1. 1. Preprocessing of FMRI Data fMRI Graduate Course October 22, 2003

2. What is preprocessing? Correcting for non-task-related variability in experimental data –Usually done without consideration of experimental design; thus, pre-analysis –Occasionally called post-processing, in reference to being after acquisition Attempts to remove, rather than model, data variability

3. Signal, noise, and the General Linear Model Measured Data Amplitude (solve for) Design Model Noise Cf. Boynton et al., 1996

4. Signal-Noise-Ratio (SNR) Task-Related Variability Non-task-related Variability

5. Preprocessing Steps Slice Timing Correction Motion Correction Coregistration Normalization Spatial Smoothing Segmentation Region of Interest Identification Bias field correction

6. Tools for Preprocessing SPM Brain Voyager VoxBo AFNI Custom BIAC scripts

7. Slice Timing Correction

8. Why do we correct for slice timing? Corrects for differences in acquisition time within a TR –Especially important for long TRs (where expected HDR amplitude may vary significantly) –Accuracy of interpolation also decreases with increasing TR When should it be done? –Before motion correction: interpolates data from (potentially) different voxels Better for interleaved acquisition –After motion correction: changes in slice of voxels results in changes in time within TR Better for sequential acquisition

9. Effects of uncorrected slice timing Base Hemodynamic Response Base HDR + Noise Base HDR + Slice Timing Errors Base HDR + Noise + Slice Timing Errors

10. Base HDR: 2s TR

11. Base HDR + Noise r = 0.77 r = 0.80 r = 0.81

12. Base HDR + Slice Timing Errors r = 0.85 r = 0.92 r = 0.62

13. HDR + Noise + Slice Timing r = 0.65 r = 0.67 r = 0.19

14. Interpolation Strategies Linear interpolation Spline interpolation Sinc interpolation

15. Motion Correction

16. Head Motion: Good, Bad,…

17. … and catastrophically bad

18. Why does head motion introduce problems? ABC

19. Simulated Head Motion

20. Severe Head Motion: Simulation Two 4s movements of 8mm in -Y direction (during task epochs) Motion

21. Severe Head Motion: Real Data Two 4s movements of 8mm in –Y direction (during task epochs) Motion

22. Correcting Head Motion Rigid body transformation –6 parameters: 3 translation, 3 rotation Minimization of some cost function –E.g., sum of squared differences

23. Effects of Head Motion Correction

24. Limitations of Motion Correction Artifact-related limitations –Loss of data at edges of imaging volume –Ghosts in image do not change in same manner as real data Distortions in fMRI images –Distortions may be dependent on position in field, not position in head Intrinsic problems with correction of both slice timing and head motion

25. Prevention is the best medicine AB D C

26. Coregistration

27. Should you Coregister? Advantages –Aids in normalization –Allows display of activation on anatomical images –Allows comparison across modalities –Necessary if no coplanar anatomical images Disadvantages –May severely distort functional data –May reduce correspondence between functional and anatomical images

28. Normalization

30. Standardized Spaces Talairach space (proportional grid system) –From atlas of Talairach and Tournoux (1988) –Based on single subject (60y, Female, Cadaver) –Single hemisphere –Related to Brodmann coordinates Montreal Neurological Institute (MNI) space –Combination of many MRI scans on normal controls All right-handed subjects –Approximated to Talaraich space Slightly larger Taller from AC to top by 5mm; deeper from AC to bottom by 10mm –Used by SPM, National fMRI Database, International Consortium for Brain Mapping

31. Normalization to Template Normalization TemplateNormalized Data

32. Anterior and Posterior Commissures Anterior Commissure Posterior Commissure

33. Should you normalize? Advantages –Allows generalization of results to larger population –Improves comparison with other studies –Provides coordinate space for reporting results –Enables averaging across subjects Disadvantages –Reduces spatial resolution –May reduce activation strength by subject averaging –Time consuming, potentially problematic Doing bad normalization is much worse than not normalizing

34. Slice-Based Normalization Before Adjustment (15 Subjects) After Adjustment to Reference Image Registration courtesy Dr. Martin McKeown (BIAC)

35. Spatial Smoothing

36. Techniques for Smoothing Application of Gaussian kernel –Usually expressed in #mm FWHM –“Full Width – Half Maximum” –Typically ~2 times voxel size

37. Effects of Smoothing on Activity Unsmoothed Data Smoothed Data (kernel width 5 voxels)

39. Should you spatially smooth? Advantages –Increases Signal to Noise Ratio (SNR) Matched Filter Theorem: Maximum increase in SNR by filter with same shape/size as signal –Reduces number of comparisons Allows application of Gaussian Field Theory –May improve comparisons across subjects Signal may be spread widely across cortex, due to intersubject variability Disadvantages –Reduces spatial resolution –Challenging to smooth accurately if size/shape of signal is not known

40. Segmentation Classifies voxels within an image into different anatomical divisions –Gray Matter –White Matter –Cerebro-spinal Fluid (CSF) Image courtesy J. Bizzell & A. Belger

41. Histogram of Voxel Intensities

42. Region of Interest Drawing

43. Why use an ROI-based approach? Allows direct, unbiased measurement of activity in an anatomical region –Assumes functional divisions tend to follow anatomical divisions Improves ability to identify topographic changes –Motor mapping (central sulcus) –Social perception mapping (superior temporal sulcus) Complements voxel-based analyses

44. Drawing ROIs Drawing Tools –BIAC software (e.g., Overlay2) –Analyze –IRIS/SNAP (G. Gerig) Reference Works –Print atlases –Online atlases Analysis Tools –roi_analysis_script.m

45. ROI Examples

46. BIAC is studying biological motion and social perception – here by determining how context modulates brain activity in elicited when a subject watches a character shift gaze toward or away from a target.

47. Additional Resources SPM website –http://www.fil.ion.ucl.ac.uk/spm/course/notes01.htmlhttp://www.fil.ion.ucl.ac.uk/spm/course/notes01.html –SPM Manual Brain viewers –http://www.bic.mni.mcgill.ca/cgi/icbm_view/http://www.bic.mni.mcgill.ca/cgi/icbm_view/

48. 2. Issues in Experimental Design fMRI Graduate Course October 23, 2003

49. What is Experimental Design? Controlling the timing and quality of presented stimuli to influence resulting brain processes What can we control? –Experimental comparisons (what is to be measured?) –Stimulus properties (what is presented?) –Stimulus timing (when is it presented?) –Subject instructions (what do subjects do with it?)

50. Goals of Experimental Design To maximize the ability to test hypotheses To facilitate generation of new hypotheses

51. What are hypotheses? Statements about the relations between independent and dependent variables. ABCD Psychological Hypotheses Hemodynamic Hypotheses Neuronal Hypotheses

52. Independent Variables Aspects of the experimental design that we want to manipulate –Often have multiple levels (e.g., experimental and control conditions) –Critical design choice lies in determining how to choose stimuli to match independent variable ABC

53. Dependent Variable: BOLD signal

54. Causal and non-causal relations between variables AB Is the BOLD response epiphenomenal?

55. Detection vs. Estimation Detection: What is active? Estimation: How does its activity change over time?

56. Detection Detection power defined by SNR Depends greatly on hemodynamic response shape SNR = aM/  M = hemodynamic changes (unit) a = measured amplitude  = noise standard deviation

57. Estimation Ability to determine the shape of fMRI response Accurate estimation relies on minimization of variance in estimate of HDR at each time point Efficiency of estimation is generally independent of HDR form

58. Optimal Experimental Design Maximizing both Detection and Estimation –Maximal variance in stimulus timing (increases estimation) –Maximal variance in measured signal (increases detection) Limitations –Refractory effects –Signal saturation

59. fMRI Design Types 1)Blocked Designs 2)Event-Related Designs a)Periodic Single Trial b)Jittered Single Trial c)Staggered Single Trial 3)Mixed Designs a)Combination blocked/event-related b)Variable stimulus probability

60. 1. Blocked Designs

61. 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

62. PET Designs Measurements done following injection of radioactive bolus Uses total activity throughout task interval (~30s) Blocked designs necessary –Task 1 = Injection 1 –Task 2 = Injection 2

63. 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 allow for more transitions –Task-related variability increases (relative to non-task) with increasing numbers of transitions 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

64. Effects of Block Interval upon HDR 40s20s15s10s 8s6s4s2s

65. What baseline should you choose? Task A vs. Task B –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 –Example: Squeezing Right Hand vs. Rest –Shows you activity associated with task –May introduce unwanted results

66. Interpreting Baseline Activity From Gusnard & Raichle, 2001

67. 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 –Emotional processes –Gathering/evaluation about the world around you –Awareness (of self) –Online monitoring of sensory information –Daydreaming

68. From Shulman et al., 1997 (PET data) From Binder et al., 1999

69. From Huettel et al., 2001 (Change Detection) From Huettel et al., 2002 (Baseline > Target Detection)

70. Power in Blocked Designs 1.Summation of responses results in large variance Single, unit amplitude HDR, convolved by 1, 2, 4,8, 12, or 16 events (1s apart).

71. HDR Estimation: Blocked Designs

72. Power in Blocked Designs 2. Transitions between blocks Simulation of single run with either 2 or 10 blocks.

73. Power in Blocked Designs 2. Transitions between blocks Addition of linear drift within run.

74. Power in Blocked Designs 2. Transitions between blocks Addition of noise (SNR = 0.67)

75. Limitations of Blocked Designs Very sensitive to signal drift –Sensitive to head motion, especially when only a few blocks are used. Poor choice of baseline may preclude meaningful conclusions Many tasks cannot be conducted repeatedly Difficult to estimate the HDR

76. 2. Event-Related Designs

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

78. Why use event-related designs? Some experimental tasks are naturally event-related Allows studying of trial effects Simple analyses –Selective averaging –No assumptions of linearity required

79. Event-Related and Blocked Designs give Similar Results A BC

80. 2a. Periodic Single Trial Designs Stimulus events presented infrequently with long interstimulus intervals 500 ms 18 s

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

82. From Bandettini and Cox, 2000 ISI: Interstimulus Interval SD: Stimulus Duration

83. 2b. Jittered Single Trial Designs Varying the timing of trials within a run

84. Randomization = Jittering Dale & Buckner, 1997

85. Extracting different task components AB

86. Effects of Jittering on Stimulus Variance

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

88. 2c. Staggered Single Trial By presenting stimuli at different timings, relative to a TR, you can achieve sub-TR resolution Significant cost in number of trials presented –Resulting loss in experimental power Very sensitive to scanner drift and other sources of variability

89. Two HDR epochs sampled at a 3s TR. Each row is sampled at a different phase. +0s +1s +2s

90. Two of the phases are normal. But, one has a change in one trial (e.g., head motion) +0s +1s +2s

91. Post-Hoc Sorting of Trials From Konishi, et al., 2000 Data from old/new episodic memory test.

92. Limitations of Event-Related Designs Differential effects of interstimulus interval –Long intervals do not optimally increase stimulus variance –Short intervals may result in refractory effects Detection ability dependent on form of HDR Length of “event” may not be known

93. 3. Mixed Designs

94. 3a. Combination Blocked/Event Both blocked and event-related design aspects are used (for different purposes) –Blocked design is used to evaluate state-dependent effects –Event-related design is used to evaluate item-related effects Analyses are conducted largely independently between the two measures –Cognitive processes are assumed to be independent

95. …… Mixed Blocked/Event-related Design Target-related Activity (Phasic) Blocked-related Activity (Tonic) Task-Initiation Activity (Tonic) Task-Offset Activity (Tonic)

96. Mixed designs Donaldson et al., 2001

97. 3b. Variable Stimulus Probability Stimulus probability is varied in a blocked fashion –Appears similar to the combination design Mixed design used to maximize experimental power for single design Assumes that processes of interest do not vary as a function of stimulus timing –Cognitive processing –Refractory effects

98. Random and Semi-Random Designs From Liu et al., 2001

99. 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