fmri pre-processing Juergen Dukart
Outline Why do we need pre-processing? fmri pre-processing Slice time correction Realignment Unwarping Coregistration Spatial normalisation Smoothing
Overview fmri time-series kernel Design matrix Statistical Parametric Map Motion correction Smoothing General Linear Model (Co-registration and) Spatial normalisation Parameter Estimates Standard template
Why do we need pre-processing?
What do we want?
Movement
Distortions
Inter-subject variability
Non-gaussian distribution
fmri pre-processing Slice timing correction (optional) Realignment (Motion correction) Unwarping (Distortion correction, optional) Co-registration Link functional scans to anatomical scan Spatial normalisation (unified segmentation) Fitting images to a standard brain Smoothing Increases signal-to-noise ratio and approximates a Gaussian distribution
Slice timing (optional) TR: 2 sec 1.2 1 1.2 1 Slice1 Slice11 Slice12 0.8 Slice1 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 TR:-1 TR:0 TR:1 TR:2 TR:3 TR:4 TR:5 0 TR:-1 TR:0 TR:1 TR:2 TR:3 TR:4 TR:5
Slice timing (optional) 1.8 1.6 1.4 Slice1 Slice11 Slice12 1.2 1 0.8 0.6 0.4 0.2 0 TR:-1 TR:0 TR:1 TR:2 TR:3 TR:4 TR:5
fmri pre-processing Slice timing correction (optional) Realignment (Motion correction) Unwarping (Distortion correction, optional) Co-registration Link functional scans to anatomical scan Spatial normalisation (unified segmentation) Fitting images to a standard brain Smoothing Increases signal-to-noise ratio and approximates a Gaussian distribution
Realignment (motion correction) Translation Z Rotation Yaw Roll X Pitch Y
Realignment (motion correction) 1 0 0 Xtrans Translations 0 1 0 Ytrans 0 0 1 Zt rans 0 0 0 1 Rigid body transformations parameterised by: Pitch about X axis 1 0 0 0 0 cos( ) sin( ) 0 0 sin( ) cos( ) 0 0 0 0 1 Roll about Y axis cos( ) 0 sin( ) 0 0 1 0 0 sin( ) 0 cos( ) 0 0 0 0 1 Yaw about Z axis cos( ) sin( ) 0 0 sin( ) cos( ) 0 0 0 0 1 0 0 0 0 1 Minimizing the squared difference (error) between the images Squared Error
Realignment (motion correction
fmri pre-processing Slice timing correction (optional) Realignment (Motion correction) Unwarping (Distortion correction, optional) Co-registration Link functional scans to anatomical scan Spatial normalisation (unified segmentation) Fitting images to a standard brain Smoothing Increases signal-to-noise ratio and approximates a Gaussian distribution
Distortion correction (unwarp) Fieldmap Raw EPI Undistorted EPI
Unwarp can estimate changes in distortion from movement Resulting field map at each time point Measured field map Estimated change in field wrt change in pitch (x-axis) Estimated change in field wrt change in roll (y-axis) 0 = + + distortions in a reference image (FieldMap) subject motion parameters (that we obtain in realignment) change in deformation field with subject movement (estimated via iteration)
fmri pre-processing Slice timing correction (optional) Realignment (Motion correction) Unwarping (Distortion correction, optional) Co-registration Link functional scans to anatomical scan Spatial normalisation (unified segmentation) Fitting images to a standard brain Smoothing Increases signal-to-noise ratio and approximates a Gaussian distribution
Co-registration Normalized mutual information functional and structural images in the same space
Co-registration T2 intensity T1 intensity T2 intensity T1 intensity
Spatial normalization Normalizes structural images to a standard brain template (standard space) The obtained transformation (warping) parameters can be applied on co-registered fmri data Improved spatial normalization based on high resolution structural information
fmri pre-processing Slice timing correction (optional) Realignment (Motion correction) Unwarping (Distortion correction, optional) Co-registration Link functional scans to anatomical scan Spatial normalisation (unified segmentation) Fitting images to a standard brain Smoothing Increases signal-to-noise ratio and approximates a Gaussian distribution
Smoothing
Smoothing
Thank you for attention