… and now for some equations:  An interactive course on diffusion MRI

In two previous blog posts (Diffusion without equations Parts I and II), we described the basics of diffusion MRI qualitatively.  We have now developed a larger course for MR scientists interested in an introduction to diffusion MRI. The course consists of six lectures (plus one bonus lecture) taken from various educational sessions held at ISMRM meetings around the world. The course was put together in collaboration with the ISMRM Education Committee, and you can watch it in its entirety below.

Diffusion MRI spans a great amount of territory, from large-scale neuronal connectivity (tractography) to detailed description of neuronal microstructure, tumor microstructure, and more. This course covers the methodology used for these applications, as well as acquisition and pre-processing of diffusion MRI data.

Historically, the development of the methodology used for tractography and microstructure has been somewhat separate, and at the most recent ISMRM annual meeting in Honolulu, the diffusion study group was described as an uneasy marriage between tractography and microstructure.  We would contend that it is a marriage that keeps getting better and better, with the spouses increasingly combining their strengths to perform analysis such as microstructure informed tractography. Let us propose that Els Fieremans, Karla Miller, Alexander Leemans, Maxime Descoteaux, Ileana Jelescu, and Daniel Alexander, together with many other contributors to this field, have been valuable marriage counselors. Let’s hear what they have to say!

«

Prev

1

/

1

Next

»

Introduction to Diffusion MRI, Els Fieremans

What is Your Sequence, Karla Miller

Analysis Pipelines for Diffusion MRI Data: From Voxels to Connectomes, Alexander Leemans

Diffusion Tractography: Principles & Methods, Maxime Descoteaux

Analysis: Tissue & Signal Models, Ileana Jelescu

Basics of Diffusion Modeling: Signal Models Versus Biophysical Models, Daniel C. Alexander

«

Prev

1

/

1

Next

»

Do you think you already know this stuff? Then try the interactive quiz below! The questions also serve as course annotations, so casual readers can find out which topics are addressed in the course. Clicking on the question will take you to the exact video location where the topic is discussed. Or you can just try to answer without watching the video, and see if you got it right.

Els Fieremans – Introduction to Diffusion MRI

In biological tissue, for a very short time scale (a few ms), the diffusion of the water molecules can be described as:

a) Isotropic and Gaussian

b) Anisotropic and Gaussian

c) Isotropic and non-Gaussian

d) Anisotropic and non-Gaussian

Correct! Wrong!

2. Which of the following statements is true about the diffusion of water molecules in the brain, over a timescale of roughly 50 ms?

Which of the following statements is true about the diffusion of water molecules in the brain, over a timescale of roughly 50 ms?

a) The diffusion is anisotropic, i.e., relatively fast parallel compared to perpendicular to the fibers, in the white matter

b) The diffusion is isotropic in both gray and white matter

c) The diffusion is relatively isotropic in the gray matter compared to the white matter

a) and c)

None of the above

Correct! Wrong!

3.Which parameter should be changed to increase the diffusion length/displacement being probed in a diffusion MRI experiment?

Which parameter should be changed to increase the diffusion length/displacement being probed in a diffusion MRI experiment?

a) Echo Time (TE)

b) Diffusion time (Delta)

c) Gradient strength (G)

d) None of the above

Correct! Wrong!

4. How will the log of the normalized diffusion weighted signal change if the diffusion sensitizing gradient strength is doubled?

How will the log of the normalized diffusion weighted signal change if the diffusion sensitizing gradient strength is doubled?

a) It will increase by a factor of 2

b) It will decrease by a factor of 4

c) It will decrease by a factor of 2

d) It will increase by a factor of 4

Correct! Wrong!

5. What is the relationship between the b-value and the resolution of the displacement distribution estimated with q-space techniques?

What is the relationship between the b-value and the resolution of the displacement distribution estimated with q-space techniques?

a) They are not related

b) They are equal

c) They are related by a simple mathematical equation

Correct! Wrong!

Karla Miller – What is Your Sequence?

1. What is the key effect that creates diffusion weighted MR contrast?

What is the key effect that creates diffusion weighted MR contrast?

a) Magnetization dephasing caused by discrepancies in the magnetic field seen by different spins

b) Average signal phase induced by discrepancies in the radio frequency (RF) pulse series

c) Average signal phase induced by the magnetic field gradients

d) Magnetization dephasing caused by discrepancies in the relaxation time of different spins

Correct! Wrong!

2. Which of the following are advantages of single shot echo planar imaging (ss-EPI)?

Which of the following are advantages of single shot echo planar imaging (ss-EPI)?

a) It produces images with high resolution and few distortions

b) It is fast

c) It freezes the motion

a) and b)

a) and c)

b) and c)

Correct! Wrong!

3. How can one significantly reduce the scan time in diffusion imaging without altering the contrast?

How can one significantly reduce the scan time in diffusion imaging without sacrificing the contrast or the volumetric coverage?

a) By doing single-slice excitation

b) By doing simultaneous multi-slice excitation

c) By reducing the gradient duration

d) By reducing the diffusion time

Correct! Wrong!

4. How does the time between the second and the third 90° radiofrequency (RF) pulses in a stimulated echo sequence affect the diffusion time dependent contrast?

How does the time between the second and the third 90° radiofequency (RF) pulses in a stimulated echo sequence affect the diffusion time dependent contrast?

a) The contrast decays as the time between the two 90° RF pulses increase

b) The contrast is being accumulated during the time between the two 90° RF pulses

c) The contrast is not affected by the time between the two 90° RF pulses

Correct! Wrong!

5. Why is the steady state free precession (SSFP) sequence good for diffusion imaging of tissues with short T2?

Why is the steady state free precession (SSFP) sequence good for diffusion imaging of tissues with short T2?

a) Because of the short TR

b) Because the duration of the diffusion gradients is short

c) Because of the inherent T1/T2 contrast of SSFP

d) Because it does not require navigator echoes to image tissue in vivo

Correct! Wrong!

Alexander Leemans – Analysis Pipelines for Diffusion MRI Data: From Voxels to Connectomes

1. How is the signal intensity drift in dMRI data best corrected?

How is the signal intensity drift in dMRI data best corrected?

a) By low-degree polynomial fitting of the b0 images

b) By low-degree polynomial fitting of the b0 and the diffusion weighted images

c) By high-degree polynomial fitting of the b0 images

Correct! Wrong!

2. When correcting for subject motion in diffusion MRI data, which processing step should be included to prevent a bias?

When correcting for subject motion in diffusion MRI data, which processing step should be included to prevent a bias?

a) Rotation of the b-matrix

b) Recalibration of the b-values

c) Normalization of the non-diffusion-weighted signal intensity

Correct! Wrong!

3. How does the size of a tract passing close to the grey matter or the ventricles determine the extent of the confound caused by partial voluming, for statistics computed for the entire tract volume?

How does the size of a tract passing close to the grey matter or the ventricles determine the extent of the confound caused by partial voluming, for statistics computed for the entire tract volume?

a) Bigger (in diameter) tracts are more affected

b) Bigger (in diameter) tracts are less affected

d) The size is irrelevant to the extent of this confound.

Correct! Wrong!

4. Which of the following could bias the Tract-based spatial statistics (TBSS) analysis?

Which of the following could bias the Tract-based spatial statistics (TBSS) analysis?

a) The skeletonization by projection might miss the pathology

b) Complex subvoxel fiber geometry

c) Motion

d) All of the above

Correct! Wrong!

Maxime Descoteaux – Diffusion Tractography: Principles & Methods

1. How often does more than one fiber orientation occupy a white matter voxel in a diffusion MRI acquisition with 2 mm isotropic voxel size?

How often does more than one fiber orientation occupy a white matter voxel in a diffusion MRI acquisition with 2 mm isotropic voxel size?

a) less than 10%

b) 10%-50%

c) 50%-90%

e) more than 90%

Correct! Wrong!

2. What is the recommended step size and the order of interpolation for tractography?

What is the recommended step size and the order of interpolation for tractography?

a) 1/10 of the voxel grid and low order interpolation

b) 1/2 of the voxel grid and low order interpolation

c) 1/10 of the voxel grid and higher order interpolation

d) 1/2 of the voxel grid and higher order interpolation

Correct! Wrong!

3. What is the main drawback of probabilistic tracking?

What is the main drawback of probabilistic tracking?

a) It produces more invalid connections in comparison to deterministic approaches

b) It cannot track highly curved structures

c) It is not reproducible

Correct! Wrong!

4. Which structures will be over-represented if the tractography is seeded from every voxel in the brain white matter?

Which structures will be over-represented if the tractography is seeded from every voxel in the brain white matter?

a) Large and short bundles

b) Thin and long bundles

c) Thin and short bundles

d) Large and long bundles

Correct! Wrong!

5. The seeding bias can be overcome by:

The seeding bias can be overcome by:

a) Including anatomical priors

b) Seeding from the grey-white matter interface

c) Particle filtering tractography

All of the above

Correct! Wrong!

Ileana Jelescu – Diffusion Analysis: Tissue & Signal Models

1. Which of the following is a main advantage of the tissue models over the signal models?

Which of the following is a main advantage of the tissue models over the signal models?

a) Tissue models have higher specificity

b) Tissue models have higher sensitivity

c) The tissue models are applicable to any type of tissue

Correct! Wrong!

2. What is the maximum b-value that is suitable for use in signal models derived from the second order Taylor expansion of bD in healthy brain?

What is the maximum b-value that is suitable for use in signal models derived from the second order Taylor expansion of bD in healthy brain?

a) 1500

b) 2500

c) 3500

c) 4500

Correct! Wrong!

3. Which of the following could be a consequence of fixed parallel diffusivity parameters in a multi-compartment model?

Which of the following could be a consequence of fixed parallel diffusivity parameters in a multi-compartment model?

a) Biased volume fractions.

b) The intra-axonal compartment will appear less anisotropic than the extra-axonal compartment.

c) Reduced specificity.

a) and b)

a) and c)

b) and c)

Correct! Wrong!

4. Which of the following has been suggested by the diffusion MRI literature?

Which of the following has been suggested by the diffusion MRI literature?

a) The parallel diffusivity in the intra-axonal compartment is less than the parallel diffusivity in the extra-axonal compartment.

b) The parallel diffusivity in the intra-axonal compartment is greater than the parallel diffusivity in the extra-axonal compartment

c) The parallel diffusivity in the intra-axonal compartment is equal to the parallel diffusivity in the extra-axonal compartment.

d) All of the above

Correct! Wrong!

5. What is a general feature of the two compartment models?

What is a general feature of the two compartment models?

a) The estimates of the parameters have two solutions

b) The estimates of the parameters have only one solution

c) The two compartment models are used only in grey matter

Correct! Wrong!

Daniel Alexander – Diffusion Modelling: Signal Models vs. Biophysical Models 

1. Which of the following metrics, derived from the reconstructed diffusion propagator, would be indicative of anisotropy?

Which of the following metrics, derived from the reconstructed diffusion propagator, would be indicative of anisotropy?

a) Return to origin probability

b) Angular variance

c) Mean displacement

d) Radial variance

Correct! Wrong!

2. Which of these entities can characterize departure of the signal from a single Gaussian?

Which of these entities can characterize departure of the signal from a single Gaussian?

a) Diffusion kurtosis

b) The diffusion tensor

c) The high angular resolution ADC profile

d) Anomalous diffusion parameters

e) Bi-exponential models

f) Mean diffusivity

a) , c), d) and e)

b) , c), d) and f)

a) , c), e) and f)

Correct! Wrong!

3. Which of the following parameters are estimated by the AxCaliber model, but not by the Neurite Orientation Dispersion and Density Imaging (NODDI) model?

Which of the following parameters are estimated by the AxCaliber model, but not by the Neurite Orientation Dispersion and Density Imaging (NODDI) model?

a) It estimates the axon diameter

b) It models the distribution of the myelin content wrapped around the axon

c) It models the the hindered and the restricted diffusion compartments

d) It estimates the diffusivity

e) It estimates the orientation dispersion

c) and d)

a) and c)

a) and e)

a) and d)

Correct! Wrong!

4. What assumption is made by some diffusion models to aid in distinguishing the intra-axonal from the extra-axonal space?

What assumption is made by some diffusion models to aid in distinguishing the intra-axonal from the extra-axonal space?

a) The axons are modelled as sticks with zero diffusivity perpendicular to the stick direction

b) The axons are modelled as sticks with zero diffusivity parallel to the stick direction

c) The anisotropy of the extra-axonal space is assumed to be higher than that of the intra-axonal space

d) The diffusivity perpendicular to the stick direction is assumed to be equal to that of cerebrospinal fluid (CSF)

Correct! Wrong!

5. Rank the models in ascending order, with respect to their complexity (number of parameters):

Rank the models in ascending order, with respect to their complexity (number of parameters):

1. Diffusion Tensor Imaging (DTI);
2. Neurite Orientation Dispersion and Density Imaging (NODDI);
3. Bi-exponential;
4. Diffusion Kurtosis Imaging (DKI);
5. Apparent Diffusion Coefficient (ADC)

1. Apparent Diffusion Coefficient (ADC);
2. Bi-exponential;
3. Neurite Orientation Dispersion and Density Imaging (NODDI);
4. Diffusion Tensor Imaging (DTI);
5. Diffusion Kurtosis Imaging (DKI)

1. Bi-exponential;
2. Apparent Diffusion Coefficient (ADC);
3. Neurite Orientation Dispersion and Density Imaging (NODDI);
4. Diffusion Tensor Imaging (DTI);
5. Diffusion Kurtosis Imaging (DKI)

1. Diffusion Kurtosis Imaging (DKI);
2. Diffusion Tensor Imaging (DTI);
3. Neurite Orientation Dispersion and Density Imaging (NODDI);
4. Apparent Diffusion Coefficient (ADC);
5. Bi-exponential

1. Apparent Diffusion Coefficient (ADC);
2. Neurite Orientation Dispersion and Density Imaging (NODDI);
3. Bi-exponential;
4. Diffusion Tensor Imaging (DTI);
5. Diffusion Kurtosis Imaging (DKI)

1. Apparent Diffusion Coefficient (ADC);
2. Neurite Orientation Dispersion and Density Imaging (NODDI);
3. Diffusion Tensor Imaging (DTI);
4. Bi-exponential;
5. Diffusion Kurtosis Imaging (DKI)

Correct! Wrong!

 

Now that you are done, we would love to get your feedback! Feel free to discuss the current questions in the comments section, or even better, contribute new ones! It doesn’t matter whether you are an expert or a novice, your knowledge (or lack thereof) could contribute to a fun experiment in crowdsourcing MR educational content. Our aim is to explore new ways to educate MR scientists. If successful, this kind of course structure may be considered for CME credit in the future.

P.S. For a historical perspective on the field, we highly recommend the interview that Derek Jones conducted with John Tanner for last year’s edition of the MRM Highlights magazine.