Académique Documents
Professionnel Documents
Culture Documents
Timetable
FSL Course
Basic Anatomy
The brain is full of neurons. These are organised into two types of tissues: - Grey Matter - White Matter
Post-Mortem
CSF
Grey Matter
MRI Neurons
White Matter
Basic Anatomy
Neurons are densely connected and have many dendrites Axons conduct electrical signals and are surrounded by myelin
Diffusion MRI
Mackay et al.
THERAPY
Functional MRI
Johansen-Berg, et al., Brain 2002
Functional MRI
The APOE 4 allele is a risk factor for late-life pathological changes that is also associated with anatomical and functional brain changes in middle-aged and elderly healthy subjects. We investigated structural and functional effects of the APOE polymorphism in 18 young healthy APOE 4-carriers and 18 matched noncarriers (age range: 20 35 years). Brain activity was studied both at rest and during an encoding memory paradigm using blood oxygen level-dependent fMRI. Resting fMRI revealed increased default mode network (involving retrosplenial, medial temporal, and medial-prefrontal cortical areas) coactivation in 4-carriers relative to noncarriers. The encoding task produced greater hippocampal activation in 4-carriers relative to noncarriers. Neither result could be explained by differences in memory performance, brain morphology, or resting cerebral blood ow. The APOE 4 allele modulates brain function decades before any clinical or neurophysiological expression of neurodegenerative processes.
hippocampus memory neuroimaging resting connectivity
polipoprotein E (apoE, protein; APOE, gene) is a very-lowdensity lipoprotein that removes cholesterol from the blood
fMRI studies have tested for early life associations of the APOE polymorphism with changes in brain function. Filbey et al. (18) reported greater activation in 8 APOE 4-carriers compared with 8 noncarriers in medial frontal and anterior cingulate areas using a working memory paradigm. Mondadori et al. (17) reported reduced activation with an associative learning paradigm in 13 4-carriers relative to 11 2-carriers and 10 3-homozygotes. Both studies therefore suggest that the APOE genotype influences brain functions even early in adulthood. Here, we used a multimodal MRI protocol to investigate structural and functional neurophysiological characteristics of 18 APOE 4-carriers and 18 noncarriers, with ages ranging from 20 to 35 years old. Our first aim was to measure differences in spontaneous fluctuations in resting brain function in 4-carriers relative to noncarriers using resting-state fMRI. Brain regions showing a strong temporal coherence (coactivation) in lowfrequency fluctuations (less than 0.1 Hz) are defined as resting state networks (RSNs), and they reflect intrinsic properties of functional brain organization (21). We were specifically interested in studying the effect of the APOE 4 allele on an RSN called the default mode network (DMN), which includes the
NEUROSCIENCE
1-P
Lat
Post
Ant
! !
Connectivity information shows link between atrophy in thalamus & amygdala/ hippocampus in AD
(Zarei et al., NeuroImage 2010)
Hippocampus-Thalamus Connectivity
Vertex Analysis
!
FA reduction GM reduction
FA is a surrogate measure of white-matter integrity
FSL Course
Overview
Structural MRI Diffusion MRI Functional MRI Complementary techniques Introduction to using FSL
- Variety of acquisitions - Measurement basics - Limitations & artefacts - Analysis principles - Acquisition tips
Structural MRI
T1-weighted
T2-weighted
Proton Density
Images gross brain anatomy Time depends on SNR & resolution (typ. 5-15 mins) Many different (and good) varieties of sequences to
White
Mini MR Physics
MRI Scanner Cutaway
Radio Frequency Coil Subject Patient Table
B0 Field
M=0
(e.g. molecular tumbling speed) but are sensitive to micro-environment and hence tissue type
Intensity is usually a complicated
Low-res T2-weighted
resolution and noise (Signal to Noise Ratio = SNR) For analysis it is Contrast to Noise Ratio (CNR) that is often more important, and contrast depends on MR sequence too
Non-uniform RF eld causes smooth variations in intensity Can sometimes be very large - less smooth for higher eld (e.g. 7T) & multi-coil arrays Need to compensate in analysis
Hardware/settings
Ghosting
Motion
Wrap-around
should always be avoided
RF Interference
RF Spiking
+ others... e.g. chemical shift (fat-water misaligned) Most are serious but uncommon (easy to spot)
Brain Extraction
Segmentation (structure)
Registration (alignment)
Local shape changes Needed to investigate group-level changes/relations Local thickness changes
Basic tips for acquisition T1-weighted images tend to be the best for the SNR/time/resolution tradeoff (1mm voxels are typical) Use the locally optimised sequence on your scanner - there are many sequences names/types that give T1-weighted images and most are equally good If subjects are likely to move a lot then (ofine) averaging several shorter acquisitions can be better Sub-cortical contrast can be enhanced with different sequences or parameter choices Turn on some fat-suppression (helps brain extraction) Isotropic voxels are much better for analysis in general Do not do upsampling on scanner (sometimes the default)
Double Inversion FLAIR WM nulled Recovery (DIR) By changing timing and signals (RF/gradients) of MR sequence can null certain tissues or ows Useful for highlighting lesions/pathologies Also can give better sub-cortical contrast
e.g. Brainstem; Globus Pallidus internal/external
then need to manually create a mask use mask in registrations and segmentations can also apply this strategy to localised artefacts
For subtle/small pathologies (e.g. diffuse hyper-intensities)
SWI / QSI MT Veno/Angio-grams By sensitising the sequence to different properties can detect other features of tissue: SWI/QSI: S=susceptibility; magnetic eld changes due to iron content (primarily) and myelin/WM MT: Magnetisation Transfer; bound/free water Veno/Angio-grams: ow/blood iron/contrast agent
Structural MRI
Many other types of structural MRI, for example...
Susceptibility-Weighted Imaging (SWI) Quantitative Susceptibility Imaging (QSI) Magnetization Transfer (MT) MR Spectroscopy (MRS) Angiograms & Venograms
Sensitive to: iron (and myelin) chemical species/ environment arteries & veins
Quantitative T1 and T2 maps (relaxometry) tissue structure Myelin maps myelin B0 map (eldmap) B1 map (RF)
Diffusion MRI
integrity, particularly in WM Provides information on anatomical connectivity Need to acquire many directions: 5-30 min scan
tensor
local Local WM structure via tensor: directions, Mean colour coded directions Diffusivity (MD) and Fractional Anistropy (FA) direction surrogate measures of microstructure
Target masks Segmented seed region (thalamus) - based on highest target probability
images
Mini MR Physics
MRI Scanner Cutaway
Radio Frequency Coil Subject Patient Table
B0 Field
Mini MR Physics
Diffusion sensitizing (encoding) gradient
Moving spins in the gradient direction change phase and reduce in coherence (less signal)
Mini MR Physics
Diffusion sensitizing (encoding) gradient
Moving spins in the gradient direction change phase and reduce in coherence (less signal)
Bodammer et al. 2004 [1 0 0], b=1000 [-1 0 0], b=1000 [1image 0 0], b=1000 Hardware-related Top-Down image Bottom-Up Estimated field Eddy Currents dw-image dw-image dw-image Extrapolate estimated
ine b m Co
Top-Down image
distortions and apply to high b-value image. th wi
Top-D after
Eddy currents: both acquisition and analysis xes needed Distortion due to B inhomogeneity (air in sinuses) - acquisition-and-analysis related xes needed Bulk motion is corrected for in acquisition (navigators) Pulsatile motion is more problematic
ing d Top-Down lu c n i Co after corr Estimated field el mb Additional lowb scans needed d o ine M wi Diffusion gradient-directions must be paired th M Top-D o de Bastin et al. 2001 Estimated field Use for after lw ith correction t n CSF-suppressed [1 0 0], b=1000 o e u Bottom-Up image etm mo v o after subject movement Use for b0-image dw-image ve m 0 m g correction en in d t u l nc i Co el mb Register using d o ine M Cross-Correlation wi th No additional M Top-Down od Estimated field el after corr wi tho u
Distortion
nt e m ve o m
tensor-component images
Bodammer et al. 2004 [1 0 0], b=1000 [1 0 0], b=1000 dw-image dw-image mated
0 0], b=1000 Basic stages in[-1 the diffusion analysis pipeline: dw-image
pply mage.
Simulated [1 0 0], b=1000 [1 0 0], b=1000 Tensor No additional scansdw-image needed dw-image
Probabilistic Tractography
Tractography-based Segmentation
Basic tips for acquisition Best parameters can be quite hardware dependent (esp. gradients) so check what is optimised for your scanner In general, b-value of 1000-1500 s/mm2 and 60+ directions (tractography) or 12+ directions (FA, etc.) tend to give good results (but the more directions the better) Get one b=0 image for every 8-10 diffusion-weighted images Get even distribution of directions on a sphere Choose a sequence that compensates for eddy currents (e.g. twice refocussed sequence, modied S-T, etc.) Get a eldmap (B0) for distortion correction - or alternatively, a blip-up-blip-down sequence (FSL tool coming very soon) Isotropic voxels (or close) are better for analysis generally Do not do upsampling on scanner (sometimes the default) Both single/multi-shell give good tractography (single for FSL now)
Functional MRI
Measures haemodynamic response to neural activity Task-based or resting-state-connectivity Intrinsic contrast (BOLD) or explicit tag (ASL) Take many fast images (EPI): 5-60 min scan
Task FMRI
Tag Control
incoherent - dephases
z
x
Information Transfer:
- large changes in
axon
cell body
spiking
- chemical neuro-
Information Transfer:
- large changes in
O2 capillary
- chemical neuro-
astrocyte
Pre synapse
7% neurotransmitter glucose
Post synapse
74%
6%
O2 capillary
haemodynamic response
- blood ow - blood volume - blood oxygenation
EEG MEG
BOLD Effect
arterioles capillary bed venules = HbO2 = Hbr
BOLD Effect
Increased Neuronal Activity
arterioles CBF capillary bed venules = HbO2 = Hbr
CBV
BOLD response, % positive BOLD response
3 2 1 0
CMRO2
initial dip
overshoot
stimulus
time
Distortion due to B0 inhomogeneity (air in sinuses) - acquisition-and-analysis related xes needed (eldmap) Physiological noise is more problematic near brainstem - acquire physiological measurements & do analysis x Motion can also be a signicant problem (some analysis xes)
Distortion Signal Loss Physiological Noise ... plus most diffusion artefacts (not eddy currents) and all structural artefacts
Analysis
MR Physics
Simple visual stimulus - ashing chequerboard Different activities include: sensory stimulus, cognitive tasks & motor tasks
Experimental Design
A (rest)
B (activation)
A (rest)
(rest) (activation)
10
20
30
40
50
time (TRs)
- stimulus vs baseline - constant stimulus intensity - constant block lengths - many repetitions: ABABA Need baseline (rest) condition to measure change
volumes are taken during the FMRI experiment Optimised for BOLD sensitivity and speed Take one volume every ~3 seconds Often take around 200 volumes (10 minutes) An FMRI volume is shown here in orthogonal view
The whole time series - a single slice taken from each volume
Used to map activations onto Best way to identify anatomy Better accuracy for registration of
results to standard space
Compare high-resolution voxels with low-resolution Limits how accurate spatial locations can be
Analysis
MR Physics
time
initial dip
overshoot
stimulus
time
Find which voxels have signals that match the model Good match implies activation related to stimulus
measured timeseries at marked voxel
time
predicted response
0 10 20 30
seconds
time
Predicted Response
The process can be modelled by convolving the activity curve with a "haemodynamic response function" or HRF
time
Predicted response
Location 1
BOLD signal
Location 1
Location 1
Fitted Amplitude
Residual Noise
Location 1
BOLD signal
Location 2
Location 2
Fitted Amplitude
- e.g. thin slices, slice angulation, z-shims, parallel imaging, ... Isotropic voxels (or close) are better for analysis generally Do not do upsampling on scanner (sometimes the default) For small FOV also take one single whole-brain EPI Biggest interaction of exp. design-acquisition-analysis so
think carefully about all parts before acquiring data!
Complementary techniques
Complementary techniques
Structural MRI: - CT (bones/membranes/vessels/tumours) - Histology (microstructure) Diffusion MRI: - Tracer studies (individual bres) - Histology (myelin/axon dimensions/glia) Functional MRI: - PET/SPECT (metabolic/ligands/low res.) - EEG/MEG (electrical activity/high temporal res.) - NIRS (haemodynamics/high temporal res.) - TMS/TDCS (alter regional brain function) - Electrodes (single cells/cell groups)
FSL Overview
Running FSL
Main FSL GUI: type fsl and access the other
GUIs through this or type the name starting with a capital letter: e.g. Melodic (or Melodic_gui on Mac) are lowercase: e.g. melodic
exibility and the ability to script (automate) from processing and analysis
Viewing tool (FSLView) completely separate Help via web-docs, usage messages, email list, ...
Image Format
The quick guide for FSL is: DICOM NIFTI Analyze - No - Yes - sooo last century!
Just need to convert DICOM to NIFTI once, after acquisition, then run everything else with NIFTI. Many tools available to do the conversion. For example: dcm2nii from mricron mri_convert from FreeSurfer
FSL Overview
Structural BET: brain extraction FAST: tissue segmentation FIRST: subcortical segmentation FLIRT: linear registration FNIRT: nonlinear registration FUGUE: EPI unwarping SIENA: atrophy analysis FSL-VBM: grey matter density Functional FEAT: model-based FMRI analysis MELODIC: model-free FMRI analysis FLOBS: optimal HRF basis functions FABBER: perfusion analysis Diffusion FDT: diffusion & tractography TBSS: voxelwise DTI analysis Other tools FSLView: display tool Inference (randomise, SMM) Brain atlases POSSUM: FMRI simulator FSLUTILS misc. utilities e.g. fslmaths, fslstats, etc.
FSLView
No processing/analysis tools FMRI timeseries viewing (model vs. data) interaction with several probabilistic atlases 3D rendering
FSLUTILS
fslhd / fslinfo / fslval - show header info fslcreatehd - create header fslmerge - concatenate images in x/y/z/t fslsplit - split 4D image into lots of 3D volumes fslroi - extract region-of-interest from 3D/4D fslmaths - general image calculator fslstats - estimate summary image statistics fslmeants - extract mean/masked timeseries fslreorient2std - reorient image axes
FSLUTILS
Compulsory arguments look like <this> Optional arguments look like [this] - A set of options looks like [this/that/other] where only one
should be used in any command Any double minus args need an equals sign (with no spaces) e.g. fsl_tsplot -i inputle -o outputle --start=2
each cortical hemisphere to a spherical surface Align across subjects on the cortical surface Display activation on inated/attened surface Cortical thickness measurements Multi-subject FMRI stats on standard spherical surface (reduces subject variability) Subcortical segmentation Easy to pass data between FSL & FreeSurfer