Statistical analysis of PET data using FMRISTAT Keith

Statistical analysis of PET data using FMRISTAT (!) Keith Worsley Department of Mathematics and Statistics, Mc. Connell Brain Imaging Centre, Mc. Gill University

CBF non kinetic Task l Unnormalized data (z = -6 mm) c: /keith/f. MRI/manou/cbf non kin/bach allan -h 1 tal 200008161003. mnc, slice 21 6 11 16 21 25 30 35 40 45 50 55 60 65 1 1 x 10 4 3. 5 7 12 17 22 26 31 36 41 46 51 56 61 66 3 8 13 18 23 27 32 37 42 47 52 57 62 67 3 2 2 2. 5 3 2 9 14 19 24 28 33 38 43 48 53 58 63 68 4 4 1. 5 base 1 5 10 15 20 29 34 Subjects 39 44 49 54 59 64 69 0. 5 0 4

CBF non-kinetic Normalized: thresh at ½ max, average, divide Task l c: /keith/f. MRI/manou/cbf non kin/normalized, slice 21 6 11 16 21 25 30 35 40 45 50 55 60 65 2 7 12 17 22 26 31 36 41 46 51 56 61 66 1. 5 2 1 1 1 8 13 18 23 27 32 37 42 47 52 57 62 67 4 9 14 19 24 28 33 38 43 48 53 58 63 68 5 10 15 20 29 34 39 44 49 54 59 64 69 base 4 3 3 Subjects 0. 5 0

Correlation models Independent scans DF: (#subj-1) × (#scans-1) = 51 bias SPM? Autocorrelation AR(1) Autocorrelation AR(2) Depends on correlations, contrast All correlations (#subj-1) = 13 Standard error of contrasts: variance Safest: DOT, FMRISTAT boost df by pooling/smoothing

Is pooling sd valid? Is sd constant across the brain? Unsmoothed sd assuming independent scans, 51 df: Pooled Sd = 0. 027

FMRISTAT: smoothing instead of pooling Effective df depends on FWHMsd: dfeff = dfresidual pooled sd, dfeff = infinity 100 dfeff no smoothing, dfeff = 51 dfeff = 13 ) 3/2 e. g. FWHMdata = 8. 3 mm: Infinity Target = 100 df ( FWHMsd 2 2 +1 2 FWHMdata t n e d n pe 0 s n o e d in 50 all 0 5 rr o c ti a l e FWHM = 10. 0 mm 10 FWHM = 4. 4 mm FWHMsd Infinity

FMRISTAT: smooth sd by FWHM = 4. 4 mm, df = 100

CBF non kinetic (z = 57 mm) Effect is always same T statistic SD: Voxel Smooth Pooled DF: 51 100 infinite

Stat_summary task 4 – base CLUSTERS: clus 1 2 3 4 5 6 vol resel Pval (one) 4688 10. 92 0 ( 0) 7067 10. 58 0 ( 0) 947 1. 6 0. 033 (0. 001) 543 0. 88 0. 22 ( 0. 01) 415 0. 75 0. 323 (0. 015) 169 0. 39 0. 788 ( 0. 06) Effective FWHM: 15 PEAKS: clus 2 1 1 2 2 1 1 1 2 3 2 9 3 5 peak 8. 03 6. 8 6. 58 6. 34 6. 33 5. 41 5. 4 5. 38 5. 25 5. 08 5 4. 9 4. 85 4. 81 Pval 0 0 0. 019 0. 021 0. 037 0. 076 0. 104 0. 155 0. 191 0. 223 (one) ( 0) ( 0) (0. 001) (0. 002) (0. 004) (0. 005) (0. 007) (0. 008) ( 0. 01) Qval 0 0 0 0 0. 001 ( ( ( (i 35 81 76 39 42 70 68 69 30 17 31 14 16 65 j 62 40 43 59 64 42 41 42 57 76 59 67 75 77 k) 64) 13) 11) 67) 70) 12) 13) 63) 42) 64) 12) 42) 61) ( x (-38. 9 ( 22. 8 ( 16. 1 (-33. 5 (-29. 5 ( 8 ( 5. 4 ( 6. 7 (-45. 6 ( -63 (-44. 2 ( -67 (-64. 3 ( 1. 3 y -19. 4 -57. 3 -52. 1 -24. 6 -16 -53. 8 -55. 6 -53. 8 -28 4. 6 -24. 6 -10. 8 2. 9 6. 4 z ) 58. 5) -18) -21) 63) 67. 5) -19. 5) -18) 57) 25. 5) 58. 5) -19. 5) 25. 5) 54) 10 5 0

CBF kinetic Task l Unnormalized, z= -6 mm c: /keith/f. MRI/manou/cbf kin/bach allan -h 1 sumpetkinetic tal 200008161003. mnc, slice 21 6 11 19 24 29 34 39 44 49 54 59 64 1 1 80 70 7 12 3 8 13 20 25 30 35 40 45 50 55 60 65 21 26 31 36 41 46 51 56 61 66 60 2 2 16 50 3 40 9 14 17 22 27 32 37 42 47 52 57 62 67 4 4 30 base 20 5 10 15 18 23 28 33 Subjects 38 43 48 53 58 63 68 10 0

Is pooling sd valid? Is sd constant across the brain? Unsmoothed sd assuming independent scans, 51 df: Pooled Sd = 7. 2

CBF kinetic (z = 57 mm) Effect is always same T statistic SD: Voxel Smooth Pooled DF: 51 100 infinite

T stat, smoothed sd, 100 df CBF non kinetic vs. kinetic

Stat_summary task 4 – base (search region is where CBF non kinetic T > 5) CLUSTERS: clus vol 1 1338 2 363 resel Pval (one) 3. 11 0 (0. 002) 1 0. 008 (0. 043) Effective FWHM: PEAKS: clus 1 1 1 2 2 2 1 1 1 6 2 2 1 3 1 2 peak 6. 24 5. 32 5. 14 4. 82 4. 48 4. 32 4. 3 4. 19 4. 16 3. 91 3. 64 3. 52 3. 5 3. 42 3. 33 3. 16 Pval 0 0. 001 0. 003 0. 005 0. 007 0. 016 0. 034 0. 047 0. 05 0. 062 0. 077 0. 116 (one) ( 0) (0. 001) (0. 003) (0. 009) (0. 015) (0. 021) (0. 023) (0. 049) (0. 099) (0. 135) (0. 143) (0. 175) (0. 214) (0. 313) Qval 0 0 0 0 0. 001 0. 002 0. 003 (i (33 (32 (35 (79 (78 (81 (40 (32 (31 (82 (39 (30 (38 (82 j 63 62 63 43 44 42 62 61 60 59 41 41 58 58 65 38 k) 61) 63) 60) 11) 12) 11) 70) 69) 66) 64) 10) 11) 69) 64) 68) 10) ( x (-41. 5 (-42. 9 (-38. 9 ( 20. 1 ( 18. 8 ( 22. 8 (-30. 8 (-32. 2 (-42. 9 (-44. 2 ( 22. 8 ( 24. 1 (-33. 5 (-45. 6 (-34. 8 ( 24. 1 y -17. 7 -19. 4 -17. 7 -52. 1 -50. 4 -53. 8 -19. 4 -21. 2 -22. 9 -24. 6 -55. 6 -26. 3 -14. 3 -60. 7 z ) 54) 57) 52. 5) -21) -19. 5) -21) 67. 5) 66) 61. 5) 58. 5) -22. 5) -21) 66) 58. 5) 64. 5) -22. 5) 15 10 5 0

CMRO kinetic Unnormalized, z = -6 mm c: /keith/f. MRI/manou/cmro kin/bach allan -o 1 sumpetkinetic tal 200008161158. mnc, slice 21 1 6 11 16 21 26 31 36 2 7 12 17 22 27 32 37 45 50 55 60 65 46 51 56 61 66 300 1 Task l 2 41 250 200 3 8 13 18 23 28 33 38 42 47 52 57 62 67 3 150 9 14 19 24 29 34 39 43 48 53 58 63 68 5 10 15 20 25 30 35 40 44 49 54 59 64 69 base 4 4 Subjects 100 50 0

CMRO kinetic l Unnormalized, smoothed 16 mm FWHM

Is pooling sd valid? Is sd constant across the brain? Unsmoothed sd assuming independent scans, 48 df: Pooled Sd = 12. 7

CMRO kinetic T statistic, z = 57 mm SD: Voxel Smooth Pooled DF: 48 100 infinite T statistic, z = -6 mm SD: Voxel Smooth Pooled DF: 48 100 infinite

CMRO kinetic smoothed 16 mm, “safe” analysis of task 3 - base Sd 0 18 1 16 1 24 1 32 1 40 1 48 1 56 1 64 1 72 1 Unsmoothed sd, 12 df Sd smoothed 26 mm, 100 df 20 0 28 2 16 2 24 2 32 2 40 2 48 2 56 2 64 2 72 2 10 0 T statistics Unsmoothed 0 sd, 12 df T = 6. 79 18 Sd smoothed 0 26 mm, 100 df T = 4. 14 28 0 38 Pooled sd, infinite df T = 3. 93 30 1 16 1 24 1 32 1 40 1 48 1 56 1 64 1 72 1 6 4 2 16 2 24 2 32 2 40 2 48 2 56 2 64 2 72 2 2 3 16 3 24 3 32 3 40 3 48 3 56 3 64 3 72 3 0 -2

Stat_summary task 3 – base CLUSTERS: clus 1 2 3 4 vol 927 211 138 266 resel 0. 25 0. 06 0. 05 0. 04 Pval 0. 179 0. 485 0. 545 0. 554 (one) (0. 121) (0. 409) (0. 485) (0. 498) Effective FWHM: 30 PEAKS: clus 1 1 1. . . 4 4 peak 4. 42 4. 41 4. 37 4. 32 40 Pval 0. 065 0. 068 0. 077 0. 089 (one) (0. 034) (0. 036) (0. 041) (0. 047) Qval 0. 651 (i (37 (35 (36 (37 (36 j 107 106 106 107 k) 40) 39) 40) 42) 38) ( x (-36. 2 (-38. 9 (-37. 5 (-36. 2 (-37. 5 y 58 56. 2 58 z ) 22. 5) 21) 22. 5) 25. 5) 19. 5) 20 10 3. 44 1. 072 ( 0. 54) 1. 091 (0. 549) 1. 003 (83 (82 75 75 21) 22) ( 25. 5 ( 24. 1 2. 9 -6) -4. 5) 0