Final Presentation - MNTP

paraderollΤεχνίτη Νοημοσύνη και Ρομποτική

17 Νοε 2013 (πριν από 4 χρόνια και 1 μήνα)

75 εμφανίσεις

Diffusion


Tensor



Imaging

Tim Hughes & Emilie
Muelly

1

DTI Module


Learning objectives


Acquisition


Fiber orientation distribution function (ODF)


Tractography


Projects


Combining
fMRI

+ DTI to explore face recognition
& working memory


Comparing and contrasting DTI parameters


2

Diffusion Tensor Imaging


DTI acquisition:


Non
-
diffusion weighted images


Diffusion weighted images (DWI)



Magnitude of diffusion weighting

(e.g. b=1200 or 2400)



b
-
value :
angular resolution


signal:noise



Output measures


Apparent Diffusion Co
-
efficient, Mean Diffusivity


Fractional Anisotropy (FA)

3

Acquired b0 image


Acquired b0 (b=0 s/mm
2
): a reference for
DTI analysis


Problematic with partial volumes


Neuronal tissue


Free water (
cerebrospinl

fluid, extracellular
fluid, and edema)


Effect on ADC, FA value, and fiber tracking


Partially fixed by FLAIR,


Incomplete saturation (mainly corrects for CSF)


Increased scan time

4

Synthetic b0


Developed as a result of last year’s MNTP (Jung
et al
)


Uses max signal intensity (from any direction) at each voxel
to create synthetic b0 image


Designed to minimize free water effect


No impact on scan time

R

=
max(DW images)

(Image contrast enhanced using gamma
corection
: gamma=0.5)


Tensor model


Single orientation at voxel (single ODF)


6+ directions with 1 b0


No information regarding fiber crossing



Constrained Spherical
Deconvolution

(CSD)


HARDI (high angular resolution diffusion imaging)


23+ DW directions with multiple b0


Informative crossing


Fiber ODF Analysis Methods

Tournier

et al., 2007

6

Methods:
Acquisition & Pre
-
prossessing


4 subjects


Acquire diffusion weighted images


Siemens 3T MRI; TR = 6900ms, TE = 115ms


50 directions, 5 b0 values (across time)


b
-
values = 1200 s/mm
2

or 2400 s/mm
2



2 acquisitions per subject, per b
-
value


Pre
-
process the data:


Motion correction (rotation of vector table)


Create Synthetic b0




7

Methods:
ODF and
Tractography


ExploreDTI v4.8.0 (A.
Leemans
)


ODF analysis (Tensor or CSD)


Identified tracts using regions of interest








Obtained tract
-
based statistics (mean FA value,
standard deviation, number of “fibers”)

8

Fornix

Cingulum

Uncinate

Fasciculus (UF)

Inferior
fronto
-
occipital

fasciculus (IFOF)

Methods:
Fiber analysis


Parameters


Diffusion weighting:



b0 images:




ODF method:





SAS v9.2


GLM, compare effects of each parameter on outcomes




Evaluated effects of all first order interactions on
outcomes


9



Number

of fibers



䵥慮⁆䄠v慬ae

戱㈰b

vs.

b2400

Acquired

vs.

Synthetic

DTI

vs.

CSD

Raw Data (UF)

Synthetic b0

Acquired b0

Tensor

b2400

CSD

b2400

Right_UF

tensor

b2400

acq

b0

b2400

synthb0

b1200

acq

b0

b1200

synth

b0

FA value

0.45

0.70

0.49

0.73

Std Dev

0.12

0.10

0.13

0.10

#

fibers


3465

3530

5017

5023

10

Right_UF

CSD

b2400

acq

b0

b2400

synth

b0

b1200

acq

b0

b1200
synth

b0

FA value

0.34

0.60

0.43

0.66

Std Dev

0.14

0.15

0.13

0.14

#

fibers


6345

4162

5519

5455

Effect of DTI parameters on

number of fibers



p
-
value



Subject

0.0922

5%

Tracts
*

<0.0001

70%

b1200
vs. b2400

0.7702

0%

Acquired b0 vs. Synthetic b0
*

0.7357

0%

Tensor

vs. CSD

0.0041

6%

11

*

Tract
-
based analysis indicates that synthetic b0 significantly increases the
number of fibers in the
fornix

only.

Effect of DTI parameters on

mean FA
value



p
-
value



Subject

0.9077

0%

Tract
*

<0.0001

17%

b1200
vs. b2400

0.1239

2%

Acquired b0 vs. Synthetic b0
*

<0.0001

67%

Tensor

vs. CSD

0.0013

8%

Number
of fibers



0.0046

6%

12

(positive correlation)

*
Significant interaction between b0 method and tract on mean FA value

Mean FA value

13

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Fornix
UF
Cingulum
IFOF
mean

FA

SD

acquired

0.49

0.09

synthetic

0.65

0.06

Effect of Synthetic b0 on FA Value


Differences by Tract

Conclusions


Changing DTI parameters can significantly alter
the number of fibers and FA values


Diffusion weighting


No significant differences in
b1200

and
b2400


b0 images


Synthetic

b0 FA compared to
acquired

b0


Effects of both FA and # fibers are most dramatic in
the fornix


ODF methods


CSD

method # fibers, mean FA values compared to
tensor based
method

14

Acknowledgments


MNTP program


Seong
-
Gi

Kim


Bill Eddy


Tomika

Cohen



DTI module

Mentor


Kwan
-
Jin Jung

TA


Xiaohan

Huang

15