I. Improving SNR (cont.) II. Preprocessing

1. I. Improving SNR (cont.)II. Preprocessing BIAC Graduate fMRI Course October 12, 2004

2. Increasing Field Strength

3. Theoretical Effects of Field Strength • SNR = signal / noise • SNR increases linearly with field strength • Signal increases with square of field strength • Noise increases linearly with field strength • A 4.0T scanner should have 2.7x SNR of 1.5T scanner • T1 and T2* both change with field strength • T1 increases, reducing signal recovery • T2* decreases, increasing BOLD contrast

4. Adapted from Turner, et al. (1993)

5. Measured Effects of Field Strength • SNR usually increases by less than theoretical prediction • Sub-linear increases in SNR; large vessel effects may be independent of field strength • Where tested, clear advantages of higher field have been demonstrated • But, physiological noise may counteract gains at high field ( > ~4.0T) • Spatial extent increases with field strength • Increased susceptibility artifacts

9. Trial Averaging • Static signal, variable noise • Assumes that the MR data recorded on each trial are composed of a signal + (random) noise • Effects of averaging • Signal is present on every trial, so it remains constant through averaging • Noise randomly varies across trials, so it decreases with averaging • Thus, SNR increases with averaging

10. Fundamental Rule of SNR For Gaussian noise, experimental power increases with the square root of the number of observations

11. Example of Trial Averaging Average of 16 trials with SNR = 0.6

13. Increasing Power increases Spatial Extent Subject 1 Subject 2 Trials Averaged 4 500 ms 500 ms 16 … 36 16-20 s 64 100 144

14. A B

15. Effects of Signal-Noise Ratio on extent of activation: Empirical Data Subject 1 Subject 2 Number of Significant Voxels VN = Vmax[1 - e(-0.016 * N)] Number of Trials Averaged

16. 1000 Voxels, 100 Active Active Voxel Simulation Signal + Noise (SNR = 1.0) • Signal waveform taken from observed data. • Signal amplitude distribution: Gamma (observed). • Assumed Gaussian white noise. Noise

17. Effects of Signal-Noise Ratio on extent of activation:Simulation Data SNR = 1.00 SNR = 0.52 (Young) Number of Activated Voxels SNR = 0.35 (Old) SNR = 0.25 SNR = 0.15 SNR = 0.10 Number of Trials Averaged

18. Explicit and Implicit Signal Averaging A B r =.82; t(10) = 4.3; p < .001 r =.42; t(129) = 5.3; p < .0001

19. Caveats • Signal averaging is based on assumptions • Data = signal + temporally invariant noise • Noise is uncorrelated over time • If assumptions are violated, then averaging ignores potentially valuable information • Amount of noise varies over time • Some noise is temporally correlated (physiology) • Nevertheless, averaging provides robust, reliable method for determining brain activity

20. II. Preprocessing of FMRI Data

21. 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

23. Quality Assurance

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

27. Slice Timing Correction

28. 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

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

31. Base HDR: 2s TR

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

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

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

35. Interpolation Strategies • Linear interpolation • Spline interpolation • Sinc interpolation

36. Motion Correction

37. Head Motion: Good, Bad,…

38. … and catastrophically bad

39. Why does head motion introduce problems? A B C

40. Simulated Head Motion

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

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

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

44. Effects of Head Motion Correction

45. 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

46. What is the best approach for minimizing the influence of head motion on your data?

49. Coregistration

50. 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