Diffusion MRI Analysis

Sonia Pujol, Ph.D.

Surgical Planning Laboratory

Harvard University
Brain Anatomy

•White matter ~45% of the

brain
•Myelinated nerve fibers
(~ 10 μm axon diameter)
White Matter Exploration

Jules Joseph Dejerine

(Anatomie des centres
nerveux (Paris, 1890-1901):
Atlas of Neuroanatomy based
on myelin stained preparation
Diffusion Tensor Imaging (DTI)
• First non-invasive
window on white
matter anatomy
• Measurement of the
motion of water
molecules using MRI
techniques.
• Three-dimensional
reconstruction of the
trajectory of white
matter bundles
Diffusion Weighted Imaging (DWI)

S0 S0 S1 S2 S3

S4 S5 S6 S7 S8

S9 S10 S11 S12

In this example, the DWI scan was acquired with 12 diffusion sensitizing gradient
directions (S1-S12) and 2 non-diffusion sensitizing gradients (S0)
 From DWI to DTI
DWI DTI
S0 S0 S1 S2 S3

S4 S5 S6 S7 S8

S9 S10 S11 S12

DWI dataset acquired with DTI dataset

12 gradient and 2 baseline

- bgˆi T D gˆ i
Si = S 0 e
Si: DWI volume acquired with
Stejskal-Tanner (1965) ith gradient
So: Baseline volume
Diffusion Tensor Imaging
- bgˆi T D gˆ i
Si = S 0 e
Diffusion Tensor Imaging
- bgˆi T D gˆ i
Si = S 0 e

D xx D xy D xz
D=
D yx D yy D yz
D zx D zy D zz
Diffusion Tensor Imaging
- bgˆi T D gˆ i
Si = S 0 e

D xx D xy D xz
D=
D yx D yy D yz
D zx D zy D zz
 Tractography

DTI tractography provides 3D reconstruction of the

trajectory of white matter pathways
Tutorial Outline
This tutorial is an
introduction to the
fundamentals of
Diffusion MRI analysis,
from the estimation of
diffusion tensors to the
interactive 3D
visualization of fiber
tracts.
 Tutorial Dataset
The tutorial dataset DiffusionMRI_tutorialData is a
Diffusion Weighted MR scan of the brain acquired
with 41 gradient directions and one baseline.

Diffusion Diffusion
Sensitizing Weighted
Gradients Images

Download the dataset at:

https://www.slicer.org/w/images/e/e6/Dti_tutorial_data.zip
 3D Slicer

The tutorial uses the 3D Slicer (Version 4.8.1,

revision 26813, Stable Release) software
available at:
http://download.slicer.org

Disclaimer
It is the responsibility of the user of 3DSlicer to comply with both the terms of the license
and with the applicable laws, regulations and rules. Slicer is a tool for research, and is not
FDA approved.
 SlicerDMRI

An open-source project to improve and

extend diffusion magnetic resonance
imaging software in 3D Slicer:
http://dmri.slicer.org

Disclaimer
It is the responsibility of the user of 3DSlicer to comply with both the terms of the license
and with the applicable laws, regulations and rules. Slicer is a tool for research, and is not
FDA approved.
Install SlicerDMRI

First, start Slicer4

Install SlicerDMRI
Click on the Modules menu

Click on the Diffusion

menu; then, click on Install
Slicer Diffusion Tools
Install SlicerDMRI

Click on the Open Extension

Manager menu
 Install SlicerDMRI

Search for SlicerDMRI in

Extension Manager

Click on SlicerDMRI to
install
Install SlicerDMRI

Click Yes to install UKFTractography

Install SlicerDMRI

Restart Slicer to finish installation

Install SlicerDMRI

Related modules appear in the

Diffusion menu
 Learning Objectives
Following this tutorial, you’ll be able to
1) Estimate a tensor volume from a set of
Diffusion Weighted Images
2) Understand the shape and size of the
diffusion ellipsoid
3) Reconstruct DTI tracts from a pre-defined
region of interest
4) Interactively visualize DTI tracts seeded
from a fiducial
MR Diffusion Analysis Pipeline

DWI Tensor Scalar 3D

Acquisition Calculation Maps Visualization
Part 1:
From DWI images
to Tensors
 Understanding the DWI Dataset

….

The Diffusion Weighted Imaging (DWI) dataset is composed of

41 volumes acquired with 41 different diffusion-sensitizing
gradient directions, and one baseline image acquired without
diffusion weighting.
Loading the DWI Dataset
In your files archive, locate the
file dwi.nrrd in the dataset
folder for this tutorial

Drag and drop the file dwi.nrrd

onto the viewer of the Slicer
application
 Loading the DWI Dataset

Exit the archive folders window,

and click OK to load the
dataset to Slicer
Loading the DWI Dataset

Slicer displays DWI

volume of the brain
Loading the DWI Dataset

Click on the Modules

menu and select the
module Volumes
Loading the DWI Dataset

The baseline image

corresponds to the DWI
Component #0.
Select the DWI
Component #10, which
corresponds to the 10th
diffusion sensitizing
gradient
Loading the DWI Dataset

For image contrast adjustment, you may use

the Manual W/L slider.
Loading the DWI Dataset

Position your mouse over

the pin icon, then click on
the link icon and the fit
image to window icon
 Loading the DWI Dataset

Click on the Slicer layout menu

and select the Red slice only
layout
 Loading the DWI Dataset

Slicer displays only

the Axial anatomical
slice in the Viewer
Creating a brain mask

Click on the Modules menu, then

select Diffusion -> Process ->
Diffusion Brain Masking
Creating a brain mask

-select the Input DWI volume ‘dwi’

-select Output Baseline Volume
‘Create new Volume as...’, and name
it ‘baseline’
-select Output Diffusion Brain Mask
‘Create new LabelMapVolume as...’,
and name it ‘brain_mask’
-click on Apply.
 Creating a brain mask

Slicer displays the

edited brain mask
 Creating a brain mask

Change the label

layer to None to
make the mask
invisible
Estimating the tensor

Click on the Modules menu, then

select Diffusion -> Process ->
Diffusion Tensor Estimation
Estimating the tensor

-Set the Input DWI volume to ‘dwi’

-Set the Input Brain Mask to
‘brain_mask’
- Select Output DTI Volume ‘Create
DiffusionTensorVolume as ...’, and
name it ‘dti’
-Set Output Baseline Volume to
‘baseline’
-Under ‘Advanced Settings’, set
Fitting Methods to ‘WLS’ (Weighted
Least Squares)
-Click on Apply.
 Estimating the tensor

Position your mouse over

the pin icon, click on the
double arrow and select
the dti in the B field, set
the F and L to none.
 Exploring the DWI Dataset

Slicer displays the DTI

volume in color by
orientation mode:
Red: right-left
Green: anterior-posterior
Blue: inferior-superior
 Diffusion Tensor Data
- bgˆi T D gˆ i
Si = S 0 e
Stejskal-Tanner equation (1965)

D xx D xy D xz
D= D yx D yy D yz
D zx D zy D zz

The diffusion tensor D in the voxel (I,J,K) is a 3x3 symmetric matrix.

 Diffusion Tensor
• The diffusion tensor D in each voxel can be visualized as
a diffusion ellipsoid, with the eigenvectors indicating the
directions of the principal axes, and the ellipsoidal
proportional to the square root of the eigenvalues
defining the
• Scalar maps can be
derived from the
rotationally invariant
eigenvalues λ1, λ2, λ3
to characterize the size
and shape of the
diffusion tensor.
 Diffusion Tensor Shape

l1= l2= l3 l1>> l2, l3 l1~l2>> l3

Isotropic media Anisotropic media

(Cerebrospinal (white matter)
Fluid, gray matter)
Exploring the DWI Dataset

Click on the Slicer layout menu

and select the Yellow slice only
layout.
 Corpus Callosum

The corpus callosum

is a broad thick bundle
of dense myelinated
fibers that connect the
left and right
hemisphere. It is the
largest white matter
structure in the brain

Image from Gray’s Anatomy

 Corpus Callosum

Corpus Callosum
 Characterizing the Size of the
tensor: Trace
Trace(D) = λ1+ λ2+λ3

•Trace(D) is intrinsic to the tissue and is independent

of fiber orientation, and diffusion sensitizing gradient
directions
•Trace(D) is a clinically relevant parameter for
monitoring stroke and neurological condition (degree
of structural coherence in tissue)
•Trace(D) is useful to characterize the size of the
diffusion ellipsoid
 Trace

Click on the Modules menu, then select

Diffusion -> Quantify -> Diffusion Tensor
Scalar Maps
Trace

Type in the following information in

the IO menu:
-select the Operation ‘Trace’
-set Input DTI Volume to ‘dti’
-select Output Volume ‘Create new
Volume as...’ and name it ‘trace’
-click on Apply to calculate the trace
map of the tensor volume
 Trace

Set L as none.

The trace image appears in

the yellow viewer
 Trace

Adjust window level by

right-dragging up and
down.
 Trace

Position your mouse over

the pin icon and then
select the ‘>>’ icon to
display this table and fill in
the following information:
-Select the volume ‘trace’
in the Background viewer
-Select the volume ‘dti’ in
the Foreground viewer
Set the opacity of the dti
volume to 0.40
 Trace
Position your mouse within the
region of the Corpus Callosum
and observe the trace values in
the Data Probe
 Trace

Note how the Trace values are

fairly uniform in both white and gray
matter, even if the tissues are
different in structure.
Scalar Maps: Fractional Anisotropy
2 2 2
( λ1 − λ2 ) + ( λ1 − λ3 ) + ( λ2 − λ3 )
FA(D) =
2 2 2
2 λ +λ +λ
1 2 3

•FA(D) is intrinsic to the tissue and is independent

of fiber orientation, and diffusion sensitizing
gradient directions
•FA(D) is useful to characterize the shape (degree
of ‘out-of-roundness’) of the diffusion ellipsoid
•Low FA: High FA:
Fractional Anisotropy

Fill in the following information:

-Set Input DTI Volume to ‘dti’
-Select Output Scalar Volume
‘Create new Volume as ...’ and
name it ‘fa’
-In ‘Scalar Measurement’,
select ‘Fractional Anisotropy’
-Click on Apply to calculate the
Fractional Anisotropy map of the
tensor volume
 Fractional Anisotropy

Set L as none.

The FA image appears in the

yellow viewer
 Fractional Anisotropy

Position your mouse over

the pin icon and click the
‘>>’ icon to display this
table. Set the background
volume to ‘fa’ and be sure
the foreground volume is
still set to ‘dti’ with opacity
at 0.40
 Fractional Anisotropy
Explore the FA values in the Corpus
Callosum and in adjacent gray matter
areas. Note how the FA values are
high in the white matter areas, and
low in gray matter regions
 Fractional Anisotropy

Change to Conventional view

Part 2:
Visualizing the
tensor data
3D Visualization: Glyphs

Click on the Modules menu and

select the module Volumes
3D Visualization: Glyphs

Position the mouse over the pin

icon and select the ‘<<‘ icon to
display the axial slice toolbar.
Set the Foreground to ‘fa’ and
the Background to ‘dti’, with
the Foreground opacity set to
1.00
3D Visualization: Glyphs

Set the Active Volume to ‘dti’

and the Scalar Mode to
‘ColorOrientation’
3D Visualization: Glyphs

Scroll down the module panel and in

the Glyphs on Slices Display section:
-Check off the option for Red, Yellow,
and Green Slice Visibility

-Set the Color by Scalar parameter to

‘ColorOrientation’

-Set the Glyph Type to ‘Ellipsoids’

3D Visualization: Glyphs

The glyphs appear in all

3 slice viewers
3D Visualization: Glyphs

Position your mouse over the

pin icon select the eye icon
to display the axial, coronal,
and sagittal slices in the 3D
viewer
 3D Visualization: Glyphs

Slicer displays the anatomical

slices in the 3D viewer
 3D Visualization: Glyphs

Zoom in to observe the glyphs.

The ellipsoids represent the
principal direction of diffusion
(main eigenvector)
Diffusion MRI tractography

Deselect the option for Red,

Yellow, and Green Slice
Visibility, and deselect the
eye icon
Diffusion MRI tractography
Click L to reset the 3D
view to left

Position your mouse over the

pin icon and change the
Foreground to ‘None’ and the
background to ‘fa’
Part 3:
From tensors
to tracts
 DTI tractography
• Definition of a region of interest (ROI) for
seeding tract in an FA map (Editor module)

• Single-tensor tractography (Tractography

Interactive Seeding module)

• Fiducial-seeding tractography (Tractography

Interactive Seeding module)
Diffusion MRI tractography

Select the module Editor

Diffusion MRI tractography

Click on OK
 Diffusion MRI tractography
Select the Yellow slice only
layout
 Diffusion MRI tractography

Select the DrawEffect tool

 Diffusion MRI tractography

Outline the contour of the

Corpus Callosum with the
DrawEffect tool and press
enter. Repeat this step with 3
adjacent sagittal slices
 Diffusion MRI tractography

In the next section, we will seed

tracts from this anatomical
region of interest.
Diffusion MRI tractography

Click on the Modules and then select Diffusion

-> Tractography->Tractography Seeding.
 Step1: I/O

Change to Conventional view

-Set the Input DTI Volume to ‘dti’

-Set Output Fiber Bundle to
‘corpusCallosum’ by renaming the
default parameter ‘Fiber Bundle’
-Set the Input Fiducials, Model or
Label Map to ‘fa-label’
Step 2: Seeding parameters

Select the default Tractography Seeding

parameters:
- Threshold Type: FractionalAnistropy
- Seeding Threshold:0.30
- Stopping Threshold: 0.25

Click Update to generate tractography

 Step 3: Generate Tracts

The tracts generated in the

corpus callosum area appear in
the 3D viewer.
Step 4: Undesirable track removal

Click on the Modules menu,

then select Diffusion ->
Tractography -> Tractography
Display
Step 4: Undesirable track removal

Click Fiber Bundle Selection

Step 4: Undesirable track removal

In ‘Fiber Bundle Selection’, under ROI

for Fiber Select, create a new
AnnotationROI as ‘ROI node’ and
select Disable ROI.
Step 4: Undesirable track removal

Adjust the ROI frame to include the

undesirable tracks, using the
colorful spheres provided.
Step 4: Undesirable track removal

Click on Negative ROI to finish

Step 4: Undesirable track removal

Uncheck ROI Visibility

Fiducial Seeding

Click on the Modules and then select Diffusion

-> Tractography->Tractography Seeding.
 Fiducial Seeding
Position the mouse over the pin icon
and click on the eye icon to display the
axial slice in the 3D viewer

Unlink the link icon

Fiducial Seeding
Select the module Markups
 Fiducial Seeding

Click on the arrow icon to

create a fiducial
 Fiducial Seeding

Position the fiducial in the left cingulum

of the coronal slice
 Fiducial Seeding

Double click on the fiducial and change

the name to LeftCingulum
Fiducial Seeding

Click on the Modules and then select Diffusion ->

Tractography->Tractography Seeding.

Set the Input DTI volume to ‘dti’

Set the Input Fiducials, Model or Label Map to ‘F’
Select the Output Fiber Bundle ‘Create New Fiber
Bundle as ...’ and name it ‘Cingulum’
Click Update
 Fiducial Seeding

Part of the left cingulum

appears in the 3D viewer.

Move the fiducial and

update the Left Cingulum
fiducial to explore the
spatial relationship
between the left cingulum
and the corpus callosum
Tractography ‘on-the-fly’
Select the module Markups
Tractography ‘on-the-fly’

Select the List ‘Create new

MarkupsFiducial’

Click on the arrow icon to

create a new fiducial, and
position it in the 3D viewer
Tractography ‘on-the-fly’

Click on the Modules and then select Diffusion ->

Tractography->Tractography Seeding.

Set the Input DTI volume to ‘dti’

Set the Input Fiducials, Model or Label Map to ‘F_1’
Select the Output Fiber Bundle ‘Create New Fiber
Bundle as ...’ and name it ‘TractOnTheFly’
Check Update (check for interactive)
 Tractography ‘on-the-fly’

Move the fiducial F_1-1

in the 3D viewer to
explore the dti dataset
Tractography ‘on-the-fly’

The Fiducial Seeding functionality

allows you to do tractography ‘on-
the-fly’ to explore white matter
structures interactively
DTI Analysis

Select the module Data to

display the list of elements
that have been generated
in this tutorial
Conclusion
This tutorial guided you
through the different
steps of a Diffusion MR
analysis pipeline, from
tensor estimation to 3D
tracts visualization, for
exploring and studying
the 3D architecture of
the brain white matter.
 Acknowledgments
• Open Source Diffusion MRI Technology For Brain Cancer
Research NIH U01CA199459

• National Center for Image Guided Therapy (NCIGT)

NIH P41EB015898

• Neuroimage Analysis Center (NAC)

NIH P41EB015902

• Fan Zhang, Ph.D.

Brigham and Women’s Hospital, Harvard Medical School