencesstudy design
• Sense of important design issues
![Page 7: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/7.jpg)
What (good) is fMRI?
What it can tell you:
• Relative local “neural” activity (LFP’s ?)• NOT absolute neural activity• NOT excitation vs inhibition• NOT about necessity of a given region for a task• NOT fine-grained temporal information
![Page 8: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/8.jpg)
Key Points
• What can fMRI tell you?
• Always comparing across conditions
• Characteristics of the hemodynamic response(HRF) and how this affected the sequentialdevelopment of fMRI paradigms and influencesstudy design
• Sense of important design issues
![Page 9: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/9.jpg)
Subtraction ParadigmDonder’s method:
Ex: How to measure time of a mental transformation?
A random series of A’s and B’s presented and the subject must:
1. Respond whenever an event occurs (RTi)
2. Respond only to A not to B (RTii)
3. Respond X to A and Y to B (RTiii)
RTi = RT(detect) + RT(response)
RTii = RT(detect) + RT(discrimination) + RT(response)
RTiii = RT(detect) + RT(discrimination) + RT(choice) + RT(response)
THUS, RT(discrimination) = RTii - RTi
RT(choice) = RTiii - RTii
![Page 10: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/10.jpg)
Criticisms of Subtraction Paradigm
1. That we already know what ‘counts’ as a single mentalprocess (i.e. choice is a single mental process?)
2. Assume that adding components does not affect otherprocesses (i.e. assumption of pure insertion)
THUS, one should pick tasks that differ along ONEdimension (either change the task OR the stimuli but notBOTH!)
And a resting baseline is good to include, however, theinterpretation should be taken lightly…(more later)
![Page 11: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/11.jpg)
The loose task comparisonThe loose task comparison
Does Does notnot hold all variables constant BUT: hold all variables constant BUT:
(1) Uses a low level reference task(1) Uses a low level reference task
(2) Allows the data to be examined for predictable(2) Allows the data to be examined for predictablestimulus or response drivenstimulus or response driven activations activations
(3) Allows the more extensive activation pattern to(3) Allows the more extensive activation pattern tobe observedbe observed
![Page 12: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/12.jpg)
The “loose” TaskThe “loose” TaskComparisonComparison
STISTISTIPREPREPRE
DOGDOGDOG
+++ +++ +++ +++
GREGREGRE
STISTISTIPREPREPRE
PAINTPAINTPAINT STRSTRSTR
STISTISTIPREPREPRE
EGGEGGEGGTASK 1TASK 1
TASK 2TASK 2
![Page 13: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/13.jpg)
The tight task comparisonThe tight task comparison
Try to hold all variables constant including:Try to hold all variables constant including:
•• Stimulus display (nominally or statistically) Stimulus display (nominally or statistically)
•• Response and response selection characteristics Response and response selection characteristics
•• Performance level- especially if comparing cohorts Performance level- especially if comparing cohorts
•• Eye movements Eye movements
•• Emotional state (minimize anxiety and boredom) Emotional state (minimize anxiety and boredom)
![Page 14: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/14.jpg)
The “tight” TaskThe “tight” TaskComparisonComparison
STISTISTIPREPREPRE
DOGDOGDOG
+++ +++ +++ +++
GREGREGRE
STISTISTIPREPREPRE
PAINTPAINTPAINT STRSTRSTR
STISTISTIPREPREPRE
EGGEGGEGGTASK 1TASK 1
TASK 3TASK 3
TASK 2TASK 2
![Page 15: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/15.jpg)
22 mmiinnuuss 11TASK 1TASK 1 TASK 2TASK 2
BRAIN AREAS THAT DIFFERBRAIN AREAS THAT DIFFER
ALL ACTIVE BRAINALL ACTIVE BRAINAREASAREAS
Thanks to Randy Buckner
![Page 16: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/16.jpg)
Example...
Interested in semantic processing and how it affectsmemory...
![Page 17: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/17.jpg)
Parameters to specify in any experiment
1. Subjects: normal vs special populations
2. What part of brain look at? How many slices canyou have for your TR?
3. Choosing your TR: How often can you take a full setof pictures
4. What coil will you use?surface coils: higher SNR, only partial coveragehead coils: lower SNR, complete coverage
5. Toggle many times between conditions within a scan
6. Run as many scans as possible within a subject
![Page 18: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/18.jpg)
Key Points
• What can fMRI tell you?
• Always comparing across conditions
• Characteristics of the hemodynamic response(HRF) and how this affected the sequentialdevelopment of fMRI paradigms and influencesstudy design
• Sense of important design issues
![Page 19: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/19.jpg)
Visual Stimulation - 2 sec Flashes
0 10 20 30 40 50 60 70 80 90 100 110 120
TIME (seconds)
Per
cent
Sig
nal C
hang
e
0.0
1.0
2.0
3.0
4.0
-1.0
[Blamire, Ogawa et al., PNAS, 1992]
![Page 20: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/20.jpg)
Visual Cortex During Brief Visual Stimulation
0.0
1.0
2.0
5 0 5 10 15
34
100
1000msec
TIME (seconds)
msec
msec
[Courtesy of Robert Savoy, Kathleen O’Craven MGH-NMR Center]
![Page 21: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/21.jpg)
Blocked design fMRI
BLOCKED:
HORSE
(abstract or concrete?)
love
(upper or lowercase?)
![Page 22: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/22.jpg)
“Blocked” fMRI:Memory Paradigm
STIPRE
HORSE
+ +
0 8 74 118 140 184 206 250
GRE
STIPRE
LOVE STR
STIPRE
HUMOR
TIME (SEC)52
+ +
STR
STIPRE
CHAIR
+
[Wagner et al., OHBM, 1998]
![Page 23: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/23.jpg)
Typical Blocked-Design Response
-0.25
0
0.25
0.50
Per
cent
Sig
nal C
hang
e
0 20 40 60 80 100 120 140 160 180 200 220 240
Time
HORSE LOVE HUMOR CHAIR
![Page 24: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/24.jpg)
13 SlicesPer
Brain Image
Thanks to Robert Savoy
![Page 25: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/25.jpg)
80 brain imagesper 4 minute
run
Thanks to Robert Savoy
![Page 26: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/26.jpg)
For purposes of illustration…….
Thanks to Robert Savoy
![Page 27: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/27.jpg)
Examine thedata fromone slice
of the brainas a function
of time
Thanks to Robert Savoy
![Page 28: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/28.jpg)
Thanks to Robert Savoy
![Page 29: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/29.jpg)
Are these voxellevels statistically
different from each other?
Combinethese
Combinethese
Thanks to Robert Savoy
![Page 30: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/30.jpg)
Combinethese
Combinethese
Consider EACH voxelacross all time points
Thanks to Robert Savoy
![Page 31: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/31.jpg)
Typical Blocked-Design Response
-0.25
0
0.25
0.50
Per
cent
Sig
nal C
hang
e
0 20 40 60 80 100 120 140 160 180 200 220 240
Time
HORSE LOVE HUMOR CHAIR
![Page 32: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/32.jpg)
Event-Related fMRI
BLOCKED:
SPACED EVENT-RELATED:
16 sec
![Page 33: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/33.jpg)
“Spaced Event-Related” fMRI:Language Paradigm
[Buckner, Bandettini et al., PNAS, 1996]
TIME (seconds)
0 32 64 96
COU PRE TRA AFT STA PEL DRI
![Page 34: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/34.jpg)
“Single-Trial” Response Across a Run“Single-Trial” Response Across a Run
TIME (SEC)TIME (SEC)
00 3232 6464 9696 128128 160160 192192 228228 256256
375375
376376
377377
378378
379379
380380
381381
ME
AN
MR
SIG
NA
LM
EA
N M
R S
IGN
AL
![Page 35: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/35.jpg)
“Event-Related” Selectively Averaged Response
0
.5
0 1 2 3 4 5 6 7 8 9 101112131415
TIME (seconds)
PE
RC
EN
TS
IGN
AL
CH
AN
GE
375
376
377
378
379
380
381
ME
AN
MR
SIG
NA
L
![Page 36: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/36.jpg)
374
375
376
377
378
379
0 2 4 6 8 10 12 14TIME (sec)
Broca’s Area During Language Paradigm
Thanks to Randy Buckner
![Page 37: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/37.jpg)
“Rapid Event-Related” fMRI
BLOCKED:
SPACED EVENT-RELATED:
16 sec
RAPID EVENT-RELATED:2 sec
![Page 38: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/38.jpg)
20 sec0 sec
1 secon
Assessing the Linearity Hypothesis
[Dale and Buckner, Hum. Brain Map., 1997]
![Page 39: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/39.jpg)
Response to Averaged Single Trials
-1
0
1
2
3
4
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19TIME (sec)
PE
RC
EN
T M
R S
IGN
AL
Split-half reliability
Thanks to Randy Buckner
![Page 40: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/40.jpg)
20 sec0 sec
20 sec0 sec 5 sec
Assessing the Linearity Hypothesis:5 Second ITI
Thanks to Randy Buckner
![Page 41: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/41.jpg)
-1
0
1
2
3
4
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19TIME (sec)
PE
RC
EN
T M
R S
IGN
AL
Response to Averaged Double Trials
Thanks to Randy Buckner
![Page 42: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/42.jpg)
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
TIME (SEC)0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
TIME (SEC)
Assessing the Linearity Hypothesis:Separation of Responses
Thanks to Randy Buckner
![Page 43: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/43.jpg)
-1
0
1
2
3
4
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19TIME (sec)
Assessing the Linearity Hypothesis:Separation of Responses
FIRST TRIAL
ESTIMATEDSECOND TRIAL
Thanks to Randy Buckner
![Page 44: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/44.jpg)
20 sec0 sec
0 sec2 sec 20 sec
0 sec2 sec 20 sec4 sec
Assessing the Linearity Hypothesis:2 Second ITI
Thanks to Randy Buckner
![Page 45: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/45.jpg)
-1
0
1
2
3
4
5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19TIME (SEC)
-1
0
1
2
3
4
5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19TIME (SEC)
RAW DATA ESTIMATED RESPONSES
Responses to Multiple Rapidly Intermixed Trials
THREE TRIAL
TWOTRIAL
ONETRIAL
FIRST TRIAL
ESTIMATEDSECOND TRIAL
ESTIMATEDTHIRD TRIAL
Thanks to Randy Buckner
![Page 46: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/46.jpg)
Structuring Event-Related Trial Presentations
[Burock et al., Neuroreport 1998]
Fixed Interval Presentation
Randomized Presentation
ITI = 16 sec 3 sec 1 sec
Nor
mal
ized
% S
igna
l Cha
nge
8
6
4
2
0
8
6
4
2
00 100 200 300 400
Time (sec)
![Page 47: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/47.jpg)
Variance Associated with Fixed Interval Designs
Seven unknowns, BUTonly three independentequations
INDIVIDUALBOLD
RESPONSE
TIME
h7 + h4 + h1
MEASUREDBOLD
RESPONSE
h6 + h3
h5 + h2
h4 + h1 + h7
h3 + h6 h2 + h5
h1 + h4 + h7
= TR
6 12 18 24 30
Thanks to Randy Buckner
![Page 48: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/48.jpg)
Variance Associated with Jittered Designs
INDIVIDUALBOLD
RESPONSE
TIME
Seven unknowns, ANDmore than sevenindependent equations
MEASUREDBOLD
RESPONSE
h7 + h5 + h3
h6 + h4 + h2
h5 + h3 + h1
h4 + h2
h3 + h1 + h7 h2 + h6
h1 + h5
h7 + h4
h6 + h3
Thanks to Randy Buckner
![Page 49: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/49.jpg)
Does the neural correlate of priming vary with the lagbetween the first and second episode within asemantic task?
PEACE+
ANVIL
PEACE+
TABLE
+ENERGY
ANVIL~2 min
~25 hr
ABSTRACT or CONCRETE?
Sorting Based onExperimenter Determined Conditions
[Wagner et al., J. Cognitive Neuroscience 2000]
![Page 50: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/50.jpg)
Rest
Rest
Novel
Long Prime
Short Prime
Long Prime
Short Prime
Novel
![Page 51: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/51.jpg)
Rest
Rest
Novel
Long Prime
Short Prime
Long Prime
Short Prime
Novel
Group datatogether
![Page 52: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/52.jpg)
Shorter Lags Yield Greater Neural Priming
Anterior IFG
-0.1
0
0.1
0.2
0.3
% SignalChange
0 2 4 6 8 10 12 14Time (s)
ShortLongNovel
Thanks to Anthony Wagner
![Page 53: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/53.jpg)
TRUTH
HOUSE
PEACE
ANVIL
TAPEREMEMBERED
FORGOTTEN
Sorting Based on Subject Behavior:Subsequent Memory Performance
[Wagner et al., Science 1998]
![Page 54: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/54.jpg)
Neural Regions Predicting Subsequent Memory
ForgottenRemembered
Inferior Prefrontal Gyrus
-1
01234
0 2 4 6 8 10 12 14Time (s)
-101234
0 2 4 6 8 10 12 14Time (s)
B
Left Posterior Parahippocampus
Thanks to Anthony Wagner
![Page 55: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/55.jpg)
Key Points
• What can fMRI tell you?
• Always comparing across conditions
• Characteristics of the hemodynamic response(HRF) and how this affected the sequentialdevelopment of fMRI paradigms and influencesstudy design
• Sense of important design issues
![Page 56: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/56.jpg)
Critical issues in paradigm designCritical issues in paradigm design
•• Poorly defined Poorly defined neuroanatomical neuroanatomical hypothesishypothesis
•• Poorly controlled baseline Poorly controlled baseline
•• Attentional Attentional effects effects
•• Learning effects Learning effects•• Stimulus habituation or sensitization Stimulus habituation or sensitization
•• System and physiological drift System and physiological drift
![Page 57: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/57.jpg)
Baseline, what is it?
Ex: if want to say something about verb generation andcompare it only to reading aloud..
BUT, still do not know if these regions are involved inreading only (thus can include a low level reading condition..)
No inherent “0” baseline for cognition,i.e. what are subjects doing when asked to do nothing?
• Ans: they are doing a lot• how interpret deactivations?
![Page 58: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/58.jpg)
Issues: Generality vs SpecificityHypothesis: Region X is involved in process Y.
Evidence: Region X is activated when subjects do an instanceof process Y
Problem: Without running several further conditions, we can’ttell whether region X might instead be involved in somethingeither more SPECIFIC or more GENERAL than Y.
Example:
Hypothesis Space: Function of Region X
hands
feeteyes
2-toned faces
facesProfileFaces
FacesW/ohair
IDphotos
Any human body part
Anything animate
Thanks to Nancy Kanwisher
![Page 59: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/59.jpg)
Issues: Attentional ConfoundsA given region might respond more strongly in condition A thancondition B simply because A is more interesting/attention-capturing than B.
Solutions:
1. Double Dissociations, i.e. faces versus objects?
2. Test conditions with opposite attentional predictionsi.e. passive viewing vs 1-back task
Thanks to Nancy Kanwisher
![Page 60: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/60.jpg)
Issues: Statistical Significance vs.Theoretical Significance
P levels alone are not sufficientFor example, the FFA may respond significantly more to pineapplesthan watermelons, but the response to pineapples might nonethelessbe much lower than the response to faces.
Solutions:Quantity effect size, e.g. with percent signal change
Provide “benchmark” conditions within the same scan to give thesemagnitudes meaning
Objects Watermelon Pineapple Faces
0.6
0.6
0.7
0.7
0.9
1.8
2.0
2.0Thanks to Nancy Kanwisher
![Page 61: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/61.jpg)
Acknowledgements
Anthony Wagner
Nancy Kanwisher
Randy Buckner
Robert Savoy
Randy Gollub
Chantal Stern
![Page 62: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/62.jpg)
Data AnalysisA. General Issues• Individual vs Group Analyses: brains are very different BUT wantTo make a general claim ANS: do both if can
• Multiple Comparisons….if doing 20,000 T-tests, better not acceptp< .05
B. Methods• Simple comparisons, is X > Y ?, look in each voxel..
• Conjunction Analyses, are any voxels significant for both X>Yand A>B?
• Regression Analyses, obtaining weights for different regressors
• ROI-based Analyses
![Page 63: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/63.jpg)
[Boyton et al., J. Neuroscience 1996]
Hemodynamic Response Summation:Linear Systems Approach
The fMRI response to a stimulus lasting a duration of NT is roughly alinear summation of N temporally shifted responses to a stimulus lastinga duration of T
![Page 64: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/64.jpg)
“Mixed” fMRI
BLOCKED:
RAPID EVENT-RELATED:2 sec
MIXED:
[Chawla et al., Nat. Neuroscience 1999; Donaldson et al., NeuroImage, 2001]
![Page 65: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/65.jpg)
TIME (SEC)
= OLD WORD = NEW WORD
MEMORY
FIXATION
MEMORYA
B
FIXATION FIXATION
Mixed Blocked/Event-relatedDesign
Donaldson et al., NeuroImage, 2001 (see also Chawla et al., Nature Neurosci., 1999)
![Page 66: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/66.jpg)
TIME
TRANSIENT BOLDRESPONSETO EVENTS
MEASUREDBOLD
RESPONSE
SUSTAINED BOLDRESPONSE
TO SET
TIME
![Page 67: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/67.jpg)
“Mixed” fMRI:Trial Separation with Task Blocking
Task 1 Task 2 Task 1
Analysis Strategies:
– Event-related analyses
• Task 1 trials ( ) vs. Task 2 trials ( )
• Trial type A ( ) vs. Trial type B ( )
[Badre et al., in prep.]
![Page 68: FMRI Experimental Design - MITweb.mit.edu/hst.583/www/course2001/LECTURES/hst583_lect...Because fMRI BOLD data is not an absolute measure of neuronal activity, all study designs must](https://reader033.fdocuments.in/reader033/viewer/2022052722/5f0bdf667e708231d432a1fb/html5/thumbnails/68.jpg)
“Mixed” fMRI:Trial Separation with Task Blocking
Task 1 Task 2 Task 1
Analysis Strategies:
– Event and State effects• Same event contrasts
• Also model NULL components within blocks to explore “state” effects
CAVEAT: correlation between event and state regressors
vs vs
Task 1 vs. Task 2 Nulls