1. 7/10/2014 Yingying

2. MRI studies brain anatomy. Functional MRI (fMRI) studies brain function. MRI vs. fMRI 2

3. Slice Thickness e.g., 6 mm Number of Slices e.g., 10 SAGITTAL SLICE IN-PLANE SLICE Field of View (FOV) e.g., 19.2 cm VOXEL (Volumetric Pixel) 3 mm 6 mm Slice Terminology Matrix Size e.g., 64 x 64 In-plane resolution e.g., 192 mm / 64 = 3 mm 3

4. 4 Coordinates - Anatomy 3 Common Views of Brain: Coronal (head on) Axial (bird’s eye), aka Transverse. Sagittal (profile) sagittalcoronal axial

5. Overview of SPM 5

6. Overview Categorical designs Subtraction - Pure insertion, evoked/differential responses Conjunction - Testing multiple hypotheses Parametric designs Linear - Adaptation, cognitive dimensions Nonlinear - Polynomial expansions, neurometric functions Factorial designs Categorical - Interactions and pure insertion Parametric - Linear and nonlinear interactions - Psychophysiological interactions 6

7. Cognitive subtraction Aim Neural structures underlying a single process Y (e.g. face recognition)? Procedure: Contrast: [Task with Y] – [control task without Y] = Y  the critical assumption of pure insertion Example:[Task with Y] – [task without Y] = Y 7

8. Cognitive subtraction Reality: 8

9. Cognitive subtraction: Baseline problems Which neuronal structures support face recognition? Distant stimuli Several components differ! Related stimuli Y implicit in control condition? Same stimuli, different task Interaction of task and stimuli (i.e. do task differences depend on stimuli chosen)? Name Person! Name Gender! President? Aunt Yingying 9

10. Evoked responses 10

11. Differential responses 11

12. A categorical analysis 12

13. Categorical design 13

14. Overview Categorical designs Subtraction - Pure insertion, evoked/differential responses Conjunction - Testing multiple hypotheses Parametric designs Linear - Adaptation, cognitive dimensions Nonlinear - Polynomial expansions, neurometric functions Factorial designs Categorical - Interactions and pure insertion Parametric - Linear and nonlinear interactions - Psychophysiological interactions 14

15. Conjunctions One way to minimize the baseline/pure insertion problem is to isolate the same process by two or more separate comparisons, and inspect the resulting simple effects for commonalities A test for such activation common to several independent contrasts is called “conjunction” Conjunctions can be conducted across a whole variety of different contexts: tasks stimuli senses (vision, audition) etc. Note: the contrasts entering a conjunction must be orthogonal ! 15

16. Conjunctions 16

17. 17

18. Test of global null hypothesis: Significant set of consistent effectsTest of global null hypothesis: Significant set of consistent effects  “Which voxels show effects of similar direction (but not necessarily individual significance) across contrasts?”  Null hypothesis: No contrast is significant: k = 0  does not correspond to a logical AND ! Test of conjunction null hypothesis: Set of consistently significant effectsTest of conjunction null hypothesis: Set of consistently significant effects  “Which voxels show, for each specified contrast, significant effects?”  Null hypothesis: Not all contrasts are significant: k < n  corresponds to a logical AND A1-A2 B1-B2 p(A1-A2) <  + p(B1-B2) <  + Friston et al.Neuroimage, 25:661-667. Friston et al. (2005). Neuroimage, 25:661-667. Nichols et al. (2005). Neuroimage, 25:653-660. Two types of conjunctions 18

19. SPM offers both conjunctions Friston et al.Neuroimage, 25:661-667. Friston et al. 2005, Neuroimage, 25:661-667. specificity sensitivity Global null: k = 0 Conjunction null: k < n 19

20. F-test vs. conjunction based on global null Friston et al.Neuroimage, 25:661-667. Friston et al. 2005, Neuroimage, 25:661-667. 20

21. Using the conjunction null is easy to interpret, but can be very conservative Friston et al.Neuroimage, 25:661-667. Friston et al. 2005, Neuroimage, 25:661-667. 21

22. Overview Categorical designs Subtraction - Pure insertion, evoked/differential responses Conjunction - Testing multiple hypotheses Parametric designs Linear - Adaptation, cognitive dimensions Nonlinear - Polynomial expansions, neurometric functions Factorial designs Categorical - Interactions and pure insertion Parametric - Linear and nonlinear interactions - Psychophysiological interactions 22

23. Parametric designs Parametric designs approach the baseline problem by: –Varying the stimulus-parameter of interest on a continuum, in multiple (n>2) steps... –... and relating measured BOLD signal to this parameter Possible tests for such relations are manifold: Linear Nonlinear: Quadratic/cubic/etc. (polynomial expansion) Model-based (e.g. predictions from learning models) 23

24. Parametric modulation of regressors → inverted ‘U’ response to increasing word presentation rate in the DLPFC SPM{F} Polynomial expansion: f(x) ~ f(x) ~ b 1 x + b 2 x 2 + b 3 x 3... Linear Quadratic F-contrast [0 1 0] on quadratic parameter SPM5 offers polynomial expansion as option for parametric modulation of regressors Büchel et al. 1996, NeuroImage 4:60-66 24

25. Parametric modulation of regressors by time Büchel et al. 1998, NeuroImage 8:140-148 25

26. “User-specified” parametric modulation of regressors Polynomial expansion & orthogonalisation Büchel et al. 1998, NeuroImage 8:140-148 26

27. Investigating neurometric functions (= relation between a stimulus property and the neuronal response) Stimulus awareness Stimulus intensity Pain intensity Pain threshold: 410 mJ P1 P2 P3 P4 P0-P4: Variation of intensity of a laser stimulus applied to the right hand (0, 300, 400, 500, and 600 mJ) Büchel et al. 2002, J. Neurosci. 22:970-976 27

28. Neurometric functions  Stimulus presence  Pain intensity  Stimulus intensity Büchel et al. 2002, J. Neurosci. 22:970-976 28

29. Model-based regressors general idea: generate predictions from a computational model, e.g. of learning or decision-making Commonly used models: Rescorla-Wagner learning model temporal difference (TD) learning model Bayesian models use these predictions to define regressors include these regressors in a GLM and test for significant correlations with voxel-wise BOLD responses 29

30. Model-based fMRI analysis Gläscher & O‘Doherty 2010, WIREs Cogn. Sci. 30 A A B B t outco me

31. Gläscher & O‘Doherty 2010, WIREs Cogn. Sci. 31 0 1

32. Fixation cross Auditory Visual Fixation cross Time (ms) 020 0 40 0 60 0 80 0 100 0 120 0 200 0 ± 500 or Visual “Distractor ” Target “Distracto r” Target 140 0 Incidental learning of audio-visual associations Hypothesis: Incidental learning of this relation is reflected by prediction-error dependent changes in connectivity between auditory and visual areas. 80%  den Ouden et al. 2009, Cereb. Cortex 32

33. Rescorla-Wagner model of associative learning p < 0.05, corrected random effects, n=16 -V1 & PUT increasingly activate the more surprising the visual outcome is -V1 & PUT increasingly deactivate the more expected the visual outcome is Rescorla-Wagner learning curve (  =0.075): During learning, predictive tones V1 PUT den Ouden et al. 2009, Cereb. Cortex 33

34. Overview Categorical designs Subtraction - Pure insertion, evoked/differential responses Conjunction - Testing multiple hypotheses Parametric designs Linear - Adaptation, cognitive dimensions Nonlinear - Polynomial expansions, neurometric functions Factorial designs Categorical - Interactions and pure insertion Parametric - Linear and nonlinear interactions - Psychophysiological interactions 34

35. Main effects and interactions A1A2 B2B1 Task (1/2) Viewing Naming Stimuli (A/B) Objects Colours Colours Objects Colours Objects Colours Objects Colours Objects interaction effect (Stimuli x Task) (Stimuli x Task) ViewingNaming Main effect of task:(A1 + B1) – (A2 + B2)Main effect of task:(A1 + B1) – (A2 + B2) Main effect of stimuli: (A1 + A2) – (B1 + B2)Main effect of stimuli: (A1 + A2) – (B1 + B2) Interaction of task and stimuli: Can show a failure of pure insertionInteraction of task and stimuli: Can show a failure of pure insertion (A1 – B1) – (A2 – B2) 35

36. Factorial design 36 A1A2 B2B1 Task (1/2) Viewing Naming Stimuli (A/B) Objects Colours A1 B1 A2 B2 Main effect of task: (A1 + B1) – (A2 + B2)

37. 37 A1A2 B2B1 Task (1/2) Viewing Naming Stimuli (A/B) Objects Colours A1 B1 A2 B2 Main effect of stimuli: (A1 + A2) – (B1 + B2) Factorial design

38. 38 A1A2 B2B1 Task (1/2) Viewing Naming Stimuli (A/B) Objects Colours A1 B1 A2 B2 Interaction of task and stimuli: (A1 – B1) – (A2 – B2) Interaction of task and stimuli: (A1 – B1) – (A2 – B2) Factorial design

39. Example: evidence for inequality-aversion Tricomi et al. 2010, Nature 39

40. Parametric interactions p < 0.05, corrected random effects, n=16 V1 & PUT increasingly activate the more surprising the visual outcome is V1 & PUT increasingly deactivate the more expected the visual outcome is Significant four-way interaction in V1 and putamen: V1 PUT A+A- V+V+ V-V- V+V-V- V-V- V+V+ V-V- A+A- Primary visual cortex Rescorla-Wagner learning curves (  =0.075): den Ouden et al. 2009, Cereb. Cortex 40

41. Psycho-physiological interactions (PPI) We can replace one main effect in the GLM by the time series of an area that shows this main effect. E.g. let's replace the main effect of stimulus type by the time series of area V1: Task factor Task A Task B Stim 1 Stim 2 Stimulus factor T A /S 1 T B /S 1 T A /S 2 T B /S 2 GLM of a 2x2 factorial design: main effect of task main effect of stim. type interaction main effect of task V1 time series  main effect of stim. type psycho- physiological interaction 41

42. PPI example: attentional modulation of V1→V5 attention no attention V1 activity V5 activity SPM{Z} time V5 activity Friston et al. 1997, NeuroImage 6:218-229 Büchel & Friston 1997, Cereb. Cortex 7:768-778 V1 V1 x Att. = V5 V5 Attention 42

43. PPI: interpretation Two possible interpretatio ns of the PPI term: V1V1 Modulation of V1  V5 by attention Modulation of the impact of attention on V5 by V1. V1V5 V1V1 V5 attention V1 attention 43

44. FMRI Design Types 44 1)Blocked Designs 2)Event-Related Designs a)Periodic Single Trial b)Jittered Single Trial 3)Mixed Designs - Combination blocked/event-related

45. Block Design 45

46. Block Design Sequences Consider the simplest case, a block design with two conditions e.g., alternate tapping of two fingers vs. rest let’s assume 2 sec/volume How long should a run be? Short enough that the subject can remain comfortable without moving or swallowing. Long enough that you’re not wasting a lot of time restarting the scanner. Ideal is ~5 ± 2 minutes images time course of activation haemodynamic response function finger tapping baseline rest Source: Jody Culham’s web slidesweb slides 46

47. Adapted from Gusnard & Raichle (2001) (E - Bad Control Design) 47

48. Block Design Sequences How fast should the conditions cycle? Every 4 sec (2 images) signal amplitude is weakened by HRF not too far from range of breathing frequency (every 4-10 sec)  could lead to respiratory artifacts if design is a task manipulation, subject is constantly changing tasks, gets confused pre-HRF post-HRF Every 96 sec (48 images) more noise at low frequencies linear trend confound subject will get bored very few repetitions – hard to do eyeball test of significance post-HRF Source: Jody Culham’s web slidesweb slides 48

49. Block Design Sequences Every 16 sec (8 images) allows enough time for signal to oscillate fully not near artifact frequencies enough repetitions to see cycles by eye a reasonable time for subjects to keep doing the same thing post-HRF Other factors: symmetric design – some use longer rest vs. activation periods add a few extra images at the end to allow the hemodynamic response to catch up add extra time at the beginning to allow for the magnet to warm up and the subject to warm up (let the startle response die down) Source: Jody Culham’s web slidesweb slides 49

50. But I have 4 conditions to compare! Adding conditions makes things way more complicated. There’s no right answer, but like everything else in fMRI, there are various tradeoffs. Let’s consider the case of four conditions plus a baseline. 1. Main condition epochs all separated by baseline epochs Pro: simplest case if you want to use event-related averaging to view time courses and don’t want to have to worry about baseline issues Con: spends a lot of your “n” on baseline measures A. Orderly progression Pro: Simple Con: May be some confounds (e.g., linear trend if you predict green&blue > pink&yellow) B. Symmetric order Pro: No linear trends to confound Con: Each condition occurs at a different frequency, including some (e.g., yellow in top sequence) that occur at low frequencies which are noisier C. Random order in each run Pro: order effects should average out Con: pain to make various protocols, no possibility to average all data into one time course, many frequencies involved Source: Jody Culham’s web slidesweb slides 50

51. 2. Clustered conditions with infrequent baselines. Kanwisher 4 condition design A. Kanwisher lab design sets of four main condition epochs separated by baseline epochs each main condition appears at each location in sequence of four two counterbalanced orders (1 st half of first order same as 2 nd half of second order and vice versa) – can even rearrange data from 2 nd order to allow averaging with 1 st order Pro: spends most of your n on key conditions, provides more repetitions Con: not great for event-related averaging because orders are not balanced (e.g., in top order, blue is preceded by the baseline 1X, by green 2X, by yellow 1X and by pink 0X). As you can imagine, the more conditions you try to shove in a run, the thornier ordering issues are and the fewer n you have for each condition. Jody’s rule of thumb: Never push it beyond 4 main + 1 baseline. Source: Jody Culham’s web slidesweb slides 51

52. Power in Blocked Designs 1. Summation of responses results in large signals then plateaus (~10 sec) 1. Response Duration does not plateau and onset does not change Stimulus duration and interval compared with HRF ISI = 1 sec

53. Choosing Length of Blocks Longer block lengths allow for stability of extended responses Hemodynamic response saturates following extended stimulation After about 10s, activation reaches plateau Many tasks require extended intervals Brain processing may differ throughout the task period Shorter block lengths move your signal to higher temporal frequencies Away from low-frequency noise: scanner drift, etc. Not possible in O-15 PET rCBF studies Periodic blocks may result in aliasing of other periodic signals in the data Example: if the person breathes at a regular rate of 12 per min and the blocks are 10s long (6 blocks/min) Could be problem if the aliased signal falls within the range of desired signals From Scott Huettel, Duke

54. Limitations of Blocked Designs Sensitive to signal drift or MR instability Poor choice of conditions/baseline may preclude meaningful conclusions Many tasks cannot be conducted well repeatedly

55. Non-Task Brain Processing In experiments activation can be greater in baseline conditions than in task conditions! Requires different processing for interpretation Suggests the idea of baseline/resting mental processes Gathering/evaluation about the world around you Awareness (of self) Online monitoring of sensory information Daydreaming Neurons that are wired together fire together This collection of resting state brain processes is often called the “Default Mode Network” (DMN)

56. Blocked Design Pros.Cons. Excellent detection power (knowing which voxels are active) Useful for examining state changes Poor estimation power (knowing the time course of an active voxel) Relatively insensitive to the shape of the hemodynamic response. Stim ON 10s Stim OFF 16s 15Hz Stim ON 10s Stim OFF 16s 10 repetitions

57. Event-related Design 57

58. Buckner et al., 1998 Event Related

59. What are Event-Related Designs? Event-related designs associate brain processes with discrete events, which may occur at any point in the scanning session. Can detect transient BOLD responses Supports adapting task to response such as changing difficulty based on error rate

60. Why use event-related designs? Some experimental tasks are naturally event-related (future stimuli based on response) Allows studying within-trial effects Improves relation to behavioral factors (behavior changes within blocks may be masked) Simple analyses Selective averaging General linear models (GLM)

61. Same Event Averaging Sorting Into Common Groups - Behavior - Physiological Measure - Outlier Rejection - Transient vs. Task level Responses

62. Periodic Single Trial Designs Stimulus events presented infrequently with long inter-stimulus intervals (ISIs) 500 ms 18 s

63. Trial Spacing Effects: Periodic Designs ISI = 8sec (~12 trials)ISI = 4sec (~45 trials) ISI = 20sec (9 trials)ISI = 12sec (15 trials) A 20 A4A4 A8A8 A 12 Want to maximize amplitude times number of trials per study

64. Bandettini & Cox, 2000 The optimal inter-stimulus interval (ISI) for a stimulus duration (SD), was determined. Empirical Observation: For SD=2sec, ISI=12 to 14 sec. Theory Predicts: For SD<=2 sec, the optimal repetition interval (RI=ISI+SD) Theory Predicts: For SD>2sec, RI = 8+(2*SD). The statistical power of ER-fMRI relative to blocked-design was determined Empirical: For SD=2 sec, ER-fMRI was ~35% lower than that of blocked-design Simulations that assumed a linear system demonstrated estimate ~65% reduction in power Difference suggest that the ER-fMRI amplitude is greater than that predicted by a linear shift-invariant system models.

65. Jittered Single Trial Designs Varying the timing of trials within a run Varying the timing of events within a trial Trial 1Trial 2Trial 3Trial 4 2 events3 events2 events1 event

66. Effects of Jittering on Response Stimulus Response Jittering allows us to sample BOLD response in more states

67. Effects of ISI on Detectability Birn et al, 2002 Jittered ISI Constant ISI Detectability Estimated Accuracy of HRF Max when ½ stims are task state and ½ stims are control state

68. Dale and Buckner (1997) Detecting Using Selective Averaging Low Response Fewer Samples Mid Response More Samples Large Response Most samples Visual stim duration = 1 s; acquisition 240 sec Trials subtracted then correlation analysis with predicted response

69. Variability of HRF: Evidence Aguirre, Zarahn & D’Esposito, 1998 HRF shows considerable variability between subjects Within subjects, responses are more consistent, although there is still some variability between sessions different subjects same subject, same sessionsame subject, different session

70. Variability of HRF: Implications Aguirre, Zarahn & D’Esposito, 1998 Generic HRF models (gamma functions) account for 70% of variance Subject-specific models account for 92% of the variance (22% more!) Poor modeling reduces statistical power Less of a problem for block designs than event-related (do you know why?) Biggest problem with delay tasks where an inappropriate estimate of the initial and final components contaminates the delay component Possible solution: model the HRF individually for each subject Possible caveat: HRF may also vary between areas, not just subjects Buckner et al., 1996: noted a delay of 0.5-1 sec between visual and prefrontal regions vasculature difference? processing latency? Bug or feature? Menon & Kim – mental chronometry

71. Block/epoch design v.s. event- related designs 71 Randomised trials order

72. efMRI: Post-hoc classification of trials 72 Post-hoc subjective classification of trials

73. Post-Hoc Sorting of Trials From Kim and Cabeza, 2007 Using information about fMRI activation at memory encoding to predict behavioral performance at memory retrieval. True Memory Formation vs. False Memory Formation

74. “Event” model of block design 74

75. Modeling block designs: Epochs vs Events 75

76. Modeling block designs: Epochs vs Events 76

77. Limitations of Event-Related Designs Low power (maybe) Collecting lots of data, many runs The key issues are: Can my subjects perform the task as designed? Are the processes of interest independent from each other (in time, amplitude, etc.)?

78. Event-Related Design Pros.Cons. Good at estimating shape of hemodynamic response Provides good estimation power (knowing the time course of an active voxel) Can have reduced detection power (knowing which voxels are active) Sensitive to errors in predicted hemodynamic response Event 1Event 2Event 3Event 4

79. Mixed Design 79

80. Mixed: Combination Blocked/Event Both blocked and event-related design aspects are used (for different purposes) Blocked design: state-dependent effects Event-related design: item-related effects Analyses can model these as separate phenomena, if cognitive processes are independent. “Memory load effects” vs. “Item retrieval effects” Or, interactions can be modeled. Effects of memory load on item retrieval activation.

81. Blocked (solid) Event-Related (dashed) Event-related model reaches peak sooner… … and returns to baseline more slowly. In this study, some language-related regions were better modeled by event-related. From Mechelli, et al., 2003 You can model a block with events…

82. Mixed Design

83. Summary of Experiment Design Main Issues to Consider What design constraints are induced by my task? What am I trying to measure? What sorts of non-task-related variability do I want to avoid? Rules of thumb Blocked Designs: Powerful for detecting activation Useful for examining state changes Event-Related Designs: Powerful for estimating time course of activity Allows determination of baseline activity Best for post hoc trial sorting Mixed Designs Best combination of detection and estimation Much more complicated analyses

84. What is fMRI Experimental Design? Controlling the timing and quality of cognitive operations to influence brain activation What can we control? Stimulus properties (what is presented?) Stimulus timing (when is it presented?) Subject instructions (what do subjects do with it?) What are the goals of experimental design? To test specific hypotheses (i.e., hypothesis-driven) To generate new hypotheses (i.e., data-driven)

85. Appendix (more information) 85

86. BOLD impulse response 86

87. BOLD impulse response 87

88. Design efficiency 88

89. Sinusoidal modulation, f=1/33 89

90. Blocked, epoch=20 second 90

91. Randomized, SOAmin=4s, highpass filter = 1/120s 91

92. Checking your design efficiency 92

93. Thank you Questions? Thank for the internet resources. I was able to borrow a lot of slides from different presentations. 93