FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to...
Transcript of FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to...
![Page 1: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/1.jpg)
FMRI Pre-Processing and Model-Based Statistics
• Brief intro to FMRI experiments and analysis
• FMRI pre-stats image processing
• Simple Single-Subject Statistics
• Multi-Level FMRI Analysis
• Advanced FMRI Analysis
![Page 2: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/2.jpg)
FMRI Pre-Statistics
• Brief intro to FMRI analysis
FMRI pre-statistical image processing:
• Reconstruction from k-space data• Motion correction• Slice timing correction• Spatial filtering• Temporal filtering• Global intensity normalisation
![Page 3: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/3.jpg)
FMRI Experiments
• Simple paradigm design:- stimulus vs baseline- constant stimulus “intensity”- constant block lengths - many repetitions: ABABA
• Need baseline (rest) condition to measure change
A(baseline)
A(baseline)
A(baseline)
B(stimulus)
B(stimulus)
time (TRs)0 1 2 3 4 5
e.g. flashing chequerboard
Stimulus
![Page 4: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/4.jpg)
The Haemodynamic Response
• Field changes (perturbations) --> dephasing --> T2* effect• BOLD-tuned MRI (T2*-weighted) is sensitive to this effect
venulesarterioles
= Hbr= HbO2
capillary bed
capillary bed
venulesarterioles
Basal State Activated State
CBF CBV CMRO2Field
changeMRIsignal
Activation leads to:
[Hbr]
![Page 5: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/5.jpg)
Predicted Response
⊗ =HRF Predicted neural activity
Predicted response
• The process can be modelled by convolving the activity curve with a "haemodynamic response function" or HRF
time
![Page 6: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/6.jpg)
FMRI Experiments: Analysis
measured timeseries at marked voxel
time
• Each voxel contains a time-varying signal (BOLD signal)
![Page 7: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/7.jpg)
FMRI Experiments: Analysis
• Find which voxels have signals that match the model
• Good match implies activation related to stimulus
predicted response
• Model the stimulus-induced change in BOLD signal (predicted response)
time
BOLD response, %
time
initialdip
post stimulusundershootovershoot
1
2
3
0
stimulus
measured timeseries at marked voxel
• Each voxel contains a time-varying signal (BOLD signal)
![Page 8: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/8.jpg)
• Correlate model at each voxel separately
• Measure residual noise variance
• t-statistic = model fit / noise amplitude
Standard GLM Analysis
• Threshold t-stats and display map
![Page 9: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/9.jpg)
Predicted Response (model)
FittedAmplitude
![Page 10: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/10.jpg)
Residual Noise
![Page 11: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/11.jpg)
• Correlate model at each voxel separately
• Measure residual noise variance
• t-statistic = model fit / noise amplitude
Signals of no interest (e.g. artifacts) can affect both activation strength and
residual noise variance
Use pre-processing to reduce/eliminate some of these effects
• Threshold t-stats and display map
Standard GLM Analysis
![Page 12: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/12.jpg)
FMRI Pre-Statistics
FMRI pre-statistical image processing:
• Reconstruction from k-space data• Motion correction• Slice timing correction• Spatial filtering• Temporal filtering• Global intensity normalisation
![Page 13: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/13.jpg)
Image Reconstruction• Convert k-space data to images:
• reconstruction algorithms
• Correct using custom-built initial analysis stages• Scanner artefacts can be found by:
• Occasionally get problematic data• e.g. slice timing errors, RF spikes, RF interference
![Page 14: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/14.jpg)
FMRI Pre-Statistics
FMRI pre-statistical image processing:
• Reconstruction from k-space data• Motion correction• Slice timing correction• Spatial filtering• Temporal filtering• Global intensity normalisation
![Page 15: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/15.jpg)
Motion Correction: Why?• People always move in the scanner
• Even with padding around the head there is still some motion
• Need each voxel to correspond to a consistent anatomical point for each point in time
• Motion correction realigns to a common reference
• Very important correction as small motions (e.g. 1% of voxel size) near strong intensity boundaries may induce a 1% signal change > BOLD
![Page 16: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/16.jpg)
Motion Correction
Register each FMRI volume to target separately
Use rigid body (6 DOF)
= multiple registration
Select a MC target (reference) for all FMRI volumes.
Can use either one original volume, mean of several, standard space
image etc.
![Page 17: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/17.jpg)
Effect of Motion Correction
Uncorrelated Motion
Stimulus Correlated Motion
Without MC With MC
![Page 18: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/18.jpg)
Motion Parameter Output
Summary of total motion (relative and absolute)
Note that large jumps are more serious than slower drifts, especially in the relative motion plot
Relative = time point to next time point - shows jumpsAbsolute = time point to reference - shows jumps & drifts
![Page 19: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/19.jpg)
FMRI Pre-Statistics
FMRI pre-statistical image processing:
• Reconstruction from k-space data• Motion correction• Slice timing correction• Spatial filtering• Temporal filtering• Global intensity normalisation
![Page 20: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/20.jpg)
Slice Timing Correction
Almost all FMRI scanning takes each slice separately
Each slice is scanned at a slightly different time
Slice order can be interleaved (as shown) or sequential (up or down)
![Page 21: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/21.jpg)
Slice TimingWithout any adjustment, the model timing is always the same
Model
slice 10
slice 9
Slice 9 acquired 1.5 secs after
slice 10 acquisition timing (TRs)
![Page 22: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/22.jpg)
slice 9
slice 10
Slice Timing... but the timing of each slice’s data
is differentData
Slice 9 acquired 1.5 secs after
slice 10 acquisition timing (TRs)
![Page 23: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/23.jpg)
Slice TimingCan get consistency by shifting
the dataData
slice 10
slice 9
Slice 9 acquired 1.5 secs after
slice 10 acquisition timing (TRs)
![Page 24: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/24.jpg)
Slice Timing... and then interpolating the data
= slice timing correctionData
slice 10
slice 9
Slice 9 acquired 1.5 secs after
slice 10 acquisition timing (TRs)
![Page 25: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/25.jpg)
Slice Timing... and then interpolating the data
= slice timing correctionData
slice 10
slice 9
Slice 9 acquired 1.5 secs after
slice 10 acquisition timing (TRs)
![Page 26: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/26.jpg)
slice 9
slice 10
Slice TimingThe result of slice timing correction is that the data is changed (degraded) by interpolation
Data
Slice 9 acquired 1.5 secs after
slice 10 acquisition timing (TRs)
![Page 27: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/27.jpg)
Slice TimingAlternatively, can get consistency by
shifting the modelModel
slice 10
slice 9
Slice 9 acquired 1.5 secs after
slice 10 acquisition timing (TRs)
![Page 28: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/28.jpg)
Slice Timing
One way to shift the model is to use the
temporal derivative in
the GLM
Shifted Model
Original Model
Temporal Derivative
=
-Based on Taylor approx:m(t+a) = m(t) + a.m′(t)
![Page 29: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/29.jpg)
Slice Timing
Shifting the model also accounts for variations in the HRF delay • as the HRF is known
to vary between subjects, sessions, etc.
This is the recommended solution
for slice timing
![Page 30: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/30.jpg)
Motion Problems
Such artefacts can severely degrade functional resultsSeverity usually worse for stimulus-correlated motion
Other motion artefacts persist including: Spin-history changes, B0 (susceptibility) interactions & Interpolation effects
Motion correction eliminates gross motion changes but assumes rigid-body motion - not true if slices acquired at different times
![Page 31: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/31.jpg)
Motion Problems
Some potential analysis remedies for motion artefacts include:1. - including motion parameter regressors in GLM2. - removing artefacts with ICA denoising3. - outlier timepoint detection and exclusion (via GLM)4. - rejection of subjects displaying “excessive” motion
No simple rule of thumb defining "too much" motion
Such artefacts can severely degrade functional resultsSeverity usually worse for stimulus-correlated motion
Other motion artefacts persist including: Spin-history changes, B0 (susceptibility) interactions & Interpolation effects
Motion correction eliminates gross motion changes but assumes rigid-body motion - not true if slices acquired at different times
![Page 32: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/32.jpg)
FMRI Pre-Statistics
FMRI pre-statistical image processing:
• Reconstruction from k-space data• Motion correction• Slice timing correction• Spatial filtering• Temporal filtering• Global intensity normalisation
![Page 33: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/33.jpg)
Spatial FilteringWhy do it?1. Increases signal to noise ratio
if size of the blurring is less than size of activation
However:• Reduces small activation areas• Safest option is to do a small amount of smoothing• Alternative thresholding/stats eliminates the need for smoothing (e.g. randomise, TFCE)
2. Need minimum "smoothness" to use Gaussian random field theory for thresholding
![Page 34: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/34.jpg)
Spatial Filtering: How?
Each voxel intensity is replaced by a weighted average of neighbouring intensities
A Gaussian function in 3D specifies weightings and neighbourhood size
Spatial filtering done by a 3D convolution with a Gaussian (cf. 1D convolution with HRF for model)
1.0
0.4
0.8
0.8
0.8 0.8
0.4
0.4 0.4
0.6 0.6
0.60.6
0.3 0.3
0.30.3
0.3 0.3
0.30.3
0.1 0.1
0.10.1
Weights
![Page 35: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/35.jpg)
Spatial Filtering: How?
Each voxel intensity is replaced by a weighted average of neighbouring intensities
Spatial filtering done by a 3D convolution with a Gaussian (cf. 1D convolution with HRF for model)
1.0
0.4
0.8
0.8
0.8 0.8
0.4
0.4 0.4
0.6 0.6
0.60.6
0.3 0.3
0.30.3
0.3 0.3
0.30.3
0.1 0.1
0.10.1
Specify amount by Full Width Half Maximum (FWHM) = distance between 0.5 values
FWHM
Weights
FWHM0
0.5
1
![Page 36: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/36.jpg)
Spatial Filtering: Results at Different FWHM
![Page 37: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/37.jpg)
FMRI Pre-Statistics
FMRI pre-statistical image processing:
• Reconstruction from k-space data• Motion correction• Slice timing correction• Spatial filtering• Temporal filtering• Global intensity normalisation
![Page 38: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/38.jpg)
Temporal Filtering: Why?• Time series from each voxel contain scanner-related and physiological signals + high frequency noise• Scanner-related and physiological signals (cardiac cycle, breathing etc) can have both high and low frequency components• These signals + noise hide activation
What is temporal filtering?
• Removal of high frequencies, low frequencies or both, without removing signals of interest
Raw Signal
![Page 39: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/39.jpg)
Temporal Filtering: Highpass
• Removes low frequency signals, including linear trend
• Must choose cutoff frequency carefully (lower than frequencies of interest = longer period)
Raw Signal Highpass Filtered
![Page 40: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/40.jpg)
Temporal Filtering: Lowpass
Removes high frequency noise
Only useful if the predicted model does not also contain high frequencies...
Lowpass FilteredRaw Signal
![Page 41: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/41.jpg)
Filtering & Temporal Autocorrelation
• In these cases, lowpass filtering removes too much signal
Some designs also contain high frequency content in the model e.g. Dense Single-Event Model:
• Also, need noise data to correctly estimate autocorrelation (to make statistics valid - see later) ➜ avoid lowpass filtering
⊗
Recommendations:- Use Highpass only- Ensure cutoff frequency higher than model frequencies (can use the Estimate button in the GUI - see practical)- Lower limit on cutoff frequency for good autocorrelation estimation (e.g. for TR=3s, cutoff period > 90s )
Spectrum of model
![Page 42: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/42.jpg)
Effect of Temporal Filtering
No Temporal Filtering
Highpass Temporal Filtering
![Page 43: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/43.jpg)
FMRI Pre-Statistics
FMRI pre-statistical image processing:
• Reconstruction from k-space data• Motion correction• Slice timing correction• Spatial filtering• Temporal filtering• Global intensity normalisation
![Page 44: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/44.jpg)
Global Intensity Normalisation• Mean intensity of the whole
dataset changes between subjects and sessions• due to various uninteresting
factors (e.g. caffeine levels)
• Want the same mean signal level for each subject (taken over all voxels and all timepoints: i.e. 4D)
• Scale each 4D dataset by a single value to get the overall 4D mean (dotted line) to be the same
• Automatically done within FEAT
Time (TR)M
ean
Inte
nsity
(o
ver
3D v
olum
e)
Subject 1
Subject 2
Subject 3
![Page 45: FMRI Pre-Processing and Model- Based Statistics...Image Reconstruction • Convert k-space data to images: • reconstruction algorithms • Correct using custom-built initial analysis](https://reader033.fdocuments.in/reader033/viewer/2022060302/5f08a43c7e708231d42302f9/html5/thumbnails/45.jpg)
SummaryReconstruction Create image and
remove gross artefacts
Motion Correction Get consistent anatomical coordinates (always do this)
Slice Timing Get consistent acquisition timing (use temporal derivative instead)
Spatial Smoothing Improve SNR & validate GRF
Temporal Filtering Highpass: Remove slow driftsLowpass: Avoid for autocorr est.
Intensity Normalisation 4D: Keeps overall signal mean constant across sessions