Introduction to Diffusion-weighted Imaging Joelle Sarlls, Ph.D. NIH MRI Research Facility National Institute of Neurological Disorder and Stroke National Institutes of Health

Motivation Magnetic resonance imaging provides information about the spatial distribution of water. Diffusion-weighted MRI (DWI) provides information about the motion of water. DWIs are sensitive to cellular architecture and tissue integrity. DWI can provide quantitative measures that are directly comparable. Diffusion imaging can be used to identify specific white matter tracts Over 1000 publications combining fMRI and DWI –120 since I gave this lecture last years ago

Outline What is diffusion? How do we measure diffusion in MRI? How do we extract directional information? What are the practical problems and limitations? Beyond the diffusion tensor

Outline

Diffusion Diffusion refers to the random translational (Brownian) motion of molecules that results from the thermal energy of theses molecules (for sphere) Stokes-Einstein

Gaussian Distribution 1D2D3D Large number of particles that are free to diffuse have a squared displacement of a Gaussian form Einstein, A. Ann Physik (1905) 4:

Diffusion For H 2 O at 37 o C D ≈ 3.0x10 -3 mm 2 /s T dif ≈ 30 ms r ≈ 25 μm If the motion of water is hindered by cell membranes, macromolecules, etc. the displacement will be less and D will appear lower.

Outline What is diffusion? How do we measure diffusion in MRI? How do we extract directional information? What are the practical problems and limitations? Beyond the diffusion tensor

Image Intensity in MRI Physical property of tissue water –ρproton density –T1relaxation time –T2relaxation time –T2*relaxation time –Ddiffusion coefficient Experimentally controlled parameters –Sequence Spin-echo/gradient echo –TRTime of Repetition –TETime to echo –b-valuediffusion-weighting factor Rotational motion, Magnetic field strength } Concentration of water Translational motion

Frequency ω0ω0 0+Z-Z GzGz Gradients make the resonance frequency a function of spatial position ω = γ B = γ B 0 + γ zG z

Basic Diffusion-weighting 90° +x S total -x 0 GxGx

Phase Twist +Z -Z 0

Basic Diffusion-weighting 90° +x S total -x 0 stationary moving GxGx

b-value = 1000 s/mm 2 b-value = 0 s/mm 2 Guess the intensity

Spin-echo Diffusion Preparation 90°180° RF G diff  Δ G Stejskal, EO and Tanner, JE. J Chem Phys (1965) 42 : echo

DWI Non-diffusion-weighted signal intensity B-value sec/mm 2 Diffusion Coefficient mm 2 /sec

Typical DWI Single-shot “spin-echo” Echo Planar Imaging ParameterValueComment TE50-100msLimited by b-value TR>5sFully relaxed Matrix96 x x 2.5 mm Slice Thickness2.5 mmEqual dimensions B-value~1000 s/mm 2 For brain* *Jones D., et al. Mag Res Med (1999) 42 : 515

b (s/mm 2 ) Diffusion map S = S 0 e -bD Calculate Diffusion Parameters

b = 0 s/mm 2 I 0 b = 1100 s/mm 2 Gz Dz

8 μm EM of mouse corpus callosum Water Diffusion in Tissue D perp << D par D perpendicular D parallel X Not Free Cell membranes Myelin Organelles Extracellular matrix Anisotropy

ADC b = 0 s/mm 2 Gz Gy Gx

b = 0 s/mm s/mm 2 Diffusion Map 0.75 x mm 2 /s0.43 x mm 2 /s Acute Stroke Warach S., et al. Ann Neurol (1995) 37:

Outline What is diffusion? How do we measure diffusion in MRI? How do we extract directional information? What are the practical problems and limitations? Beyond the diffusion tensor

8 μm EM of mouse corpus callosum Water Diffusion in Tissue D perp << D par D perpendicular D parallel X Not Free Cell membranes Myelin Organelles Extracellular matrix Anisotropy

Anisotropic Diffusion

The Diffusion Tensor x y z Basser, P, et. al. J Magn Reson B (1994) 3 :

b = 0 s/mm 2 GxGyGz GxyGxz Gyz DTI b = 1100 s/mm 2

Calculate Diffusion Tensor

Diagonalize DT x y z ε3ε3 ε2ε2 ε1ε1 EigenvaluesEigenvectors

Quantitative Parameters Fractional Anisotropy 0 ≤ FA ≤ 1 Average Diffusivity x y z ε3ε3 ε2ε2 ε1ε1 <D><D>

isotropicanisotropic

FA

Directional Encoding for DTI z x y ε3ε3 ε2ε2 ε1ε1 Z X Y Pajevic S. and Pierpaoli C., Magn Reson Med (1999) 43 :

Directional Encoded Color Map

FA DECNo sym DEC LinearLine FieldPlanarSpherical

Applications of DTI Cerebral Ischemia (Stroke) Brain Cancer and Effects of Radiotherapy Multiple Sclerosis Epilepsy Metabolic Disorders Normal Brain Maturation and Aging Traumatic Brain Injury Alzheimer’s Disease Amyotrophic Lateral Sclerosis Niemann-Pick type C Disease Dementias Connectivity

Pediatric DIPG T2-weightedMDFADEC

Outline What is diffusion? How do we measure diffusion in MRI? How do we extract directional information? What are the practical problems and limitations? Beyond the diffusion tensor

Typical DW SSEPI Low Resolution Distortions - Field inhomogeneities Distortions - Diffusion weighting Insensitive to Bulk motion Time Efficient PROCON

CON: Distortions from field inhomogeneities Non-diffusion-weighted SSEPI T2-weighted FSE SSEPI corrected

CON: Distortions from DW FA maps DW SSEPI volumes

Other Common Problems in DTI Low SNR Incomplete Fat Suppression Bulk movement Cardiac pulsation

Low SNR 1.7mm iso1.3mm iso2.5 mm iso

Low SNR 1.7mm iso mm 3 1.3mm iso mm mm iso mm 3

Low SNR 1.7mm iso mm 3 1.3mm iso mm mm iso mm 3

Incomplete Fat Suppression b=1100 s/mm 2 MDFA

Cardiac Pulsation Diffusion weighting in Z

Bulk Movement

Outline What is diffusion? How do we measure diffusion in MRI? How do we extract directional information? What are the practical problems and limitations? Beyond the diffusion tensor

What isTractography? The use of orientation information from diffusion imaging to reconstruct estimates of white matter pathways in the brain.

Limitation to DTI comes from partial volume effects Typical resolution for SSEPI DTI 2.5 x 2.5 x 2.5 mm Cortical projection systems of left cerebral hemisphere

Partial Volume Effect DT ellipsoid distribution

Sub-millimeter DTI

Beyond Standard DTI High Angular Resolution Diffusion Imaging (HARDI) –Multi-tensor models –Non-parametric algorithums DSI, Qball, SD, PAS

Non-parametric Algorithms fODFdistributionfODFdistribution

b = 0 s/mm 2

ACKNOWLEDGEMENTS Peter Bandettini, PhD Carlo Pierpaoli, MD,PhD Ted Trouard, PhD Lindsay Walker, MS Kathy Warren, MD Emilie Steffen Dan Handwerker, PhD THANK YOU

Diffusion Profile IsotropicAnisotropic

Inherent Motion in Living Systems ei1ei1 ei2ei2 einein kyky kxkx Diffusion-Weighted MRI S n (t) = F[ k x (t), k y n ] e i  n S n (t) = F[ k x (t)+  k x n, k y n +  k y n ] e i  n

kyky kxkx PRO: insensitive to bulk motion S(t) = F[ k x (t)+  k x, k y +  k y ] e i 

DWI Non-diffusion-weighted signal intensity B-value sec/mm 2 Diffusion Coefficient mm 2 /sec Take two measures of signal and solve for D.

Average (Trace) Image

Calculated the ADC I ave b-value = 0 ADC map (mm 2 /sec)

Trace b = 0 s/mm 2 FA DEC Gx Gy GzGxy Gxz Gyz λ 1 +λ 2 +λ 3

Not all processing software is created equal! TORTOISECompetitor

Diffusion-Weighted MRI (DWI) Sensitizes MRI image intensity to small, thermally induced random motion of water molecules The motion of water within tissue is extremely sensitive to the microscopic architecture and integrity of the tissue For H 2 O at 37 o C D ≈ 3.0x10 -3 mm 2 /s T dif ≈ 30 ms r ≈ 25 μm Einstein, A. Ann Physik (1905) 4:

Identical anisotropy maintained throughout the entire voxel Anisotropy a small fraction of the voxel. Not experimentally observed Complete anisotropy, but variable orientation. Experimentally isotropic. Limitation to DTI : Spatial resolution