White Matter Tractography Using Random Vector RAVE Perturbation

1
White Matter Tractography Using Random Vector
(RAVE) Perturbation

• Mariana Lazar
• and Andrew L. Alexander
• Departments of Physics and Computer Science,
• University of Utah
• Departments of Medical Physics, University of
  Wisconsin

2

• Supported by
• NIMH R0162015

3
Overview

• Goal To develop an algorithm that accounts
  for the probabilistic nature and measurement
  uncertainty of the diffusion tensor
• Possible application
• Designing structural connectivity
  measures between different neural
  centers in the brain

4
Outline

• Diffusion Tensor and White Matter Tractography
• RAVE algorithm - algorithm description
  - human brain fiber tracking using RAVE
• Discussion

5
White Matter Tractography

• Goal Reconstruct the fiber connections between
  different brain regions using the directional
  information provided by diffusion tensor imaging
• Common approach- use e1 to estimate fiber
  directions- works well in regions with highly
  linear anisotropy

6
White Matter Tractography

• Limitations -ambiguous e1 - poor
  SNR, encoding - non-prolate tensor
  shape (e.g. laminar) - partial
  voluming ?1 insufficient to describe
  multiple fiber directions
  within a voxel

7
RAVE perturbation algorithm

• Probabilistic nature of diffusion tensor should
  be considered in regions of poor e1 specificity
  
• New approach RAVE (Random Vector) Perturbation
• -from a single seed multiple pathways are
  generated by calculating a perturbed eigenvector
  direction at discrete points along the trajectory

8
RAVE perturbation algorithm
Reference frame
Measurement frame
9
RAVE perturbation algorithm
x
z
y
?? - degree of perturbation
10
RAVE perturbation algorithm

• Diagonalize the tensor - rotate
  tensor to the reference frame
• Perturb major eigenvector
• Rotate back to the measurement frame

11
Fiber Tracking

• Pre-assigned seed regions
• Project both forward and backward perturbing the
  major eigenvector at each step
• Trajectories terminated for FA lt threshold (e.g.,
  FA 0.15 - 0.2) or if angle between two
  consecutive steps is greater then 40 degrees.

12
Fiber Tracking

• For each seeding point a family of 500 tracts
  were generated
• The number of times an image voxel was
  intersected by trajectories was counted resulting
  in a volumetric density of the fiber pathways
• The results were displayed using volume rendering

13
Human Brain Fiber Tracking

• DW - EPI / b 1000 sec/ mm3
• Subject 1 - 260x260x110 mm3 field-of-view
  - 1 mm3 isotropic voxelsSubject 2, 3 -
  1.96x1.96x3 mm3 voxel interpolated to
  0.98 mm isotropic voxels

14
Superior longitudinal fasciculus seed
Subject 1
GORDON KINDLMANN
? 0.2
15
Superior longitudinal fasciculus seed
? 0.2
? 0.4
16
Superior longitudinal fasciculus seed
? 0.6
? 0.8
17
Cortical white matter seed
? 0.2
Subject 1
18
Cortical white matter seed
19
Internal capsule seed
? 0.2
Subject 1
20
Internal capsule seed
21
Tumor patient Cortico-spinal Tract and Fornix
? 0.2
Subject 2
22
Subject 2
23
Full tract reconstruction Superior longitudinal
fasciculus
? 0.2
Subject 3
24
Discussion

• We demonstrated a probabilistic tractography
  approach that can describe multiple possible
  pathways from a single point and their likelihood
• The method has the potential to reveal fiber
  pathways other than the ones obtained using basic
  streamlines techniques

25
Discussion

• The technique indicates possible pathways - some
  of branching might be erroneous- the results
  should be interpreted with caution - correlate
  to additional information (FMRI or anatomy)

26

• Acknowledgements
• UU-Gordon Kindlmann
• UW-Madison Aaron Field, Victor Haughton, Howard
  Rowley, Benham Badie, Konstantinos Arfanakis
• NIMH 62015