Advanced search
1 file | 15.99 MB

Validation methods for diffusion weighted magnetic resonance imaging in brain white matter

Els Fieremans (UGent)
(2008)
Author
Promoter
(UGent) and (UGent)
Organization
Abstract
Magnetic resonance imaging or MRI is a powerful medical imaging technique. This non-invasive technique utilizes a strong static magnetic field and radio frequent electromagnetic waves to align the magnetic dipole moments of water molecules in the body. Images are created based on the local density of water molecules and their mutual interactions. MRI provides great contrast between soft tissues and is generally regarded to be a safe imaging method since no ionizing radiation is used. An important modality of MRI is diffusion weighted MRI (DWMRI), which provides unique biologically and clinically relevant information of water in tissues that is not available from other imaging modalities. The overall effect observed in a diffusion weighted MRI image voxel reflects, on a statistical basis, the displacement distribution of water molecules which are present in that voxel. During their random diffusion driven displacement, molecules probe tissue structure at a microscopic scale. As diffusion is a three-dimensional process, the molecular mobility is not necessarily equal in all directions, which enables the detection of tissue anisotropy. An interesting application is the diffusion anisotropy in brain white matter (WM) originating from its specific organization in bundles of more or less myelinated axonal fibres running in parallel. Assuming that the direction of the fastest diffusion indicates the overall orientation of the fibres, DW-MRI provides a way to map the spatial orientation of the white matter tracks in the brain. Fibre tractography algorithms are used to reconstruct and visualize the neural fibre tracts in the WM. DW-MRI is also useful in the diagnosis of stroke and to investigate white matter pathologies such as Alzheimer disease, multiple sclerosis (MS) and brain tumours. Despite the fact that the first clinical applications of DW-MRI were presented in 1986, problems with validation, quantification and interpretation of the DW-MRI data still persist. A hardware diffusion phantom serving as a gold standard for the quantitative validation of DW-MRI is crucial for clinical purposes but still not available. In addition, phantoms with a well-known anisotropic structure would be useful to develop and test fibre tractography algorithms. Moreover, the origin of the DW-MRI signal inWM is not completely understood and the contributions of the different compartments (diffusion in the intra-, extracellular space and exchange between the intra-, extracellular compartment) are elusive. A well-known physical or simulated phantom would be useful to gain more insight in this matter. In this Ph.D. dissertation, both hardware and software diffusion phantoms are presented for the validation and interpretation of DWMRI data. Hardware diffusion phantoms can be classified into isotropic and anisotropic phantoms. Isotropic phantoms contain liquids such as water. They are widely used to test diffusion sequences and to evaluate related imaging artefacts. Anisotropic phantoms are essential in order to evaluate quantitatively measured parameters expressing the anisotropy such as the diffusion tensor and the fractional anisotropy (FA). They are also useful in analysing the variability of different MR-scanners in terms of anisotropy and fibre orientation. Both biological diffusion phantoms (plants and excised spinal cord) and synthetic diffusion phantoms are used for the validation of DW-MRI. We focus here on the development of anisotropic fibre phantoms because on the one hand, they have a well-known and reproducible structure and on the other hand they can be used to imitate complex geometries such as curved fibres and fibre crossings. A fibre phantom bundle consists of parallel fibres placed in water and surrounded by a shrinking tube to pack the fibres densely. In order to investigate how the different fibre material properties, size of the fibres and packing density influence the outcome of the DW-MRI experiment, fibre bundles are created with varying fibre density and made of different fibre materials (carbon fibre, nylon fibre, Dyneema R and fibreglass). The fibre density and fibre diameter are the two major factors determining the diffusion properties such as the FA, while the SNR is mainly determined by the surface relaxation and the magnetic susceptibility of the fibre. The most appropriate fibres to manufacture diffusion phantoms turn out to be densely packed fibre bundles made from a hydrophobic material with a magnetic susceptibility close to water. Of the tested fibre materials, Dyneema R fibres show to have the best fibre characteristics for making diffusion phantoms and are used further in this work. The diffusion in the Dyneema R fibre bundles is measured using DW-MRI and bulk NMR diffusion measurements. The measured diffusion properties are compared to simulations. The diffusion coefficient and kurtosis in the interstitial space between fibres is modelled using Monte Carlo (MC) simulations of random walkers. The timedependent apparent diffusion coefficient and kurtosis are simulated in geometries of parallel fibres with varying packing geometries and packing densities. The MC simulations confirm the accuracy and validity of the existing analytical models for ordered packing geometries. The simulations in the random packed fibre geometries show a higher FA and a longer diffusion transition time between the short and longtime diffusion limit in comparison with ordered packing geometries. Based on the correspondence between simulations and experimental measurements, the fibre phantoms are shown to be useful for the quantitative validation of DW-MRI on clinical MRI-scanners. Next, the MC simulations are elaborated in a geometry with intra and extracellular compartments imitating the WM. The simulations are extended to simulate the DW-MRI signal and exchange is enabled between the different compartments. The bi-exponential model and the cumulant expansion form are evaluated as models for the DWMRI signal in WM. The validity of these models is investigated by evaluation of the models as a function of the considered b-interval and considering the effect of exchange. The simulated DW-MRI signal shows to be rather pseudo-biexponential because there is a strong dependence on the diffusion time and considered b-interval. Moreover, when assuming that a biexponential function models the diffusion in two compartments, i.e. the intra- and extracellular space, this model is not straightforward interpretable in case of exchange between those compartments. The cumulant expansion form is proposed as an alternative. The diffusion coefficient and kurtosis can be accurately fitted when including higher order terms in the cumulant expansion form. The kurtosis appears to be a useful parameter to measure the exchange between the intra- and extracellular compartments. It could be potentially useful to correlate the kurtosis with the observed changes in cell permeability to cell pathological changes such as during stroke and in malignant tumours. Finally, some potential applications of Dyneema R fibre phantoms for the validation of DW-MRI are demonstrated as well in this work. As they have a well-known structure and anisotropy, they show to be suitable for sequence design, optimisation and the evaluation of imaging artefacts. Thanks to the flexibility and the variety of the shrinking tubes, an anthropomorphic head phantom containing the major WM in vivo fibre tracts can be created. In addition, a crossing fibre phantom is manufactured to test and evaluate fibre tractography algorithms.

Downloads

  • (...).pdf
    • full text
    • |
    • UGent only
    • |
    • PDF
    • |
    • 15.99 MB

Citation

Please use this url to cite or link to this publication:

Chicago
Fieremans, Els. 2008. “Validation Methods for Diffusion Weighted Magnetic Resonance Imaging in Brain White Matter”. Ghent, Belgium: Ghent University. Faculty of Engineering.
APA
Fieremans, E. (2008). Validation methods for diffusion weighted magnetic resonance imaging in brain white matter. Ghent University. Faculty of Engineering, Ghent, Belgium.
Vancouver
1.
Fieremans E. Validation methods for diffusion weighted magnetic resonance imaging in brain white matter. [Ghent, Belgium]: Ghent University. Faculty of Engineering; 2008.
MLA
Fieremans, Els. “Validation Methods for Diffusion Weighted Magnetic Resonance Imaging in Brain White Matter.” 2008 : n. pag. Print.
@phdthesis{891081,
  abstract     = {Magnetic resonance imaging or MRI is a powerful medical imaging technique. This non-invasive technique utilizes a strong static magnetic field and radio frequent electromagnetic waves to align the magnetic dipole moments of water molecules in the body. Images are created based on the local density of water molecules and their mutual interactions. MRI provides great contrast between soft tissues and is generally regarded to be a safe imaging method since no ionizing radiation is used.
An important modality of MRI is diffusion weighted MRI (DWMRI), which provides unique biologically and clinically relevant information of water in tissues that is not available from other imaging modalities. The overall effect observed in a diffusion weighted MRI image voxel reflects, on a statistical basis, the displacement distribution of water molecules which are present in that voxel. During their random diffusion driven displacement, molecules probe tissue structure at a microscopic scale. As diffusion is a three-dimensional process, the molecular mobility is not necessarily equal in all directions, which enables the detection of tissue anisotropy.
An interesting application is the diffusion anisotropy in brain white matter (WM) originating from its specific organization in bundles of more or less myelinated axonal fibres running in parallel. Assuming that the direction of the fastest diffusion indicates the overall orientation of the fibres, DW-MRI provides a way to map the spatial orientation of the white matter tracks in the brain. Fibre tractography algorithms are used to reconstruct and visualize the neural fibre tracts in the WM. DW-MRI is also useful in the diagnosis of stroke and to investigate white matter pathologies such as Alzheimer disease, multiple sclerosis (MS) and brain tumours.
Despite the fact that the first clinical applications of DW-MRI were presented in 1986, problems with validation, quantification and interpretation of the DW-MRI data still persist. A hardware diffusion phantom serving as a gold standard for the quantitative validation of DW-MRI is crucial for clinical purposes but still not available. In addition, phantoms with a well-known anisotropic structure would be useful to develop and test fibre tractography algorithms. Moreover, the origin of the DW-MRI signal inWM is not completely understood and the contributions of the different compartments (diffusion in the intra-, extracellular space and exchange between the intra-, extracellular compartment) are elusive. A well-known physical or simulated phantom would be useful to gain more insight in this matter.
In this Ph.D. dissertation, both hardware and software diffusion phantoms are presented for the validation and interpretation of DWMRI data. Hardware diffusion phantoms can be classified into isotropic and anisotropic phantoms. Isotropic phantoms contain liquids such as water. They are widely used to test diffusion sequences and to evaluate related imaging artefacts. Anisotropic phantoms are essential in order to evaluate quantitatively measured parameters expressing the anisotropy such as the diffusion tensor and the fractional anisotropy (FA). They are also useful in analysing the variability of different MR-scanners in terms of anisotropy and fibre orientation. Both biological diffusion phantoms (plants and excised spinal cord) and synthetic diffusion phantoms are used for the validation of DW-MRI.
We focus here on the development of anisotropic fibre phantoms because on the one hand, they have a well-known and reproducible structure and on the other hand they can be used to imitate complex geometries such as curved fibres and fibre crossings. A fibre phantom bundle consists of parallel fibres placed in water and surrounded by a shrinking tube to pack the fibres densely. In order to investigate how the different fibre material properties, size of the fibres and packing density influence the outcome of the DW-MRI experiment, fibre bundles are created with varying fibre density and made of different fibre materials (carbon fibre, nylon fibre, Dyneema R and fibreglass). The fibre density and fibre diameter are the two major factors determining the diffusion properties such as the FA, while the SNR is mainly determined by the surface relaxation and the magnetic susceptibility of the fibre. The most appropriate fibres to manufacture diffusion phantoms turn out to be densely packed fibre bundles made from a hydrophobic material with a magnetic susceptibility close to water. Of the tested fibre materials, Dyneema R fibres show to have the best fibre characteristics for making diffusion phantoms and are used further in this work. The diffusion in the Dyneema R fibre bundles is measured using DW-MRI and bulk NMR diffusion measurements. The measured diffusion properties are compared to simulations. The diffusion coefficient and kurtosis in the interstitial space between fibres is modelled using Monte Carlo (MC) simulations of random walkers. The timedependent apparent diffusion coefficient and kurtosis are simulated in geometries of parallel fibres with varying packing geometries and packing densities. The MC simulations confirm the accuracy and validity of the existing analytical models for ordered packing geometries. The simulations in the random packed fibre geometries show a higher FA and a longer diffusion transition time between the short and longtime diffusion limit in comparison with ordered packing geometries. Based on the correspondence between simulations and experimental measurements, the fibre phantoms are shown to be useful for the quantitative validation of DW-MRI on clinical MRI-scanners.
Next, the MC simulations are elaborated in a geometry with intra and extracellular compartments imitating the WM. The simulations are extended to simulate the DW-MRI signal and exchange is enabled between the different compartments. The bi-exponential model and the cumulant expansion form are evaluated as models for the DWMRI signal in WM. The validity of these models is investigated by evaluation of the models as a function of the considered b-interval and considering the effect of exchange. The simulated DW-MRI signal shows to be rather pseudo-biexponential because there is a strong dependence on the diffusion time and considered b-interval. Moreover, when assuming that a biexponential function models the diffusion in two compartments, i.e. the intra- and extracellular space, this model is not straightforward interpretable in case of exchange between those compartments. The cumulant expansion form is proposed as an alternative. The diffusion coefficient and kurtosis can be accurately fitted when including higher order terms in the cumulant expansion form. The kurtosis appears to be a useful parameter to measure the exchange between the intra- and extracellular compartments. It could be potentially useful to correlate the kurtosis with the observed changes in cell permeability to cell pathological changes such as during stroke and in malignant tumours.
Finally, some potential applications of Dyneema R  fibre phantoms for the validation of DW-MRI are demonstrated as well in this work. As they have a well-known structure and anisotropy, they show to be suitable for sequence design, optimisation and the evaluation of imaging artefacts. Thanks to the flexibility and the variety of the shrinking tubes, an anthropomorphic head phantom containing the major WM in vivo fibre tracts can be created. In addition, a crossing fibre phantom is manufactured to test and evaluate fibre tractography algorithms.},
  author       = {Fieremans, Els},
  isbn         = {9789085782247},
  language     = {eng},
  pages        = {X, j, 182},
  publisher    = {Ghent University. Faculty of Engineering},
  school       = {Ghent University},
  title        = {Validation methods for diffusion weighted magnetic resonance imaging in brain white matter},
  year         = {2008},
}