Transcript DTI Module

David Schaeffer
Lauren Libero
(University of Georgia)
(University of Alabama at Birmingham)
Sara Levens, Ph.D.
(University of Pittsburgh)
Instructor
Kwan-Jin Jung, Ph.D.
Technical Assistant
Nidhi Kohli
(Carnegie Mellon University)
(Carnegie Mellon University)
LEARNING OBJECTIVES

Effects of
 Segmented
sampling
 Motion correction
 Fiber orientation estimation method
 fMRI based ROIs vs. drawing ROIs

Anatomical separation of sensorimotor cortex
TERMINOLOGY

Diffusion encoding gradient direction
Vector table (x, y, z components)
 Angular resolution


Diffusion-weighting (b-values)
Duration & amplitude
 s/mm²


b0 = 0 s/mm²

No diffusion gradient
METHOD OF ACQUISITION

Segmented sampling
 Complementary
diffusion encoding directions
 64
(A) - 10 min
 64 (B) - 10 min
 128 (A + B) - 20 min
 Useful
for special
populations
MOTION CORRECTION

How to correct:
1. Estimate the motion
2. Rotate image and vector table accordingly
Intended
Collected
Head correction
WRONG
Head & vector table
correction
CORRECT
MOTION CORRECTION

No correction
 No

vector rotation
Interpolation
how much you rotate vector table
 Based on distributed b0 images – “real motion”
6
Rotation (degrees)
Rotation (degrees)
 Estimates
3
0
-3
-6
Time
6
3
0
-3
-6
Time
BEFORE
AFTER
MOTION CORRECTION

Simulation method
 Collect
1.
two diffusion scans
6 direction scan (low b-value)


2.
Why? – Fast (little time for motion)
Edges of brain are clearly defined
6 or more direction scan (higher b-value)
 Assume
no motion on scan 1, then simulate what
higher b-value volume should look like
Low b-value
(b=800 s/mm²)
DWI
(scan 1)
Find D (diffusion
tensor)
S=S0e-bD
Assume no
motion
Find S
(simulated high
b-value)
High b-value
(b=2000 s/mm²)
DWI
(scan 2)
S=S0e-bD
Co-register
volumes
(estimating motion)
Rotate vector table
FIBER ORIENTATION ESTIMATION METHOD
Fiber/voxel
Data Acquisition
Analysis
Single fiber
6 – 12 directions
Tensor
Multiple fibers
> 25 directions (HARDI)
CSD (Q-ball, multi-tensor)
FIBER ORIENTATION ESTIMATION METHOD

Tensor
 Performs
well for straight tracts (like motor)
 Performs poorly for crossing and branching fibers
(like Genu)

Constrained Spherical Deconvolution (CSD)
 Better
for detecting branching and crossing fibers
(Tournier et al., 2007)
CSD VS. TENSOR
Average Number of Tracts in
Genu
120000
100000
80000
N Fibers
Genu
CSD
60000
40000
20000
Genu
Tensor
0
CSD
Tensor
DRAWING ROIS
Manually draw ROIs
 Using fMRI

Collect fMRI data – find center of activation (x, y, z)
 Matrix transformation

 Convert
from fMRI coordinates into DWI native space
SEGMENTING SENSORIMOTOR
Finger closing fMRI results as ROI
 Separation of sensory and motor areas

 Clustering
– fiber end-point distribution
Central Sulcus
SUMMARY

Sampling schemes can be
advantageously altered for use with
special populations

Simulation is a promising method for
more accurate motion correction

CSD Fiber tracking is most appropriate
for resolving fiber crossings
SUMMARY

fMRI-based ROIs can be used to
track fibers from areas of
activation

DTI can be used as a tool to
segment brain areas that are
not separable based on diffuse
fMRI activation maps
ACKNOWLEDGMENTS
Dr. Kwan-Jin Jung
 Nidhi Kohli
 MNTP Leaders: Dr. Eddy & Dr. Kim
 MTNP Trainees & Participants
 DTI Trainees 2009 & 2010
 Funding:


NIH grants:
R90DA023420 and T90DA022761
DISTRIBUTED B0
SCANNING PARAMETERS
MOTION CORRECTION
No correction
Interpolation
Motion
Simulation