1. Group Analyses in fMRI Jody Culham Brain and Mind Institute
Department of Psychology
Western University
Group Analyses in fMRI
Last Update: November 21, 2016
Last Course: Psychology 9223, F2016, Western University

2. Group Analyses 1. Get all the subjects’ brains into a common space
Talairach space
MNI space
cortex-based alignment
2A. Do group statistics
Random Effects GLM
and/or
2B. Use a Region of Interest Approach

3. Combining Group Data

4. Brains are Heterogeneous
Slide from Duke course

5. Talairach Coordinate System
Talairach & Tournoux, 1988
made an atlas based on one brain
any brain can be squished or stretched to fit hers and locations can be described using a 3D coordinate system (x, y, z)
… from an alcoholic old lady
Note: That’s TalAIRach, not TAILarach!

6. Rotate brain into ACPC plane
Find anterior commisure (AC)
Corpus Callosum
Fornix
Find posterior commisure (PC)
ACPC line
= horizontal axis
Pineal Body
“bent asparagus”
Note: official Tal says to use top of AC and bottom of PC but I suspect few people actually do this
Source: Duvernoy, 1999

7. AC = Origin
x = 0 +
y = 0 z
z = 0 - L R + y - - + x

8. Left is what?!!! L R R L Neurologic (i.e. sensible) convention
left is left, right is right L R
x = 0 - +
Note: Make sure you know what your magnet and software are doing before publishing left/right info!
Radiologic (i.e. stupid) convention
left is right, right is left R L

9. Tip
Put a vitamin E capsule on the one side of the subject’s head or coil.

10. Deform brain into Talairach space
Mark 8 points in the brain:
anterior commisure
posterior commisure
front
back
top
bottom (of temporal lobe)
left
right
Squish or stretch brain to fit in “shoebox” of Tal system
ACPC=0
y>0
y<0 z y
AC=0
y>0
y<0
x (left < 0; right > 0)
Extract 3 coordinates

11. Talairach Tables
Source: Culham et al., 2003, Exp. Brain Res.
Talairach coordinates can be useful for other people to check whether their activation foci are similar to yours
Often it’s easiest to just put coordinates in a table to avoid cluttering text

12. Do We need a “Tarailach Atras”?
Variability between Japanese
and European brains, both male
(red > yellow > green > blue)
Variability between male and female brains, both European
Source: Zilles et al., 2001, NeuroImage

13. Talairach Pros and Cons
Advantages
widespread system
allows averaging of fMRI data between subjects
allows researchers to compare activation foci
relatively easy to use
Disadvantages
does better for central regions of cortex, but not great for most of cortex
not appropriate for all brains (e.g., group variability, patients may not fit well)
ignores left- vs. right-hemisphere asymmetries
activation foci can vary considerably – other landmarks like sulci may be more reliable

14. MNI Space
Researchers at the Montreal Neurological Institute (MNI) created a better template based on a morphed average of hundreds of brains
many different versions

15. MNI Space
Benefits
MNI Space is based on many subjects not just one brain like Talairach space
MNI transformations use nonlinear warping, which leads to better intersubject alignment
Image Source

16. Converting Between MNI and Tal space
You many want to convert between the systems
Only Tal provides Brodmann’s areas
Need to convert for meta-analyses
The MNI and Talairach coordinates are similar but not identical
e.g., temporal lobes extend 10 mm lower in MNI brain
Caveat: careful comparison requires a transformation
-- converters can be found online
Source:

17. Anatomical Localization Sulci and Gyri
gray matter
(dendrites &
synapses)
white matter
(axons)
FUNDUS
BANK
GYRUS
SULCUS
CROWN
pial surface
gray/white border
neuron
SULCUS
FISSURE
GYRUS
Source: Ludwig & Klingler, 1956,
in Tamraz & Comair, 2000

18. Variability of Sulci
Source: Szikla et al., 1977, in Tamraz & Comair, 2000

19. Effects of Sulcal Variability
Source: Frost & Goebel, 2012, NeuroImage

20. Variability of Functional Areas
Watson et al., 1995
- motion-selective area, MT+ (=V5) is quite variable in stereotaxic space
however, the area is quite consistent in its location relative to sulci
junction of inferior temporal sulcus and lateral occipital sulcus
see also Dumoulin et al., 2000

21. Cerebral cortex is a crumpled map

22. Cortical Surfaces segment gray-white matter boundary
render cortical surface
inflate cortical surface
sulci = concave = dark gray
gyri = convex = light gray

23. Cortical Inflation Movie
Movie: unfoldorig.mpeg
Source: Marty Sereno’s web page

24. Cortical Flattening Source: Brain Voyager Getting Started Guide
2) make cuts along the medial surface
(Note, one cut typically goes along the fundus of the calcarine sulcus though in this example the cut was placed below)
1) inflate the brain
3) unfold the medial surface so the cortical surface lies flat
4) correct for the distortions so that the true cortical distances are preseved
Source: Brain Voyager Getting Started Guide

25. Spherical Averaging
Future directions of fMRI: Use cortical surface mapping coordinates
Inflate the brain into a sphere
Use sulci and/or functional areas to match subject’s data to template
Cite “latitude” & “longitude” of spherical coordinates
Source: Fischl et al., 1999

26. Spherical Averaging Source: Fischl et al., 1999
Movie: brain2ellipse.mpeg
Source: Marty Sereno’s web page
Movie: morph-curv1.mpg
Source: Marty Sereno’s web page
Source: Fischl et al., 1999

27. Before and After CBA
Source: Frost & Goebel, 2012, NeuroImage

28. Before and After CBA hand motor area (M1) hand somatosensory area (S1)
Source: Frost & Goebel, 2012, NeuroImage

29. Gains in Overlap
Source: Frost & Goebel, 2012, NeuroImage

30. Voxelwise Group Analyses
Fixed vs. Random Effects

31. Example Three subjects Three conditions: Baseline, Faces, Objects
For simplicity, just consider one voxel in FFA

32. Stupid Way: Concatenated Fixed Effects (FFX)
Concatenate data, pretend you have one subject who did three runs

33. Stupid Way: Concatenated Fixed Effects (FFX)
Make one predictor for Faces and one for Objects (2 df)
Scale predictors (by beta weights)
Note why this is stupid
Assumes all subject show same magnitude of activation
Errors in this assumption   residuals

34. Better Way: Separate Subjects FFX
Don’t concatenate
Separate predictors for 3 subjects x 2 conditions
6 df

35. Problem: Separate Subjects FFX
We could do business-as-usual GLM and see if predictors account for significant variance considering noise
Effectively, we are asking how confident we are that this effect is true (not due to chance) in these three subjects (and only these three subjects)
BUT usually, we want to generalize to the population from which we sampled

36. Best Way: Random Effects (RFX)

37. Second-level analysis
Nothing too complicated… it’s effectively just a paired t-test
Subject
βFaces
βObjects
Difference
βFaces - βObjects S1
0.552
0.105
0.447 S2
2.061
1.121
0.940 S3
1.019
0.247
0.772
Mean
0.719 SD
0.250
SEM [=SD/sqrt(N)]
0.145
tcrit(df=3)
4.3
95%CI lower (= mean – (t*SEM))
0.097
95%CI lower (= mean + (t*SEM))
1.34
Estimated Distribution
of Differences
does not include zero
 significant (p<.05)
0.719
0.097
1.34
95% CI

38. Repeat for the other 60,000 voxels…
Huettel, Song & McCarthy, 2008
First-level analysis
Second-level analysis

39. If you really want to be correct…
We often refer to this type of analysis as random effects (RFX)
Since subjects is a random effect but other aspects (e.g., stimulus categories) are fixed effects, technically the proper term is Mixed Effects Analysis
Other common jargon = Second-level Analysis

40. Take-home Message
RFX enables us to generalize to the population from which we sampled subjects
Degrees of freedom comes from number of subjects, not number of time points
No need to worry about correction for serial correlations
for most fMRI studies, this means underpaid graduate students in need of a few bucks!

41. Examples from a real data set

42. Concatenated FFX Example 17 Subjects x 2 runs with Faces & Houses
df = 17 Ss x 2 runs/S x 264 vols/run - 1
one predictor per condition

43. FFX Separate Subjects …
If you’re looking at pilot data from a few Ss, with RFX it will be hard to see any effects.
You can do FFX with separate subjects.
It has the same problems as concatenated FFX but at least enables you to examine the consistency between Ss

44. Now that our df no longer depends on # volumes,
RFX S1 S2 S3 …
S17
This contrast is just like doing a paired t-test between Faces and Objects with 17 Ss
df = 17 Ss - 1
Now that our df no longer depends on # volumes,
we don’t have to worry about correction for serial correlations with RFX

45. The Problem of Intersubject Variability
The hand area lies in the “hand knob”, a section of the central sulcus that looks like an upside-down omega
Primary Motor Cortex (M1) lies along the anterior bank while Primary Somatosensory Cortex (S1) lies along the posterior bank Ω Ω
2+ cm
Although these areas are reliably localized with respect to sulcal landmarks within individuals, the sulcus has high anatomical variability between individuals

46. The Problem of Intersubject Variability
The hand area lies in the “hand knob”, a section of the central sulcus that looks like an upside-down omega
Primary Motor Cortex (M1) lies along the anterior bank while Primary Somatosensory Cortex (S1) lies along the posterior bank
Although these areas are reliably localized with respect to sulcal landmarks within individuals, the sulcus has high anatomical variability between individuals
Due to this variability, activation of the same area (e.g., the motor hand area during finger tapping) can appear scattered in stereotaxic (Talairach or MNI) space. Given that random effects (RFX) analyses require somewhat consistent effects across participants, this can handicap statistical power
This also means that it can be tricky to isolate effects in one area (e.g., M1) without contamination from adjacent areas (e.g., S1) Ω Ω
S1 β = 3
S2 β = 0
S3 β = 0
RFX non-sig S1 S2 S3
3 Ss superimposed

47. Three Solutions to Intersubject Variability
RFX non-sig
S1 β = 2.5
S2 β = 2.5
S3 β = 1
RFX sig
Solution 1:
Spatial Smoothing
increase likelihood of consistent effects across Ss
doesn’t solve problem that voxels likely contain combination of different areas (M1 and S1 in this case)
Solution 2:
Cortex-based Alignment
solution works very well and makes it less likely to blur across adjacent functional areas
generation of surfaces can be time-consuming (though automation is getting better)
cortex-centric
S1 β = 3
S2 β = 3
S3 β = 3
RFX highly sig S1 S2 S3
Solution 3:
Individual ROIs
identify ROI anatomically in each S (e.g., based on anterior bank of hand knob) or functionally in each S (based on functional localizer)
anatomical ROIs can require manual labour
functional ROIs can be somewhat subjective
doesn’t work for unexpected activation sites
S1 β = 3
S2 β = 3
S3 β = 3
RFX highly sig
Original Data

48. Random Effects Analysis
Brain Voyager recommends you don’t even toy with random effects unless you’ve got 10 or more subjects (and 50+ is best)
Random effects analyses can really squash your data, especially if you don’t have many subjects.
Though standards were lower in the early days of fMRI, today it’s virtually impossible to publish any group voxelwise data without RFX analysis

49. Strategies for Exploration vs. Publication
Deductive approach
Have a specific hypothesis/contrast planned
Run all your subjects
Run the stats as planned
Publish
Inductive approach
Run a few subjects to see if you’re on the right track
Spend a lot of time exploring the pilot data for interesting patterns
“Find the story” in the data
You may even change the experiment, run additional subjects, or run a follow-up experiment to chase the story
While you need to use rigorous corrections for publication, do not be overly conservative when exploring pilot data or you might miss interesting trends
Random effects analyses can be quite conservative so you may want to do exploratory analyses with fixed effects (and then run more subjects if needed so you can publish random effects)

50. How can we identify activation foci?
Talairach coordinates
Example: The FFA is at x = 40, y = -55, z = -10
Anatomical localization
Example: The FFA is in the right fusiform gyrus at the level of the occipitotemporal junction
Functional localization
Example: The FFA includes all voxels around the fusiform gyrus that are activated by the comparison between faces and objects
Kanwisher, McDermott & Chun,
1997, J Neurosci

51. Talairach Daemon

52. Neurosynth
Neurosynth (Dec. 2014)

53. Automated Coordinate Extraction
Yarkoni et al., 2001

54. Example: Working Memory
META-
ANALYSIS
from previous studies
FORWARD INFERENCE
What is the probability of activation here a study involving working memory?
REVERSE INFERENCE
If there is activation here, what is the probability the study involved working memory?
Yarkoni et al., 2001

55. Automated Coordinate Extraction
Yarkoni et al., 2001

56. Definition of an “Area”
Neuroimager’s definition of an area: Some blob vaguely in the vicinity (+/- ~3 cm) of where I think it ought to be that lights up for something I think it ought to light up for
Neuroanatomist’s definition of an area: A circumscribed region of the cerebral cortex in which neurons together serve a specific function, receive connections from the same regions, have a common structural arrangement, and in some cases show a topographic arrangement
may also be called a cortical field

57. Cortical Fields: Multiple Criteria
Function
an area has a unique pattern of responses to different stimuli
Architecture
different brain areas show differences between cortical properties (e.g., thickness of different layers, sensitivity to various dyes)
Connectivity
Different areas have different patterns of connections with other areas
Topography
many sensory areas show topography (retinotopy, somatotopy, tonotopy)
boundaries between topographic maps can indicate boundaries between areas (e.g., separate maps of visual space in visual areas V1 and V2)

58. Can We Use Multiple Criteria in Human Imaging?
Function
this is often the only criterion in fMRI
Architecture
there are now probabilistic maps of human brain areas available (Zilles lab)
Connectivity
DTI and functional connectivity now give us options here
Topography
topography is useful in imaging, especially for early and mid-level sensory areas

59. Brodmann Area 17 Meets 21st Century
Anatomical MRI
Functional MRI
Layer 4
Logothetis fMRI data: image from
Goense, Zappe & Logothetis, 2007, MRI
Layer 4 fMRI activation (0.3 x 0.3 x 2 mm spin echo)

60. Brodmann’s Areas Brodmann (1905):
Based on cytoarchitectonics: study of differences in cortical layers between areas
Most common delineation of cortical areas
More recent schemes subdivide Brodmann’s areas into many smaller regions
Monkey and human Brodmann’s areas not necessarily homologous

61. Brodmann Area 17

62. Retinotopic Maps
EXPANDING
RINGS
ROTATING
WEDGES

63. DTI in V1
Saentz & Fine, 2010, NeuroImage

64. Maps, Maps, Maps … even in parietal lobe … … even in frontal lobe …
Hagler & Sereno, 2006, NeuroImage
Wandell et al., 2007, Neuron

65. Other Sensory “-topies”
Touch:
Somatotopy
Servos et al., 1998
red = wrist; orange = shoulder
Audition:
Tonotopy
Sylvian fissure
temporal lobe
cochlea
Movie: tonotopy.mpeg
Source: Marty Sereno’s web page

66. Learning Brain Anatomy
Duvernoy, 1999, The Human Brain: Surface, Blood Supply, and Three-Dimensional Sectional Anatomy
beautiful pictures
good schematic diagrams
clear anatomy
slices of real brain
Springer, US$439
DISCONTINUED
Ono, 1990, Atlas of the Cerebral Sulci
great for showing intersubject variability
gives probabilities of configurations and stats on sulci
Theime, US$199
Damasio,2005, Human Brain Anatomy in Computerized Images, 2nd edition
good for showing sulci across wide range of slice planes
2nd edition much better than 1st edition
Oxford University Press, US$100
Tamraz & Comair, 2000, Atlas of Regional Anatomy of the Brain Using MRI with Functional Correlations
good overview
Springer, US$203
Talairach & Tournoux, Co-Planar Stereotaxic Atlas of the Human Brain
just because it’s the standard doesn’t mean it’s good
Theime, US$240

67. Brain Tutor
Mac/PC: free
iOS App: $1.99

68. Proposal Guidelines Jody Culham Brain and Mind Institute
Department of Psychology
Western University
Proposal Guidelines
Last Update: November 21, 2016
Last Course: Psychology 9223, F2016, Western University

69. Research Proposal Due December 12, 2016
Goals
give students an opportunity to demonstrate what they’ve learned and apply ideas to their research area
give students practice in writing grants/papers
16 double-spaced pages + figures
must be original
not thesis
not something your advisor totally worked out for you
can get some suggestions from advisor but core of proposal should be your work

70. Research Proposal partially like a grant partially like a paper
proposal for experiment
make case for why experiment should be done
“hasn’t been done before” is not good enough
clear question, hypotheses
conclusion: so what?
immunization against potential criticism
partially like a paper
just one experiment, not 5 years of experiments
in-depth methods
be clear about design (e.g., protocol) and analyses
be clear about contrasts
Appendix with budget and time line
don’t worry about formatting – spend your time on content not formatting

71. Range of Approaches
Standard univariate fMRI with hypothesis-driven GLM
Block or Event-related
Advanced designs
e.g., MVPA
Data-driven fMRI
e.g., ICA on resting state data
Approaches we’ve touched on in class
e.g., intersubject correlations
Anatomical approaches
DTI
For the more computationally inclined
better ways to analyze data
if you must use equations, explain them intuitively in text (and consider putting them in an appendix)
Be aware that we won’t discuss these in too much detail in class; therefore you would need to have some prior exposure or to do some extra reading

72. Two questions to consider whenever you write a paper or give a talk
Who is my audience?
a math-phobic professor who will be checking whether you understood the core ideas of fMRI
What is my goal?
show professor that you can find a way to use neuroimaging in your research
show professor that you understand jargon and concepts
Bonus
be clever and creative
write clearly and concisely
solidify your understanding of neuroimaging approaches
think more deeply about how to apply neuroimaging

73. Sections Introduction
Give enough information to put the research in context and lead the reader to the conclusion that the experiment you’re proposing is a reasonable next step
You don’t need to cite every paper in the history of neuroimaging
Do enough of a lit search to be fairly certain proposal hasn’t been done
Replication attempts discouraged (but may be considered with sufficient justification)
Methods
Include enough detail to demonstrate that you understand jargon and key concepts
Be clear about specific contrasts
Results
How could it turn out?
May want to include graph of hypotheses
Conclusions
Don’t just summarize; conclude! So what?!
What would it mean if the results turned out one way or another? Are there any caveats that should be acknowledged? What is the broader significance of the research?
References
whatever format you like (e.g., APA)
don’t go overboard
Figures (optional)
Appendix
How much will it cost?
How long will it take?

74. Key Methods # participants (pilot + discarded + final)
Design (block, event-related)
Timing of TR, trials, blocks, jittering, runs, session
Stimulus/task conditions + rationale
Preprocessing steps
include criterion for motion correction
Approach (localizer?; ROI and/or voxelwise; individual and/or group; for event-related convolution or deconvolution)
Contrasts
don’t just say “activation for faces” or “activation for faces vs houses, bodies & scrambled”
define specific contrast: e.g., “a conjunction of faces vs. each control category [(F>H)AND(F>B)AND(F>S)]
corrections (esp. multiple comparisons)

75. Example of Hypothesis Figure