Hands-On Data Analysis
 Kate Pirog Revill and Chris Rorden
 Data from safety training
–
–
–
–
9 subjects
Finger-tapping task (12s tapping, 12s rest)
188 scans acquired over 6:16 (TR = 2)
Hi-res structural image
 Single-subject analysis in FSL
 Single-subject analysis in SPM8
 Group analysis in FSL
 Group analysis in SPM8
1
Finding the Data
Open terminal window
– Navigate to directory home/public/coursedemo
– Choose analysis technique (/spm or /fsl)
– Choose subject (/subj1 through /subj9)
Data already converted from dicom format
– Hi-res structural: T1mprages002a1001…
– Functional: fmri12secon12secoff…
– timing.txt (fsl folders only)
Get it from the web http://www.cabiatl.com/Resources/FSL/subj9.zip
2
FSL: Single Subject
Launch FSL: fsl &
Data tab
– First, must select
functional dataset:
Select 4D data
3
FSL: Single Subject
 Click on folder to
open file selection
dialog
 Select fmri12sec…
file
 Click ok in both
boxes
4
FSL: Single Subject
Data tab:
– Set TR = 2
– Set high-pass cutoff
to 48 (2 x block
size)
– Delete 0 volumes
– Total volumes and
output directory set
automatically
5
FSL: Single Subject
Pre-Stats tab
– Set spatial
smoothing to 8mm
– Other settings ok
6
FSL: Single Subject
Stats Tab
– Don’t add motion
parameters
Choose full model
setup
7
FSL: Single Subject
One explanatory
variable: motor
– Custom format
Choose timing.txt
(3 column file with
block onsets,
durations, and
strengths)
– Uncheck temporal
derivative (block
design)
8
FSL: Single Subject
 Contrasts tab
– One contrast: areas
more active during
the motor task than
during rest
– Name it motor>rest
 Click Done, model
will appear: does it
make sense?
– (click on it to make it
disappear)
9
FSL: Single Subject
Registration
– Make sure Normal
search, 12 DOF are
selected
– Could add hi-res
structural here, but
won’t to save time
& preprocessing
steps
10
FSL: Single Subject
Save as design.fsf
Click Go to start
analysis
Web browser will
open to track
progress (takes ~
10 minutes)
11
FSL: Single Subject
 Done? Check it out…
 Click on blue links to
view results:
– Pre-stats: motion
parameters
– Stats: model design
– Post-stats: activation
map and voxel fits
– Registration: alignment
of functional image to
template
12
FSL: Group Analysis
All individual subject analyses must be
completed first
Navigate to /home/public/coursedemo/fsl and
create your own folder for group analysis
(name it after yourself). Enter this folder and
launch FSL (fsl &)
13
FSL: Group Analysis
 Set tab at top left to Higherlevel analysis
 Data Tab:
– Inputs are FEAT directories
– 9 inputs
– Output directory should be
ok if you launched FSL from
the directory I told you to
create for yourself!
14
FSL: Group Analysis
 Click on folders to get to file selection dialog and
select the fmri...feat directory for each participant
15
FSL: Group Analysis
Stats Tab
– Make sure to choose
Mixed effects: Simple
OLS
MUCH faster- doesn’t
take subject variability
into account
– Choose Full model
setup
16
FSL: Group Analysis
EVs Tab
– Only 1 group of subjects
– Want to see motor
activity for all subjects,
so set EV to 1 for each
input
17
FSL: Group Analysis
Contrasts Tab
– Only one
contrast (where
all subjects show
activity)
Click Done to
view design
18
FSL: Group Analysis
Post-Stats tab
– Lower Z
threshold to 1.64
(5%)
– Feel free to
repeat analysis
later w/ different
thresholds..
19
FSL: Group Analysis
Save as
groupdesign.fsf
Click GO to start
analysis (will launch
web browser to monitor
progress)
– Takes about 5 minutes
20
FSL: Group Analysis Results
21
Finding the Data
SPM pipeline
– Preprocess: realign (motion correct), coregister
(align T1 to fMRI), normalize (warp to standard
space), smooth
– Conduct statistics.
Data already converted from dicom using
dcm2nii
– File in /home/public/coursedemo/subj1...subj9
– Hi-res structural: T1mprage…
– Functional: fmri12secon12secoff…
Get it from the web http://www.cabiatl.com/Resources/SPM/subj9.zip
22
SPM: Single Subject
 Again, get into proper
directory:
cd /home/public/coursedemo/spm/subj9
 Launch MATLAB
matlab &
 At prompt, launch SPM8
spm8
(“spm fmri” outside CABI)
23
SPM: Single Subject
Select Realign:
Estimate & Reslice
from spatial
preprocessing bar
Realign: Estimate &
Reslice
– Click New:Session
24
SPM: Single Subject
Realign: Estimate &
Reslice
– Highlight Session and
click on Select Files
to open dialog box
25
SPM: Single Subject
 Select all
fmri12sec…
images (hold down
shift)
– Hint: type ^f.* in
filter box to only
show files starting
with f
– 188 files selected
– Click Done
26
SPM: Single Subject
Realign: Estimate &
Reslice
– Leave all options as
defaults
– Save as realign.mat
– Press green arrow to
run (2 min)
– Close batch editor
27
SPM: Single Subject
 Progress shown in
lower left window
 Results of each step
generally displayed in
Graphics window
28
SPM: Single Subject
 Select Coregister:
Estimate
 Coregistration
– Choose reference image:
meanfmri12secon…
– Choose source image:
T1mprages…
– Save as coreg.mat
– Click green arrow to run (2
min)
– Click around results to
check alignment
29
SPM: Single Subject
Segmentation: find
grey and white
matter in hi-res
structural image
(improves
normalization)
30
SPM: Single Subject
Segment
– Highlight Data and
select file
T1mprages…
– Save as segment.mat
– Run (~ 5 minutes)
To view results, click
Display
“c1T1mpr” is gray
matter, “c2T1mpr” is
white matter
31
SPM: Single Subject
Normalize: warp
images to standard
space (functionals
and structural)
Choose
Normalize:Write
32
SPM: Single Subject
Normalize
– Click on New:Subject
– Choose parameter
file:
 T1mpr…_seg_sn.mat
– Choose Images to
Write:
all 188 rfmri… files
33
SPM: Single Subject
Normalize
– Set Voxel Sizes to 3 3 3
– Save as func_norm.mat
– Run (~ 1 min)
Also normalize structural
image
– Same procedure except
Images to Write =
T1mprages
Voxel Sizes = 1 1 1
34
SPM: Single Subject
To check results of
normalization,
choose Check Reg,
then select
wT1mprage… and
one of the wrfmri…
images and see if
they match up well
35
SPM: Single Subject
Smoothing
– Highlight Images to
Smooth and select all
188 wrfmri… files
– Check that FWHM =
[8 8 8]
– Save as smooth.mat
– Run (~ 3 min)
No display
36
SPM: Single Subject
 Click Specify 1st-level
 1st-level
– Highlight Directory and select ./
(current directory) for saving
SPM.mat file
– Timing Parameters
 Units: Seconds
 Interscan interval: 2
– Data and Design: New
Subject/Session
 Highlight Scans and choose all
188 swrfmri… files
37
SPM: Single Subject
1st-level
– Conditions: New
Condition
Name: motor
Onsets: can enter as a
list of start times but
can also use MATLAB
notation  0:24:360
38
SPM: Single Subject
1st-level
– Durations: 12 (all
blocks are 12s long,
so only one value
needed)
– High Pass Filter: 48
– Save as design.mat
– Run (1min)
You will see a design
matrix when
completed
39
SPM: Single Subject
Estimate
– Choose SPM.mat file
– Run (~ 2 minutes)
40
SPM: Single Subject
Results
– Click on Results
button in Inference
bar
Lets define our
statistical contrast…
– Select SPM.mat
– Choose Define new
contrast…
41
SPM: Single Subject
Results
– Define contrast:
Name: motor>rest
Type: t-contrast
Contrast: enter 1 and
press Submit
Hit OK.
Then hit Done in
previous window
42
SPM: Single Subject
Results
– Mask with other contrasts: no
– Title: motor>rest
– P value adjustment: FWE
– P value: 0.05
– Extent threshold: 0
– Will run for < 1 min…
View Results
– Right-click on glass brain to
find global maximum voxel
43
SPM: Single Subject
 Things to try
– Overlay activation on
structural image:
 Overlays: sections
 Select wT1mprage.. file
– Select max voxel and plot
timecourse:
 Plot: Plot… Fitted Responses
 Predicted
 Against scan or time
44
SPM: Group Analysis
 Navigate to
/home/public/spm and
make your own group
directory (i.e. kategroup)
 Launch MATLAB and
spm8.
 Choose Specify 2ndLevel
45
SPM: Group Analysis
 2nd-Level
– Choose a results
directory (the one you
just made)
– Scans: navigate to each
subject’s directory and
choose con_0001.img
 Don’t hit OK until you’ve
completed all 9 subjects
– Save as groupdesign.mat
– Run (~2 min)
 The resulting matrix is a
white box (same effect for
all subjects).
46
SPM: Group Analysis
 Estimate
– Choose SPM.mat from
your group directory
– Run (~ 2 min)
 Results
– Choose SPM.mat from
your group directory
 Define new contrast
 Name: group_m>r
 T-contrast
 Contrast: 1
 Hit submit, OK, done
47
SPM: Group Analysis
Results
– Mask with other
contrasts: no
– Title: motor>rest
– P value adjustment:
FWE
– P value: 0.05
– Extent threshold: 0
– Will run for < 1 min…
48