228
V. Kurtcuoglu et al.
nerve roots based on the Visual Human Project’s Visible Man [73] and assessed
the effects of spinal cord eccentricity and cross-sectional geometry on CSF flow.
They found that spinal SAS CSF flow is inertia dominated and predicted that
patient-specific CFD models could become a noninvasive tool to explain abnormal
CSF dynamics and distribution of intrathecal drugs. In 2006, Stockman presented
a 3D lattice Boltzmann model of a subsection of the spinal canal with idealized
spinal cord nerve roots and arachnoid trabeculae [74]. Model findings indicated
that inclusion of trabecular microstructure had relatively little impact on the CSF
velocity profiles but did influence local flow mixing to a great degree and could,
therefore, impact tracer dispersion. In 2009 and 2010, Gupta et al. investigated
flow in an anatomically accurate representation of a healthy cranial subarachnoid
space and fourth ventricle [9, 10]. They imposed CSF flow velocities obtained
through MRI at the pontine and cerebellomedullary cisterns and at the foramen
magnum in the spinal canal. The subarachnoid space was modelled as an anisotropic
homogeneous porous medium. Their main finding was that the SAS microstructure
clearly contributes to the pressure drop along the domain. In contrast, Sweetman et
al. focused on the integration of spinal and cranial SAS as well as ventricular system
into one CFD model without accounting for microstructures [75], expanding their
previous 2D and 3D work on the cranial compartments [76, 77]. CSF pulsation
was produced by an imposed volume change in brain tissue transmitted to the fluid
domain through deformable superior wall sections of the lateral ventricles.
A series of studies was conducted by various researchers to understand more
deeply the impact of anatomy on CSF dynamics. Yiallourou et al. used the finite
volume method for several subject-specific 3D models of the cervical spine and
found peak CSF velocities to compare poorly with 4D phase-contrast MRI [78],
suggesting that differences could be due to lack of spinal cord nerve roots in
the model. Similarly, Lindstrøm et al. also found relatively poor agreement of
MRI-measured CSF velocities and CFD without spinal cord nerve roots [79].
In contrast, research by Clarke et al. showed relatively good agreement of CFD
and MRI-measured CSF velocities [80]. Pahlavian et al. investigated the impact
of anatomically realistic spinal cord nerve roots and denticulate ligaments on
CSF flow within the cervical spine [11]. Results showed that these anatomic
structures increased peak CSF velocities, mixing and bidirectional flow. Pahlavian
and colleagues then investigated the reliability of CFD-predicted CSF velocities
using MRI data collected from an in vitro model with spinal cord nerve roots
[81–83]. Results indicated that in vitro MRI compared well to CFD results, but
lacked agreement with in vivo measurements. Thus, Martin et al. studied interoperator reliability of MRI-based CFD prediction of CSF motion, finding fluid
dynamic and geometric results to have a high degree of reliability for healthy subject
geometries [84].
For all of the above simulations, either the SAS was modelled as a rigid structure
or CSF flow was imposed assuming a monotonically decreasing flow waveform
magnitude along the spinal axis. However, in vivo measurements indicate that the
CSF flow waveform changes in both magnitude and shape along the spine [78, 85,
86]. Thus, in 2017, Khani et al. carried out a CFD study of a cynomolgus monkey
V. Kurtcuoglu et al.
nerve roots based on the Visual Human Project’s Visible Man [73] and assessed
the effects of spinal cord eccentricity and cross-sectional geometry on CSF flow.
They found that spinal SAS CSF flow is inertia dominated and predicted that
patient-specific CFD models could become a noninvasive tool to explain abnormal
CSF dynamics and distribution of intrathecal drugs. In 2006, Stockman presented
a 3D lattice Boltzmann model of a subsection of the spinal canal with idealized
spinal cord nerve roots and arachnoid trabeculae [74]. Model findings indicated
that inclusion of trabecular microstructure had relatively little impact on the CSF
velocity profiles but did influence local flow mixing to a great degree and could,
therefore, impact tracer dispersion. In 2009 and 2010, Gupta et al. investigated
flow in an anatomically accurate representation of a healthy cranial subarachnoid
space and fourth ventricle [9, 10]. They imposed CSF flow velocities obtained
through MRI at the pontine and cerebellomedullary cisterns and at the foramen
magnum in the spinal canal. The subarachnoid space was modelled as an anisotropic
homogeneous porous medium. Their main finding was that the SAS microstructure
clearly contributes to the pressure drop along the domain. In contrast, Sweetman et
al. focused on the integration of spinal and cranial SAS as well as ventricular system
into one CFD model without accounting for microstructures [75], expanding their
previous 2D and 3D work on the cranial compartments [76, 77]. CSF pulsation
was produced by an imposed volume change in brain tissue transmitted to the fluid
domain through deformable superior wall sections of the lateral ventricles.
A series of studies was conducted by various researchers to understand more
deeply the impact of anatomy on CSF dynamics. Yiallourou et al. used the finite
volume method for several subject-specific 3D models of the cervical spine and
found peak CSF velocities to compare poorly with 4D phase-contrast MRI [78],
suggesting that differences could be due to lack of spinal cord nerve roots in
the model. Similarly, Lindstrøm et al. also found relatively poor agreement of
MRI-measured CSF velocities and CFD without spinal cord nerve roots [79].
In contrast, research by Clarke et al. showed relatively good agreement of CFD
and MRI-measured CSF velocities [80]. Pahlavian et al. investigated the impact
of anatomically realistic spinal cord nerve roots and denticulate ligaments on
CSF flow within the cervical spine [11]. Results showed that these anatomic
structures increased peak CSF velocities, mixing and bidirectional flow. Pahlavian
and colleagues then investigated the reliability of CFD-predicted CSF velocities
using MRI data collected from an in vitro model with spinal cord nerve roots
[81–83]. Results indicated that in vitro MRI compared well to CFD results, but
lacked agreement with in vivo measurements. Thus, Martin et al. studied interoperator reliability of MRI-based CFD prediction of CSF motion, finding fluid
dynamic and geometric results to have a high degree of reliability for healthy subject
geometries [84].
For all of the above simulations, either the SAS was modelled as a rigid structure
or CSF flow was imposed assuming a monotonically decreasing flow waveform
magnitude along the spinal axis. However, in vivo measurements indicate that the
CSF flow waveform changes in both magnitude and shape along the spine [78, 85,
86]. Thus, in 2017, Khani et al. carried out a CFD study of a cynomolgus monkey