Introduction to Diffusion Tensor Imaging

November 2010

Jennifer Campbell

Introduction to Diffusion Tensor Imaging

• Molecular diffusion• Diffusion weighting in MRI• The diffusion tensor in use• Problems and solutions

Introduction to Diffusion Tensor Imaging

• Molecular diffusion• Diffusion weighting in MRI• The diffusion tensor in use• Problems and solutions

Diffusion MRI measures Brownian motion of water molecules

Path of diffusing water molecule

Water displacement distribution

Diffusion MRI measures Brownian motion of water molecules

−−•

−

=d

T

d

drrDrr

DrrP

τπττ

4)()(exp

)4(1),|( 0

10

30 dDr τ6; 2 >=<

Diffusion MRI measures Brownian motion of water molecules

Restricted diffusion: Apparent Diffusion Coefficient (ADC)

Diffusion MRI measures Brownian motion of water molecules

Restricted diffusion: Apparent Diffusion Coefficient (ADC)

Diffusion MRI measures Brownian motion of water molecules

−−•

⋅

−

=d

T

d

drrrrrrP

τπττ

4)((ADC))(exp

)4(ADC1),|( 0

10

30

Tissue structures determine which directions of motion are most probable

White matter fibre bundle: oriented

structure

Water molecules prefer to travel parallel to fibre

direction

Tissue structures determine which directions of motion are most probable

−−•

−

=d

T

d

drrrrrrP

τπττ

4)()(exp

)4(||1),|( 0

10

30

DD

The diffusion tensor

λ1

λ3λ2

e1

Introduction to Diffusion Tensor Imaging

• Molecular diffusion• Diffusion weighting in MRI• The diffusion tensor in use• Problems and solutions

Diffusion weighted MR sequenceTE

G

90° 180° echo

G

Stejskal-Tanner diffusion weighted sequence: pulsed-gradient spin-echo (PGSE)

)3/(222 δδγ −∆= Gb

∆δ

bDoeSbS −=)(

bDoeSbS −=)(

b

)(bS

Diffusion weighted MR sequence

B0B0 + G B0

G

180° echo

B0, 180o

G

B0 + G

Gradients, phase, and motion

G

180° echo

G

Gradients, phase, and motion

B0 + G B0, 180o B0 + G B0B0

Diffusion weighted images (DWIs)

• series of DWIs: 3-6-60-100+ encoding directions• Full brain coverage: 15-20 minutes per 100 directions • b value: 1000-3000 s/mm2

• Voxel size: roughly 2mm x 2mm x 2mm isotropic voxels typically used

Acquisition parameters for diffusion weighted images (DWIs)

Sequence modifications: trace weighted sequence

GGx

RF

90° 180°

G

GyG G

GzG G

Computing the diffusion tensor

Computing the diffusion tensor

λ1

λ3λ2

e1

Computing the diffusion tensor

FA: anisotropy

index

trace of diffusion tensor

RGB plot: principal

eigenvector (e1) direction, scaled

by FA

Maps obtainable from diffusion tensor

)()()()(

23FA 2

32

31

23

22

21

2 λλλλλλλλλ

++−+−+−

=

Diffusion tensor imaging (DTI)

λ1

λ2∼λ3

cylindrical symmetry

Diffusion tensor imaging (DTI)

λ⊥, radial

λ||, axial

cylindrical symmetry

Introduction to Diffusion Tensor Imaging

• Molecular diffusion• Diffusion weighting in MRI• The diffusion tensor in use• Problems and solutions

Diffusion tensor imaging (DTI)

normal white matter

Diffusion tensor imaging (DTI)

myelin degeneration:•radial diffusivity ↑•FA ↓

Diffusion tensor imaging (DTI)

normal white matter

Diffusion tensor imaging (DTI)

axonal fragmentation:•axial diffusivity ↓•FA ↓

Diffusion tensor imaging (DTI)

normal white matter

Diffusion tensor imaging (DTI)

decreased axon density:•radial diffusivity ↑•FA ↓

Diffusion tensor imaging (DTI)

crossing fibres:•FA lower than for single direction

Diffusion tensor imaging (DTI)

degeneration of one fibre population in crossing case:•FA ↑

Diffusion tensor imaging (DTI)

curvature:•FA lower than for single direction

Diffusion tensor imaging (DTI)

splay:•FA lower than for single direction

normal tissue

Diffusion tensor imaging (DTI)

cellular swelling:•MD ↓

Diffusion tensor imaging (DTI)

Fibre tracking using diffusion MRI data

Diffusion MRI data

Water displacement profile

Fibre directions

Fibre tracts

Tractography

What does tractography tell us?-is A connected to B?-how are A and B connected?-how confident are we in this

connection?

Applications of tractography:-visualization in 3D: localization,

education-segmentation (both white matter and

grey matter)-investigating clinically relevant

differences between pathways in different populations-neuroanatomical questions

Tractography

Streamline integration methods:• “fibre assignment using continuous tracking”

(FACT)• Euler• RK4

Tract delineation techniques:• tract delineating ROIs (as many as desired): drawn

manually, by protocol, or based on other data• exclusion masks, stopping criteria S. Mori et al, AN 45:265-269, 1999

Tractography: warnings

Diffusion MRI tractography results include many false positives and false negatives: priors are essential.– Tractography is good for segmentation of parts of

pathways, but segmenting the entire pathway can be challenging.

– Characterizing unknown new anatomy is much more difficult than studying known anatomy.

– Tractography does not distinguish between afferent and efferent pathways, or between mono- and multisynaptic connections.

Tractography incorporating uncertainty

MamayyezSiahkal et al. 2009

Applications• basic human neuroanatomy• structure and function• understanding disease• understanding therapies• nonhuman primates; other animal

studies

• stroke• MS• cancer• trauma• dyslexia• epilepsy• schizophrenia

Applications

• ALS• dementia• CJD• surgical planning• surgical outcome• blindsight• depression

• neuroanatomy• development• aging

Introduction to Diffusion Tensor Imaging

• Molecular diffusion• Diffusion weighting in MRI• The diffusion tensor in use• Problems and solutions

Diffusion weighted signal intensity: high value in directions perpendicular to fibres

Diffusion weighted signal profile

Probability of water displacement orientation distribution function (ODF)

Diffusion tensor model

High Angular Resolution Diffusion Imaging (HARDI)

q-ball imaging

QBI DTI

High Angular Resolution Diffusion Imaging (HARDI)

Why use high angular resolution diffusion MRI?• Reduce false positive and false negative tractography results due to

crossings• In voxels where fibres cross, single fibre approaches can yield:

– ambiguous fibre direction– incorrect fibre direction– only one fibre direction

(that of fascicle with largest volume fraction) • Examples of such regions:

– basis pontis, subcortical white matter, superior longitudinal fasciculus, acoustic radiations, projection from subgenual white matter to amygdala, optic chiasm, caudate nucleus, corpus callosum, cortical spinal tract, cingulate bundle.

Crossing fibre reconstruction approaches

• multi-tensor approaches (Alexander et al., Parker et al., others)

• multi ball and stick (Behrens et al.)

• Composite hindered and restricted model of diffusion (CHARMED) (Assaf et al.)

• diffusion spectrum imaging (DSI) (Wedeen et al.)

• q-ball imaging (QBI) (Tuch et al.)

• spherical deconvolution (Tournier et al., Anderson, others)

• other variants

DTI vs. HARDI :Tractography in phantom

single fibre multi-fibreCampbell et al. ISBI 2006.

Tractography: false negatives

Lateral connections?

Signal to noise ratio (SNR); spatial resolution

(b=3000 s/mm2)(b=1000 s/mm2)

Jumping from tract system to tract system

Effects of limited spatial resolution

Jumping from tract system to tract system

Effects of limited spatial resolution

Heidemann et al. 2010.

Signal to noise ratio (SNR); spatial resolution

Higher field strength (7T)

• diffusion MRI acquisitions generally use fast imaging techniques which can create artifacts such as magnetic susceptibility-induced distortion

• these artifacts confound registration with standard structural scans

Magnetic susceptibility induced distortion

Magnetic susceptibility induced distortion

Heidemann et al. 2010.

segmented readout

Eddy-current induced distortion

• eddy currents are induced by the rapidly changing magnetic field gradients required for diffusion sensitization

• these eddy currents induce magnetic fields which distort the image.

artifacts: shifting, sheering, stretching, compression, signal loss

TE

G G

90° 180° echo

G

180°

G

Twice refocused balanced echo (TRBE)

Eddy-current induced distortion

Suggested review articles• Susumu Mori et al. Principles of diffusion tensor imaging and its

application to basic neuroscience research. Neuron 51, 527-539, 2006.

• Basser et al. Diffusion-tensor MRI: theory, experimental design and data analysis – a technical review. NMR in Biomedicine 15, 456-467, 2002.

• Mori et al. Fiber tracking: principles and strategies - a technical review. NMR in Biomedicine 15, 468-480, 2002.

• Johansen-Berg et al. Just pretty pictures? What diffusion tractography can add in clinical neuroscience. Curr Opin Neurol 19, 379-385, 2006.

Software packages

• mincdiffusion (BIC)• mincDTI (Andrew Janke - BIC)• FSL (FMRIB, Oxford)• DTI Studio (Johns Hopkins)• Camino (Manchester/London)• CINCH (Stanford)• Mrtrix (Brain Research Institute, Melbourne)• more …

Acknowledgements

G. Bruce PikeKaleem SiddiqiIlana LeppertPeter SavadjievParya Mamayyez Siahkal Steven FreyIves LevesqueLuis Concha

Nikola StikovClarisse MarkEva Alonso OrtizHalleh GhaderiMike FerreiraSilvain BeriaultChristine Tardif