Introduc)on to fmri Natalia Zaretskaya
Content fmri signal fmri versus neural ac)vity A classical experiment: flickering checkerboard Preprocessing Univariate analysis Single- subject analysis Group analysis SoHware
fmri in comparison Spa)al resolu)on Temporal resolu)on
Some fmri study examples
Recommended reading!
fmri signal Hemodynamic response S)mulus Neural ac)vity increase Oxygen consump)on increase (CMR02) Cerebral blood flow increase (CBF) Cerebral blood volume increase (CBV) MR signal changes
Blood Oxygen Level Dependent signal (BOLD) Hemodynamic response S)mulus Neural ac)vity increase Oxygen consump)on increase (CMR02) Cerebral blood flow increase (CBF) Cerebral blood volume increase (CBV) MR signal changes BOLD signal
Neural ac)vity and energy consump)on in the primate brain Postsynap)c ac)vity: 75% Ac)on poten)als: 10% Ac)va)on neural excita)on Deac)va)on neural inhibi)on aher AXwell and Iadecola, TINS 2002 See also Logothe)s et al., 2001&2003
Hemodynamic response hxp://hirnforschung.kyb.mpg.de/en/methods/func)onal- magne)c- resonance- imaging- fmri.html
Hemodynamic response func)on (hrf) 0.12 HRF 0.1 0.08 0.06 0.04 0.02 0-0.02 Brief s)mulus 0 1.5 3 4.5 6 7.5 9 10.5 12 13.5 15 16.5 18 19.5 21 22.5 24 25.5 27 28.5 30 31.5
Image types Func1onal/BOLD/ T2,T2*weighted Structural/anatomical/T1- weigted SagiXal view Coronal view Resolu)on 3x3x3 mm Transverse view Resolu)on: 1x1x1 mm
Image acquisi)on In- plane resolu)on 3x3mm Slice thickness 3mm Voxel size 3x3x3mm
Experimental design Condi1on of interest Checkerboard Control Gray background Finger tapping Rest Faces Houses Cola Pepsi
20 s on 20 s off A typical experiment 2 1.5 1 0.5 0-0.5-1 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0 62.5 65.0 67.5 70.0 72.5 75.0 77.5 80.0 82.5 85.0 87.5 90.0 92.5 95.0 97.5 100.0 102.5 105.0 107.5 110.0 112.5 115.0 117.5 Time (s) 1.5 1 0.5 0 TR 0.0 2.5 5.0 Time (s)
4D func)onal dataset 3D Repe))on )me (TR) )me Voxel 1me course 2.00 1.50 1.00 0.50 0.00-0.50-1.00 Time (s)
Analysis steps Single- subject preprocessing Single- subject analysis (1 st level analysis) Group analysis (2 nd level analysis)
Mo)on correc)on Purpose: compensate for subject movement 2 1 0 0.0 2.5 5.0 Time (s)
Mo)on correc)on Purpose: compensate for subject movement 6 parameter rigid body transform 3x 3x # volume
Slice- )me correc)on Purpose: compensate for the lag in slice acquisi)on ascending interleaved
Slice- )me correc)on Purpose: compensate for the lag in slice acquisi)on 2.5 s 0.12 0.1 0.08 0.06 0.04 Before correc1on 0.02 0-0.02 0.0 7.5 15.0 22.5 30.0 37.5 45.0 52.5 60.0 67.5 75.0 82.5 90.0 97.5 105.0 112.5 Time (s) AIer correc1on 0.12 0 s 0.1 0.08 0.06 0.04 0.02 0 0.0 7.5 15.0 22.5 30.0 37.5 45.0 52.5 60.0 67.5 Time (s) 75.0 82.5 90.0 97.5 105.0 112.5-0.02 0 4 8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32
Normaliza)on to common space Purpose1: compensate individual differences in brain shape for the group analysis hxp://publicdomainreview.org/collec)ons/phrenology- diagrams- from- vaughts- prac)cal- character- reader- 1902/
Templates & coordinate systems Purpose2: Unified coordinate system Talairach MNI hxp://imaging.mrc- cbu.cam.ac.uk/imaging/mnitalairach
Rigid transform Template 6 parameter rigid transform 3x 3x
Affine transform 12 parameter affine transform 3x 3x 3x 3x
Deforma)on field Warped image Nonlinear transform Source Template warping
Co- registra)on Subject 1 Template Group ac)va)on Subject 2 Subject n
Spa)al smoothing Purpose: increase the signal and suppress the noise 0 FWHM 5 FWHM 10 FWHM Full Max Full- Width/Half- max 2mm FWHM Half Max 5mm FWHM 10mm FWHM
More on smoothing later In the sohware sec)on!
Single- subject (first- level) analysis Univariate/voxel- wise analysis Ques)on: Which areas are ac)vated by the flickering checkerboard?
Checkerboard experiment - 1-0.5 0 0.5 1 1.5 2 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0 62.5 65.0 67.5 70.0 72.5 75.0 77.5 80.0 82.5 85.0 87.5 90.0 92.5 95.0 97.5 100.0 102.5 105.0 107.5 110.0 112.5 115.0 117.5 Time (s) S1mulus func1on
Checkerboard experiment - 1-0.5 0 0.5 1 1.5 2 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0 62.5 65.0 67.5 70.0 72.5 75.0 77.5 80.0 82.5 85.0 87.5 90.0 92.5 95.0 97.5 100.0 102.5 105.0 107.5 110.0 112.5 115.0 117.5 Time (s)
Simplest analysis - 1-0.5 0 0.5 1 1.5 2 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0 62.5 65.0 67.5 70.0 72.5 75.0 77.5 80.0 82.5 85.0 87.5 90.0 92.5 95.0 97.5 100.0 102.5 105.0 107.5 110.0 112.5 115.0 117.5 Time (s)
Simplest analysis 2 1.5 1 0.5 0-0.5-1 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0 62.5 65.0 67.5 70.0 72.5 75.0 77.5 80.0 82.5 85.0 87.5 90.0 92.5 95.0 97.5 100.0 102.5 105.0 107.5 110.0 112.5 115.0 117.5 Time (s) T- test OFF ON - 1-0.5 0 0.5 1 1.5 2
Simplest analysis 2 1.5 1 0.5 0-0.5-1 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0 62.5 65.0 67.5 70.0 72.5 75.0 77.5 80.0 82.5 85.0 87.5 90.0 92.5 95.0 97.5 100.0 102.5 105.0 107.5 110.0 112.5 115.0 117.5 Time (s) T- test OFF ON - 1-0.5 0 0.5 1 1.5 2
General Linear Model (GLM) Build a predic)on about the signal Fit predic)on to actual data If it fits nicely there is ac)va)on!
Building predic)on - 1-0.5 0 0.5 1 1.5 2 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 65.0 70.0 75.0 80.0 85.0 90.0 95.0 100.0 105.0 110.0 115.0 Time (s) S1mulus func1on - 0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30 HRF - 1.00-0.50 0.00 0.50 1.00 1.50 2.00 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0 62.5 65.0 67.5 70.0 72.5 75.0 77.5 80.0 82.5 85.0 87.5 90.0 92.5 95.0 97.5 100.0 102.5 105.0 107.5 110.0 112.5 115.0 117.5 Time (s) Regressor (signal predictor) convolu)on
Building regressors - 1-0.5 0 0.5 1 1.5 2 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 65.0 70.0 75.0 80.0 85.0 90.0 95.0 100.0 105.0 110.0 115.0 Time (s) S1mulus func1on - 0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30 HRF - 1.00-0.50 0.00 0.50 1.00 1.50 2.00 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0 62.5 65.0 67.5 70.0 72.5 75.0 77.5 80.0 82.5 85.0 87.5 90.0 92.5 95.0 97.5 100.0 102.5 105.0 107.5 110.0 112.5 115.0 117.5 Time (s) convolu)on
General Linear Model Y(t) = β * X(t) + c + e(t) Measured data Modeled data (regressors) Constant offset Error (e.g. noise, confounds)
General linear model fit Y = Xβ + e β = (X T X) -1 X T y
Sta)s)cal inference (t- test) in GLM Voxel- wise t- sta)s)c t = β/se(β) 3D t- sta)s)c map 72,221 voxels Final result Is related to the variance of error Thresholding: deal with mul)ple comparison problem!
What about two condi)ons? Condi)on A Condi)on B - 1-0.5 0 0.5 1 1.5 2 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0 62.5 65.0 67.5 70.0 72.5 75.0 77.5 80.0 82.5 85.0 87.5 90.0 92.5 95.0 97.5 100.0 102.5 105.0 107.5 110.0 112.5 115.0 117.5 Time (s)
General Linear Model Y(t) = + + c + e(t) Measured data Modeled data (regressor 1) Modeled data (regressor 2) Error (e.g. noise, confounds)
Sta)s)cal inference (t- test) in GLM Voxel- wise t- sta)s)c t = (β1 β2) / SE(β1;β2) 3D t- sta)s)c map 72,221 voxels Final result Thresholding: deal with mul)ple comparison problem!
What about X condi)ons? Design Matrix Condi)ons of interest regressors Nuissance regressors )me Y = Xβ + e
Contrast: linear combina)on of betas Contrast vector: describes how beta- es1mates are combined ( weighted ) Contrast image: a 3D volume containing the result of beta- combina1ons Y(t) = β1*x1(t) + β2* X2(t) + c + e(t) [ 1-1 ] β1- β2
Group analysis (second- level analysis) Subject 1 contrast image Subject 2 contrast image Subject n contrast image Voxel- wise one- sample t- test Is our combina)o n of betas different from zero? 3D t- sta)s)c map Thresholding: deal with mul)ple comparison problem! Final result
GLM advantages Complex experimental designs Regress out uninteres)ng effects/ confounds Es)mate HRF shape 0.12 0.1 0.08 0.06 0.04 0.02 0-0.02? 0 4 8 12 16 20 24 28 32
Different sohware, different philosophies Since 1994 hxp://www.fil.ion.ucl.ac.uk/spm/ Since 1994 hxps://github.com/wanderine/broccoli/ Doug Greve hxp://afni.nimh.nih.gov/afni/ Since 2000 hxp://fsl.fmrib.ox.ac.uk/fsl/fslwiki/ hxps://surfer.nmr.mgh.harvard.edu/fswiki/fsfast
Surface- based analysis advantages 14 mm FWHM Surface- based smoothing 5 mm apart in 3D 25 mm apart on surface Averaging with other )ssue types (WM, CSF) Averaging with other func)onal areas Surface- based inter- subject registra)on slide courtesy of Doug Greve
500 µm 0.6 mm isotropic voxels 3 mm isotropic voxels Weber et al., 2008
500 µm 0.6 mm isotropic voxels 3 mm isotropic voxels Weber et al., 2008
500 µm 0.6 mm isotropic voxels Thank you for your axen)on Also thanks to: Jon Polimeni Andreas Bartels Doug Greve for some of the slides and ideas 3 mm isotropic voxels Weber et al., 2008