Group Analyses in fMRI

1. Jody Culham Brain and Mind Institute Department of Psychology Western University Group Analyses in fMRI http://www.fmri4newbies.com/ Last Update: March 3, 2012 Last Course: Psychology 9223, W2013, 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. Corpus Callosum Fornix Pineal Body “bent asparagus” Rotate brain into ACPC plane Find anterior commisure (AC) Find posterior commisure (PC) ACPC line = horizontal axis 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. Neurologic (i.e. sensible) convention • left is left, right is right L R - + x = 0 • Radiologic (i.e. stupid) convention • left is right, right is left R L + - x = 0 Left is what?!!! Note: Make sure you know what your magnet and software are doing before publishing left/right info! Note: If you’re really unsure which side is which, tape a vitamin E capsule to the one side of the subject’s head. It will show up on the anatomical image. (Remember which side you put it on!)

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

10. 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

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

12. 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 • not appropriate for all brains (e.g., group variability, patients may not fit well) • activation foci can vary considerably – other landmarks like sulci may be more reliable

13. MNI Space • Researchers at the Montreal Neurological Institute created a better template based on a morphed average of hundreds of brains (not just one brain like Talairach) • The MNI brain is more representative of average brain shape; however, it does not provide Brodmann areas • The MNI alignment is more complex than Talairach: SPM uses it but many software packages still use Talairach • CAVEAT: The MNI and Talairach coordinate are similar but not identical -- careful comparison requires a transformation -- converters can be found online http://brainmap.org/icbm2tal/index.html Source: http://www.mrc-cbu.cam.ac.uk/personal/matthew.brett/abstracts/MNITal/mniposter.pdf

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

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

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

17. 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

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

19. Cortical Inflation Movie Movie: unfoldorig.mpeg http://cogsci.ucsd.edu/~sereno/unfoldorig.mpg Source: Marty Sereno’s web page

20. Cortical Flattening 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

21. 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 Movie: brain2ellipse.mpeg http://cogsci.ucsd.edu/~sereno/coord1.mpg Source: Marty Sereno’s web page Source: Fischl et al., 1999

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

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

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

25. Voxelwise Group Analyses Fixed vs. Random Effects

26. Subjects as Fixed Effects (FFX) • all levels of an effect (e.g., all subjects) were measured • how likely is it that the effect is true in the subjects I tested? • concatenate all the data from all the subjects • does not estimate inter-subject variability • results can arise from a few Ss with strong effects • you can only draw conclusions about the group of subjects you tested Mumford & Poldrack, 2007, SCAN

27. Sample ROI-GLM • 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

28. Subjects as Random Effects (RFX) • random effects (RFX)* • only a subset of levels of an effect (e.g., a subset of subjects) were sampled • how likely is it that the effect is true in the population from which I drew my sample? • variability across subjects ismodelled • *Point of confusion: • in fMRI, some people talk about a model in which subjects are treated as RFX as an RFX model; others call it a mixed effects model (because subjects are RFX but other aspects can be FFX) underpaid graduate students in need of a few bucks!

29. Two Levels of Analysis Second-level analysis First-level analysis Mumford & Poldrack, 2007, SCAN Another way to think about it… The first level is the same as what we have done with single-subjects’ data except that the data has been put into a common space first (e.g., Talairach, MNI or cortical alignment space). At both the first and second level, we’re making one GLM per voxel. Huettel, Song & McCarthy, 2008 First-level analysis First-level analysis

30. Sample ROI-GLM 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

31. Smoothing and Averaging anatomical variability of activation for 3 Ss without spatial smoothing each subject shows an effect but there’s not enough spatial overlap to find any voxels in an RFX analysis anatomical variability of activation for 3 Ss with spatial smoothing now there’s enough overlap between Ss that some voxels will be found with RFX analysis

32. 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

33. 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)

34. 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 basis FFX but at least enables you to examine the consistency between Ss …

35. Sample ROI-GLM S1 S2 We could just do this in a region we have defined independently. Then we don’t have to worry about the multiple comparisons problem 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

36. Region of interest (ROI) approach • Perform statistical contrasts for the ROI data in an INDEPENDENT data set • Because the runs that are used to generate the area are independent from those used to test the hypothesis, liberal statistical thresholds (e.g., p < .05) can be used Identify a region of interest images from O’Reilly et al., 2012, SCAN Functional ROI Anatomical ROI Functional-Anatomical ROI

37. Voxelwise Approach vs. ROI Approach To Localize or Not to Localise?

38. Neuroimagers can’t even agree how to SPELL localiser/localizer! To Localize or Not to Localise?

39. Methodological Fundamentalism The latest review I received…

40. Approach #1: Voxelwise Statistics Run a statistical contrast for every voxel in your search volume. Correct for multiple comparisons. Find a bunch of blobs.

41. Voxelwise Approach: Example • Malach et al., 1995, PNAS • Question: Are there areas of the human brain that are more responsive to objects than scrambled objects • You will recognize this as what we now call an LO localizer, but Malach was the first to identify LO LO (red) responds more to objects, abstract sculptures and faces than to textures, unlike visual cortex (blue) which responds well to all stimuli LO activation is shown in red, behind MT+ activation in green

42. Example of ROI Approach Culham et al., 2003, Experimental Brain Research Does the Lateral Occipital Complex compute object shape for grasping? Step 1: Localize LOC Intact Objects Scrambled Objects

43. Example of ROI Approach Culham et al., 2003, Experimental Brain Research Does the Lateral Occipital Complex compute object shape for grasping? Step 2: Extract LOC data from experimental runs Grasping Reaching NS p = .35 NS p = .31

44. Example of ROI Approach Very Simple Stats • Instead of using % BOLD Signal Change, you can use beta weights • You can also do a planned contrast in Brain Voyager using a module called the ROI GLM Then simply do a paired t-test to see whether the peaks are significantly different between conditions Extract average peak from each subject for each condition NS p = .35 NS p = .31

45. Example: The Danger of ROI Approaches • Example 1: LOC may be a heterogeneous area with subdivisions; ROI analyses gloss over this • Example 2: Some experiments miss important areas (e.g., Kanwisher et al., 1997 identified one important face processing area -- the fusiform face area, FFA -- but did not report a second area that is a very important part of the face processing network -- the occipital face area, OFA -- because it was less robust and consistent than the FFA.

46. Pros and Cons: Voxelwise Approach • Benefits • Require no prior hypotheses about areas involved • Include entire brain • May identify subregions of known areas that are implicated in a function • Doesn’t require independent data set • Drawbacks • Requires conservative corrections for multiple comparisons • vulnerable to Type II errors • Neglects individual differences in brain regions • poor for some types of studies (e.g., topographic areas) • Can lose spatial resolution with intersubject averaging and smoothing • adjacent areas become blurred together • Requires speculation about areas involved

47. Pros and Cons: ROI Approach • Benefits • Extraction of ROI data can be subjected to simple stats • Elimination of mega multiple comparisons problem greatly improves statistical power (e.g., p < .05) • Hypothesis-driven • Useful when hypotheses are motivated by other techniques (e.g., electrophysiology) in specific brain regions • ROI is not smeared due to intersubject averaging • Important for discriminating abutting areas (e.g., V1/V2) • Easy to analyze and interpret • Can be useful for dissecting factorial design data in an unbiased manner • Drawbacks • Neglects other areas that may play a fundamental role • If multiple ROIs need to be considered, you can spend a lot of scan time collecting localizer data (thus limiting the time available for experimental runs) • Works best for reliable and robust areas with unambiguous definitions • Sometimes you can’t find an ROI in some subjects • Selection of ROIs can be highly subjective and error-prone

48. A Proposed Resolution • You can do BOTH ROI analyses and voxelwise analyses • ROI analyses for well-defined key regions • Voxelwise analyses to see if other regions are also involved • Ideally, the conclusions will not differ • If the conclusions do differ, there may be sensible reasons • Effect in ROI but not voxelwise • perhaps region is highly variable in stereotaxic location between subjects • perhaps voxelwise approach is not powerful enough • Effect in voxelwise but not ROI • perhaps ROI is not homogenous or is context-specific

49. 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

50. Talairach Daemon • http://www.talairach.org