Visualization of Diffusion Tensor Imaging

1
Visualization of Diffusion Tensor Imaging

• Guus Berenschot
• May 2003

Supervisor Bart ter Haar Romeny Daily
Supervisor Anna Vilanova i Bartroli Other
committee members Carola van Pul, Klaas Nicolay
and Peter Hilbers
2
Contents

• Introduction Diffusion Tensor Imaging (DTI)
• Visualization Tool for DTI
• Demonstration
• Conclusion and Future Work

3
Introduction Diffusion Tensor Imaging

• Diffusion is the random motion of molecules, and
  is characterized by a diffusion coefficient D.
• In tissue this diffusion hindered by physical
  barriers.
• The diffusion coefficient is called Apparent
  Diffusion Coefficient (ADC)

4
Introduction Diffusion Tensor Imaging

• Diffusion Tensor Imaging is a Magnetic Resonance
  Imaging (MRI) technique.
• DTI measures the ADC in 6 directions and computes
  a symmetric diffusion tensor (D) of this
• This diffusion tensor is defined for each voxel
  in the 3D dataset

5
Diagonalization Diffusion Tensor

• Diagonalization of this tensor provides three
  eigenvectors (ev1, ev2 and ev3) with three
  corresponding eigenvalues (?1, ?2 and ?3)

6
Anisotropy Indices

• Linear case
• Planar case
• Isotropic case

7
Problem Definition

• How to extract meaningful information of a 3D DTI
  dataset???
• Neonatal brain (Maxima Medical Center, Veldhoven)
• Muscles (Magnetic Resonance Laboratory)

8
Visualization Tool For DTI

• Anatomical reference
• Displaying local tensor information
• Displaying global tensor information
• Improvements to existing techniques

9
Multi Planar Reconstruction Planes

• Anatomical reference
• Displaying local tensor information in 2D slice

10
Anisotropy Indices

• Colorcoding anisotropy indices

11
Colorcoding Main Diffusion Direction

• Colorcoding directions
• Intensity color scaled with anisotropy index

Pajevic et al. 1997
12
Glyphing

• Glyphs are icons that represent the local tensor
  information
• Two types of glyphs can be displayed
• Ellipsoids
• Cuboids (Worth et al., 1998)

13
Cuboids
14
Fiber Tracking Introduction

• Fiber tracking simplifies the tensor field to the
  vector field of the main eigenvector
• This vector field is made continuous by
  interpolation
• Consider this vector field as a velocity field
  and drop a free particle on it
• This particle will follow a trajectory
• The found trajectory can be seen as a bundle of
  fibers

Xue et al. 1999, Conturo et al. 1999, Mori et
al. 1999
15
Tracking

• The tracking can be seen as solving the following
  integral
• To solve this integral we use a second order
  Runge Kutta integration

16
Seed Points

• Manual definition of a seed point or seed region
• Start tracking in all voxels and keep the
  trajectories that pass a certain region

Stopping Criteria

• Linear Anisotropy (Cl) falls below a certain
  threshold
• Angle in a fiber is too big

17
Fiber Tracking In Healthy Volunteer
Optical tract Corpus Callosum
18

• Patient with a tumor

Neonatal brain
Mouse muscle
19
Surface Building

• Fiber tracking gives problems in regions with
  planar anisotropy the main eigenvector is not
  reliable
• Planar anisotropy can be due to kissing crossing
  or branching fibers

20
Surface Building

• If we enter a region with planar anisotropy
  follow all directions defined by local plane and
  display a surface here
• If anisotropy is linear again do the common
  fiber tracking.

21
Demonstration
22
Conclusion

• The visualization tool is considered as very
  useful by the MMC and the MRL
• Results of fiber tracking in neonatal brain is
  promising

Future Work

• Seeding is biased
• Noisy data-gt smoothing
• Quantitative information of fibers

23
Thanks

• Anna Vilanova i Bartroli (daily supervisor)
• Bart ter Haar Romeny (supervisor)
• Gustav Strijkers and Anneriet Heemskerk (MRL)
• Carola van Pul and Maurice Jansen (MMC)
• George Roos and Jan Buijs (radiologist and
  neonatologist MMC)
• Klaas Nicolay and Peter Hilbers (committee
  members)