1. 1 Experimental Design An Introduction to MRI Physics and Analysis Michael Jay Schillaci, PhD Monday, March 17 th, 2008

2. 2 Overview  Design terminology  Blocked Designs  Event-Related Designs  Mixed Designs  Summary

3. 3 Design Terminology

4. 4 Experimental Design: Terminology Variables  Independent vs. Dependent  Categorical vs. Continuous Contrasts  Experimental vs. Control  Parametric vs. subtractive Comparisons of subjects  Between- vs. Within-subjects Confounding factors Randomization, counterbalancing

5. 5 What is fMRI Experimental Design? Controlling the timing and quality of cognitive operations (IVs) to influence brain activation (DVs) 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)

6. 6 What types of hypotheses are possible for fMRI data?

7. 7 Optimal Experimental Design Maximizing both Detection and Estimation  Maximal variance in signal (incr. detect.)  Maximal variance in stimulus timing (incr. est.) Limitations on Optimal Design  Refractory effects  Signal saturation  Subject’s predictability

8. 8 Finding effects Statistics are based on the ratio of explained predictable versus unexplained variability: We can improve statistical efficiency by  Increasing the task related variance (signal) Designing Experiments (today’s lecture)  Decreasing unrelated variance (noise) Spatial and temporal processing lectures.  Good signal in our fMRI data Physics lectures Signal+Noise Noise F= Signal Noise t=

9. 9 fMRI Signal There are two crucial apects of the BOLD effect:  The HRF is very sluggish The is a long delay between brain activity and changes in fMRI images (~5s).  The HRF is additive Doing a task twice causes about twice as much change as doing it once.

10. 10 The BOLD timecourse Visual cortex shows peak response ~5s after visual stimuli. Indirect measure 0 6 12 18 24 % Signal Change 2 102 10 Time (seconds)

11. 11 Temporal Properties of fMRI Signal Hemodynamic response function (HRF) is sluggish: peak signal above 5s after activation. We predict the HRF by convolving the neural signal by the HRF. We want to maximize the amount of predictable variability. Convolved Response = Neural SignalHRF

12. 12 BOLD effects are additive Three stimuli presented rapidly result in almost 3 times the signal of a single stimuli (e.g. Dale & Buckner, 1997). Crucial finding for experimental design. Note there are limits to this additivity effect, but the basic point is that more stimuli generate more signal (see Birn et al. 2001)

13. 13 Blocked Designs

14. 14 What are Blocked Designs? Blocked designs segregate different cognitive processes into distinct time periods Task ATask BTask ATask BTask ATask BTask ATask B Task ATask BREST Task ATask BREST

15. 15 Comparing predictable HRF Consider 3 paradigms: 1. Fixed ISI: one stimuli every 16 seconds. – inefficient 2. Fixed ISI: one stimuli every 4 seconds. – Insanely inefficient: virtually no task-related variability 3. Block design: cluster five stimuli in 8 seconds, pause 12 seconds, repeat. – Very efficient. – Cluster of events is additive. Note peak amplitude is x3 the 16s design.

16. 16 Choosing Length of Blocks Longer block lengths allow for stability of extended responses  Hemodynamic response saturates following extended stimulation After about 10s, activation reaches max  Many tasks require extended intervals Processing may differ throughout the task period Shorter block lengths move your signal to higher frequencies  Away from low-frequency noise: scanner drift, etc. Periodic blocks may result in aliasing of other variance in the data  Example: if the person breathes at a regular rate of 1 breath/5sec, and the blocks occur every 10s

17. 17

18. 18

19. 19 Optimal Design Block designs are optimal.  Present trials as rapidly as possible for ~12 sec  Summation maximizes additive effect of HRF.  Consider experiment: Three conditions, each condition repeated 14 times (once every 900ms) 1. Press left index finger when you see  2. Press right index finger when you see  3. Do nothing when you see  Note huge predictable variability in signal.

20. 20 Block designs While efficient, block designs are often predictable. May not be experimentally valid. Optimal block length around 12s, followed by around 12s until condition is repeated.  Avoid long blocks: Reduced signal variability Low frequency signal will be hard to distinguish from low frequency signals such as drift in MRI signal.

21. 21 Block Designs aka ‘Box Car’, or ‘Epoch’ designs. Different cognitive processes occur in distinct time periods 1. Press left index finger when you see  2. Press right index finger when you see  3. Do nothing when you see 

22. 22 Block designs good for detection, poor for estimating HDR. Block design limitations Detection: which areas are active? Estimation: what is the timecourse of activity?

23. 23 Block design limitations While block designs offer statistical power, they are very predictable.  E.G. our participants will know they will press the same finger 14 times in a row. Many tasks not suitable for block design  E.G. Novelty detection, memory, etc.  Your can not post-hoc sort data from block designs, e.g. Konishi, et al., 2000 examine correct rejection vs hits on episodic memory task.

24. 24 Types of Blocked Design Task A vs. Task B (… vs. Task C…)  Example: Squeezing Right Hand vs. Left Hand  Allows you to distinguish differential activation between conditions  Does not allow identification of activity common to both tasks Can control for uninteresting activity Task A vs. No-task (… vs. Task C…)  Example: Squeezing Right Hand vs. Rest  Shows you activity associated with task  May introduce unwanted results

25. 25 Adapted from Gusnard & Raichle (2001)

26. 26 Adapted from Gusnard & Raichle (2001) Oxygen Extraction Fraction Cerebral Metabolic Rate of O 2 Cerebral Blood Flow Any true baseline?

27. 27 Non-Task Processing In many experiments, activation is greater in baseline conditions than in task conditions!  Requires interpretations of significant activation Suggests the idea of baseline/resting mental processes  Gathering/evaluation about the world around you  Awareness (of self)  Online monitoring of sensory information  Daydreaming This collection of processes is often called the “Default Mode”

28. 28 Default Mode! Damoiseaux 2006 analyzed separate 10-subject resting- state data sets, using Independent Components analysis. Vision. Memory.

29. 29 Power in Blocked Designs 1. Summation of responses results in large variance

30. 30 HDR Estimation: Blocked Designs

31. 31 Deeper concept… We want the changes evoked by the task to be at different parts of the frequency spectrum than non-task-evoked changes.

32. 32 Limitations of Blocked Designs Very sensitive to signal drift Poor choice of conditions/baseline may preclude meaningful conclusions Many tasks cannot be conducted repeatedly Difficult to estimate the Hemodynamic Response

33. 33 Event-Related Designs

34. 34 What are Event-Related Designs? Event-related designs associate brain processes with discrete events, which may occur at any point in the scanning session.

35. 35 Why use event-related designs? Some experimental tasks are naturally event- related Allows studying of trial effects Improves relation to behavioral factors Simple analyses  Selective averaging  General linear models

36. 36 Event related designs Much less power than block designs.  Simply randomizing trial order of our block design, the typical event related design has one quarter the efficiency.  Here, we ran 50 iterations and selected the most efficient event related design. Still half as efficient as the block design. Note this design is not very random: runs of same condition make it efficient.

37. 37 2a. Periodic Single Trial Designs Stimulus events presented infrequently with long interstimulus intervals 500 ms 18 s

38. 38 McCarthy et al., (1997)

39. 39 Trial Spacing Effects: Periodic Designs 20sec 8sec4sec 12sec

40. 40 From Bandettini and Cox, 2000 ISI: Interstimulus Interval SD: Stimulus Duration Why not short, periodic designs?

41. 41 2b. Jittered Single Trial Designs Varying the timing of trials within a run Varying the timing of events within a trial

42. 42 Effects of Jittering on Stimulus Variance

43. 43 Dale and Buckner (1997) How rapidly can we present stimuli?

44. 44 Effects of ISI on Power Birn et al, 2002

45. 45 Mixed Designs

46. 46 3a. 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.

47. 47 Permuted Blocks Permuted block designs (Liu, 2004) offer possible some unpredictability… Permuted Design: 1.Start with a block design 2.Randomly swap stimuli 3.Repeat step to for n iterations More iterations = less predictable, less power

48. 48

49. 49

50. 50 Summary

51. 51 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…

52. 52 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