17
Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis Keith Worsley Department of Mathematics and Statistics, McGill University, McConnell Brain Imaging Centre, Montreal Neurological Institute.

Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

  • Upload
    valora

  • View
    27

  • Download
    0

Embed Size (px)

DESCRIPTION

Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis. Keith Worsley Department of Mathematics and Statistics, McGill University, McConnell Brain Imaging Centre, Montreal Neurological Institute. First scan of fMRI data. Highly significant effect, T=6.59. - PowerPoint PPT Presentation

Citation preview

Page 1: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

Spatial smoothing of autocorrelations to control the degrees of freedom in

fMRI analysis

Keith Worsley

Department of Mathematics and Statistics, McGill University,

McConnell Brain Imaging Centre, Montreal Neurological Institute.

Page 2: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

0

500

1000First scan of fMRI data

-5

0

5

T statistic for hot - warm effect

0 100 200 300

870880890 hot

restwarm

Highly significant effect, T=6.59

0 100 200 300

800

820hotrestwarm

No significant effect, T=-0.74

0 100 200 300

790800810

Drift

Time, seconds

fMRI data: 120 scans, 3 scans each of hot, rest, warm, rest, hot, rest, …

T = (hot – warm effect) / S.d. ~ t110 if no effect

Page 3: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

FMRISTAT: fits a linear model for fMRI time series with AR(p) errors

• Linear model: ? ? Yt = (stimulust * HRF) b + driftt c + errort

• AR(p) errors: ? ? ? errort = a1 errort-1 + … + ap errort-p + s WNt

unknown parameters

Page 4: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

0 50 100 150 200 250 300 350-1

0

1

2Alternating hot and warm stimuli separated by rest (9 seconds each).

hot

warm

hot

warm

0 50-0.2

0

0.2

0.4

Hemodynamic response function: difference of two gamma densities

0 50 100 150 200 250 300 350-1

0

1

2Responses = stimuli * HRF, sampled every 3 seconds

Time, seconds

DESIGN example: pain perception

Page 5: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

-0.1

0

0.1

0.2

0.3

First step: estimate the autocorrelationAR(1) model: errort = a1 errort-1 + s WNt

• Fit the linear model using least squares

• errort = Yt – fitted Yt

• â1 = Correlation ( errort , errort-1)

• Estimating errort’s changes their correlation structure slightly, so â1 is slightly biased:

Raw autocorrelation Smoothed 12.4mm Bias corrected â1

~ -0.05 ~ 0~ -0.05 ~ 0

?

Page 6: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

-1

-0.5

0

0.5

1 Hot - warm effect, %

0

0.05

0.1

0.15

0.2

0.25Sd of effect, %

-6

-4

-2

0

2

4

6 T = effect / sd, 100 df

Pre-whiten: Yt* = Yt – â1 Yt-1, then fit using least squares:

Second step: refit the linear model

T > 4.93 (P < 0.05, corrected)

Page 7: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

Why bother to smooth the acor?

• Sample variability in estimated acor adds variability to sd

• Lowers effective

df of T statistic

• Increases

threshold

• Less power

• Particularly after

correction for search 0 50 100 150

0

2

4

6

8

10

12

14

Df

Th

resh

old

Corrected for whole brain search

One voxel

Page 8: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

Gautama et al. (2005): Smooth autocorrelations, choose amount of smoothing to optimally predict autocorrelations using e.g. cross-validation, model selection.

Page 9: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

Effect of variability in sample acor on dbn of T: first idea

• Why not write linear model with e.g. AR(1) errors

Yt = xt’β + ηt, ηt = a1ηt-1 + εt

where εt iid ~N(0,σ2), as

Yt = a1Yt-1 + xt’β + xt-1’(a1β) + εt

• Least-squares estimates are ~max like, so

• Non-linear l.s.: dfeff ~ n-(#a)-(#β) …. ???? or

• Linear l.s.: dfeff ~ n-(#a)-(#β)-(#a)×(#β) …. ????

• Doesn’t work (see later) because: – design matrix is random?– ~max like only for large samples i.e. df = ∞?

Page 10: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

Better idea: Harville et al. (1974), …, Kenward, Roger (1997) … SAS PROC MIXED …

• Linear model at a single voxel:

Y ~ Nn(Xβ, V(θ)), θ = (σ2, a1, …, ap)

• Fit by ReML, interested in effect

E = c’β, S = Sd(E)

• T = E / S

• E depends on β, S depends on θ

• β, θ ~independent so variability in θ only affects S

Page 11: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

• S depends on θ, and from ReML theory we know ~mean, ~variance of θ.

• Use linear approx to S2(θ) to find ~mean, ~variance of S2

• dfeff is surrogate for variability of S2:

dfeff := 2 E(S2)2/Var(S2)

• Satterthwaite: S2 ~ cons×χ2dfeff , T ~ tdfeff

Continued …

Page 12: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

Expression for dfeff

• dfeff depends on contrast(!) and θ,

– Could plug in θ, but don’t know θ in advance– Explicit expression if acors = 0– Hope it is a good approx for when acors ≠ 0

• Contrast in obs: x = X(X’X)-1c, so E = x’Y

• τj = lag j acor of x, dfresidual = least-squares df

• 1/dfeff = 1/dfresidual + 2(τ12 + … + τp

2)/dfresidual

Page 13: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

Effect of smoothing acor

• Assume ε ~ white noise smoothed by Gaussian filter, width FWHMdata, GRF(FWHMdata)

• Autocors ~ GRF(FWHMdata/√2)

• Smoothing acors in D dimensions by FWHMacor reduces variance by

f = (2 FWHMacor2/FWHMdata

2 + 1)D/2

• Define dfacor := f dfresidual

• 1/dfeff = 1/dfresidual + 2(τ12 + … + τp

2)/dfacor

Page 14: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

0 1 2 3 40

20

40

60

80

100

120

Sim, a1=

00.10.20.30.4

Hot, 1=0.61

FWHMf ilter

/FWHMdata

Eff

ectiv

e df

Residual df = 114

Theory,a

1=0

x=

0 1 2 3 40

20

40

60

80

100

120

Sim, a1=

00.10.20.30.4

Hot + Warm, 1=0.5

FWHMf ilter

/FWHMdata

Eff

ectiv

e df

Residual df = 114

Theory,a

1=0

x=

0 1 2 3 40

20

40

60

80

100

120

Sim, a1=

00.10.20.30.4

Hot - Warm, 1=0.79

FWHMf ilter

/FWHMdata

Eff

ectiv

e df

Residual df = 114

Theory,a

1=0

x=

0 1 2 3 40

20

40

60

80

100

120

Sim, a1=

00.10.20.30.4

Cubic drift, 1=0.94

FWHMf ilter

/FWHMdata

Eff

ectiv

e df

Residual df = 114

Theory,a

1=0

x=

Page 15: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

0 10 20 30

0

50

100

FWHMacor

0 10 20 300

50

100

FWHMacor

Summary

Applications: Hot stimulus Hot-warm stimulus

Target = 100 df

Residual df = 110

Target = 100 df

Residual df = 110

FWHM = 10.3mm FWHM = 12.4mm

dfacor = dfresidual(2 + 1) 1 1 2 acor(contrast of data)2

dfeff dfresidual dfacor

FWHMacor2 3/2

FWHMdata2

= +

• Variability in acor lowers df• Df depends on contrast • Smoothing acor brings df back up:

Contrast of data, acor = 0.79Contrast of data, acor = 0.61

FWHMdata = 8.79

dfeff dfeff

Page 16: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

0

0.2

0.4

Autocorrelation a1

No

smoo

thin

g

Effective df = 110

0

0.2

0.4

12.4

mm

FW

HM

sm

ooth

ing

Effective df = 1249

-5

0

5

T statistic for hot-warm

Effective df = 49

-5

0

5

Effective df = 100

P = 0.05, corrected

Threshold = 5.25

Threshold = 4.93

Application: Hot – warm stimulus

Page 17: Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis

Refinements

• Could get a rough estimate of acor first, then use this to get better estimate of dfeff, but this is time consuming

• Acor varies spatially, so dfeff varies spatially, but we don’t have any random field theory for P-values

• Could use spatially varying filter to achieve ~constant dfeff, but again this is time consuming

• All the theory built on asymptotic and/or questionable assumptions, so maybe can’t take it too far …