Modeling of Brain Shift Phenomenon for Different Craniotomies and Solid Models.
ABSTRACT This study investigates the effects of different solid models on predictions of
brain shift for three craniotomies. We created a generic 3D brain model based on
healthy human brain and modeled the brain parenchyma as single continuum and
constrained by a practically rigid skull. We have used elastic model,
hyperelastic 1st, 2nd, and 3rd Ogden models, and hyperelastic Mooney-Rivlin with
2- and 5-parameter models. A pressure on the brain surface at craniotomy region
was applied to load the model. The models were solved with the finite elements
package ANSYS. The predictions on stress and displacements were compared for
three different craniotomies. The difference between the predictions of elastic
solid model and a hyperelastic Ogden solid model of maximum brain displacement
and maximum effective stress is relevant.
Hindawi Publishing Corporation
Journal of Applied Mathematics
Volume 2012, Article ID 409127, 20 pages
Modeling of Brain Shift Phenomenon for Different
Craniotomies and Solid Models
Alvaro Valencia,1Benjamin Blas,1and Jaime H. Ortega2
1Department of Mechanical Engineering, Universidad de Chile, Beauchef 850, Santiago, Chile
2Department of Mathematical Engineering and Center for Mathematical Modeling, Universidad de Chile,
Av. Blanco Encalada 2120, Santiago, Chile
Correspondence should be addressed to Alvaro Valencia, firstname.lastname@example.org
Received 15 July 2011; Revised 4 October 2011; Accepted 20 October 2011
Academic Editor: Venky Krishnan
Copyright q 2012 Alvaro Valencia et al. This is an open access article distributed under the
Creative Commons Attribution License, which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.
This study investigates the effects of different solid models on predictions of brain shift for three
craniotomies. We created a generic 3D brain model based on healthy human brain and modeled
the brain parenchyma as single continuum and constrained by a practically rigid skull. We have
used elastic model, hyperelastic 1st, 2nd, and 3rd Ogden models, and hyperelastic Mooney-Rivlin
with 2- and 5-parameter models. A pressure on the brain surface at craniotomy region was
applied to load the model. The models were solved with the finite elements package ANSYS.
The predictions on stress and displacements were compared for three different craniotomies. The
difference between the predictions of elastic solid model and a hyperelastic Ogden solid model of
maximum brain displacement and maximum effective stress is relevant.
Neurosurgery requires high levels of accuracy due to the complexity of the brain. To do
this, surgeons have preoperative images that identify the exact area of operation. However,
during craniotomy, a change on loading condition occurs, that causes brain deformation.
The deformations carry a margin of error in the surgery area. The phenomenon known
as brain shift deformations will be studied in this work. We note that the brain shift is a
negative effect that occurs in the surgical opening of the skull ?craniotomy?. Brain shift is
produced by a pressure difference on the brain induced in the region of the craniotomy. This
changes the position of the pathology and healthy tissues from the calculated with high-
quality preoperative radiographic images.
it to the patient coordinate system. This assumes that the brain is rigid and is a source of error
in the exact determination of tumor position.
2Journal of Applied Mathematics
There are several factors that determinate the magnitude of brain shift produced by
a craniotomy: gravity, mechanical tissue properties, loss of brain-spinal fluid, anatomical
constrains, intracranial pressure, and patient variability.
A current challenge is the determination of stress and displacements in a solid model
of the brain subject to a craniotomy. The geometry of the brain is very complex, and the
characteristics of the tissue are not easy to measure and model. The results of the solid model
will help to correct the position of the brain for the surgical navigator system.
Most solid brain models use elastic model ?1? in order to model the deformations and
stress of the brain tissue. Using an elastic solid model, the Young modulus does not affect the
displacement field if the gravity is not considered ?2?.
The effects of considering hyperelastic model of brain have been considered in few
works in the literature. The use of nonlinear solid model made it possible to obtain very good
predictions of deformation of ventricles and tumor ?3?. The same authors have supposed
that the brain deformations depend very weakly on the constitutive model and mechanical
properties of the brain tissues, and therefore simple hyperelastic model can be used ?4?.
Several authors propose the use of the linear elasticity to model the stress and
deformations of the brain tissue ?5–7?. The linear elasticity considers the determination
of some parameters as the elasticity modulus ?E?, the shear modulus, or second Lam´ e’s
parameter ?μ,λ? among others. The models consider just one brain tissue, isotropic and
incompressible, which is a simplification. Then, they assume that the brain is immersed
into the cerebrospinal liquid which is contained by the rigid skull. It is clear that this liquid
which is nondeformable. The elasticity modulus is similar to the human bones, that is the
elasticity modulus of the skull is 6.5GN/m2and the Poisson constant of the skull is 0.22.
The skull is a rigid structure, which contains three elements, the brain tissue ?86%?,
blood ?4%?, and cerebrospinal liquid ?CRL, 10%?. The interaction among these elements
produces a pressure called intracranial pressure. Normally, this pressure in a health adult is
around 10mmHg ?1332.8Pa? and must not be higher than 15mmHg. The density of the CRL
is 1007kg/m3. Furthermore, the brain tissue corresponds only to the 2% of the total weight
but is the element with highest intracranial volume. The weight of the brain is between 1300
and 1600gr, and its volume is around 1000 to 1500cc. Its density is closer to the water density,
that is 1040kg/m3.
We can see that, about the properties of the brain tissue, in particular, the values of the
elasticity constants, there are several differences among the authors ?8, 9?.
It is important to remark that the linear elasticity has a suitable behavior for small
deformations, and it is clear from several authors that the relationship between stress and
deformation for soft tissue is not linear ?10, 11?.
In the present investigation, we report the effects of hyperelastic solid models on
maximal displacement and effective stress of the brain. We have calculated the brain shift
for three craniotomies.
2. Mathematical Models
The linear elasticity theory is the study of linear elastic solids undergoing small deformations.
The linearity means that the components of the stress tensor are a linear combination of the
Journal of Applied Mathematics3
The relationship that defines each element of the strain tensor is shown in ?2.1?. This
tensor is known as infinitesimal tensor of Green-Cauchy:
with i,j ? 1,2,3.
The constitutive equations of linear elasticity for an elastic solid are represented by
generalized Hooke’s law:
where i,j ? 1,2,3.
However, if the material is assumed homogeneous and isotropic, we obtain the consti-
tutive equation of Lam´ e-Hooke.
2.1. Constitutive Equation of Lam´ e-Hooke
Itiswellknownthat,consideringahomogeneous andisotropicmaterial,weobtaintheLam´ e-
Hooke constitutive equations. That means that the components of the elastic tensor depend
on two particular constants of each material, these constants are the so-called Lam´ e modulus
?λ,μ?. The relation between the elastic coefficients and the Lam´ e modulus is the following:
Cijkm? λδijδkm? μ?δikδjm? δim? δjk
Finally, after some simplifications, we have
σij? λδijεkk? 2μεij,
where εkkis the trace of the deformation tensor.
2.2. Nonlinear Elasticity
The nonlinear elasticity is an observed phenomenon in elastomeric material ?rubber? and
porous media. The origin of both materials is different, for instance, the elastomeric materials,
which are polymers, can be synthetic or natural rubber, and, on the other hand, porous
media exist in the nature in form of organic materials, vegetal and animal tissue. The main
characteristic of this material is their deformation capacity, which can arrive from 200% to
300%. Nevertheless, these large deformations can be recovered, and the material comes back
to its natural state. It is important to note that the human tissues behave as a nonlinear elastic
4 Journal of Applied Mathematics
In what follows, we will present some ideas on the nonlinear elastic models. Firstly, we
will define the relationship between the strains and the displacement vector, which is defined
in ?2.5?, which is a nonlinear relationship:
with i,j ? 1,2,3.
From the above equation, we can obtain the strain tensor or Green-Lagrange strain
tensor. This tensor helps to quantify the length changes of the material and the variation of
the angle between the material fibers.
The deformation energy is a useful function in order to define a hyperelastic material.
This function gives a relation between the stored energy with the strain and deformations
generated in the solid. Moreover, its derivatives with respect to stretch give us the stress
it is necessary to introduce the deformation gradient
This tensor represents the variation of a deformed material point with respect to its
initial state. To simplify the computations, it is interesting to obtain the Green-Cauchy
left deformation tensor ??B?? and the Green-Cauchy right deformation tensor ??C??, both
can be recovered from the deformation gradient tensor, and the Green-Cauchy invariant
deformation tensor can be easily obtained:
?B? ? ?F??F?T,
?C? ? ?F?T?F?.
deformation tensor ??B??. If we assume isotropy of the material, the energy depends on the
first three invariants of the tensor
W ? W?I1,I2,I3?.
The invariants for an isotropic material are as follows.
Journal of Applied Mathematics5
If the normal forces are parallel to the principal direction of the material, we have that
the invariants only depend on the principal elongations of the solid. It is important to remark
that this condition can occur in every isotropic material, and this is due to that in all directions
the measurements are equal.
The principal stretches are defined as the quotient between the final length and the
initial length in the direction of the deformation. The invariants are functions of the principal
where λ1, λ2, and λ3are the stretch in the principal directions.
If we suppose that the solid is almost incompressible or with a high compression
modulus, the deformation energy depends only on the first and second invariant, since the
third invariant verifies I3 ? 1. The Cauchy strains are calculated from the derivative with
respect to the deformations of the deformation energy, that is,
where p represents the pressure produced in the principal directions.
In what follows, we will present several solid models, such as, their deformation
energies and principal stress obtained under different assumptions as hyperelasticity,
isotropy, incompressibility and, under uniaxial tension. For the uniaxial tension, we have
Thus, we obtain that
I1? λ2? 2λ−1,
I2? 2λ ? λ−2.
In what follows, we will describe some different hyperelastic models used for brain tissue
6 Journal of Applied Mathematics
2.2.1. Neo-Hooke Material Model
In this case, the deformation energy model is given by
W ? C1?I3− 3?,
where C1is a given constant. The stress in the direction of the principal stretch is
σ ? C1
2.2.2. Mooney-Rivlin Material Model
In this case, the deformation energy model, with 5 parameters, is the following:
w ? C1?I1− 3? ? C2?I2− 3? ? C3?I1− 3?2? C4?I1− 3??I2− 3? ? C5?I2− 3?2,
where C1to C5are material constants. In this case, the stress direction for the principal stretch
σ ? C1
2.2.3. Odgen Material Model
In this case, the deformation energy is given by
where αkand μkare constants of the material. The stress direction of the principal stretch is
From the Cauchy tensor, it is possible to compute the equivalent stresses. The
equivalent stress. is computed using the Von Misses formula:
?σ1− σ2?2? ?σ2− σ3?2? ?σ1− σ3?2
where σ1, σ2, and σ3are the principal stresses.
Journal of Applied Mathematics7
The equivalent strain is defined as
√2?1 ? υeq
?2? ?εz− εx?2?3
where ε and γ are the components of the deformation tensor of Hencky.
3. Numerical Methods
3.1. Modeling of the Brain Shift
In order to make the numerical simulations of the brain shift, we will consider the experi-
mental data of Mehdizadeh et al. ?8?. In the experiments, the gray matter is obtained from
the parietal lobe and the white matter is obtained from the corpus callosum from a one-year-
old bovine. The tissue obtained corresponds to discs of 15mm diameter and 5mm of height.
The tests were realized with a uniform rate of deformation of 1mm/min in order to avoid
inertial forces. The used machine was a dynamic testing machine, Hct/25–400 with servo
hydraulic valve PID controller. The elastic modulus obtained was E ? 24.6kPa and ν ? 0.49.
For the Neo-Hooke model, the constant is C1? 7903Pa.
To study the hyperelastic solids, we use the data obtained for the gray matter. The
curve for the uniaxial traction for the gray matter is showed in Figure 1.
The cerebral cortex consists of gray matter, and this region is the most affected by brain
shift. Also, the ratio of the volume of grey matter to white matter in the cerebral hemisphere
for a 20-year-old man is 1.3 ?12?. The mechanical properties of gray and white matters
measured by Mehdizadeh et al. ?8? show differences for gray matter, true Young modulus
of 24.6kPa and, for the white matter, true Young modulus of 19kPa have been derived. On
the other hand, it is practically impossible to build a simple solid model considering the real
white and gray matter distribution in a human brain. Considering these reasons, we have
used the mechanical properties of gray matter for the complete brain model. Due to the facts
that the larger brain displacements are near brain surface and the brain cortex is composed
only principally of gray matter, the model predicts brain shift with acceptable precision.
The Odgen material model was studied considering the first, second, and third order.
The Mooney-Rivlin model was studied considering two of its forms, with two and five
parameters ?see Tables 1 and 2?.
3.2. CAD Geometry
To quantify accurately the deformations and stresses produced in the phenomenon of brain
shift during a brain craniotomy, the CAD model of brain is relevant. The CAD geometry
used in the present work is an approximation with characteristics similar to a real brain. We
modeled the characteristics of a healthy male brain of 35 years. The brain is approximately
a ball whose surface geometry is characterized by irregular folds, see Figure 2. In this area
circulate most of the blood vessels, veins, and arteries. The width of the brain is variable;
however, the average value is 140mm. The average length is 170mm. The height of the
brain varies with respect to the observed cross-section up to 120mm. Considering the
above measures as a reference and using MRI images of the brain, a CAD 3D is generated.
8 Journal of Applied Mathematics
Uniaxial test data
Stress (×103) (Pa)
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12
Figure 1: Stress-strain curves for gray matter.
Table 1: Odgen constants.
Table 2: Mooney-Rivlin constants.
MR 2 parameters ?Pa?
MR 5 parameters ?Pa?
3.85E ? 06
−8.71E ? 06
4.93E ? 06
The CAD software used is solid edge, as it provides the necessary tools to model complex
nonparameterized curves. The methodology is to build a hemisphere from the outer contours
of the brain. To obtain these, contours are sectioned in coronal three-dimensional models,
then the contours are drawn to generate the solid model. Once the model is built in, with the
option Mirror, the second hemisphere is created ensuring the model symmetry. The last step
is to use the option Swept Protrusion to create the final CAD model of the brain. The model
obtained is showed in Figure 3.
The cerebral cortex is characterized as a cortical layer with a convoluted topology,
Figure 2. This complex geometry is modeled as simple hemisphere as in all previously
reported investigations about brain shift, see Wittek et al. ?3, 4?. The model must be relatively
Journal of Applied Mathematics9
Figure 2: Coronal view of the brain.
Figure 3: Principal and Isometric views of the brain model.
simple to be used as predictive tool for the clinicians with a minimum error. The comparison
of model predictions with clinical results of brain shift ensures that the approximation of the
complex brain structure is correct for the goal of the present model. We will try to use this
model in brain surgery to predict brain shift after clinical validation in the future; for this
reason, the model must produce computational results in short time. Models that consider
the topology of cerebral cortex as highly convoluted sheet for investigation of the gray matter
deformation have been reported by Chung et al. ?13?. However, the model is too complex to
be applied during a surgery.
The skull is made similarly to the brain. To do this, we use the option Offset tool in the
CAD software. The goal is to keep the skull around the brain model but with a separation
between these elements. According to what was observed in the MRI images of the middle-
aged male patient, the gap between the elements was determined. Figure 4 shows the process
of generating the skull geometry.
10 Journal of Applied Mathematics
Figure 4: Design for the skull model.
Figure 5: Views of the brain and skull model.
Figure 5 shows a view of the brain and the skull used in the present investigation.
Although the models are a simplification of the real, it is important to note that they retain
In the present investigation, we do not consider the cerebrospinal fluid CSF and the
brain can be deformed in this space. in a brain craniotomy, CSF is extracted during surgery,
and, therefore, this model restriction has low effect on brain displacement. The subarachnoid
space between brain and skull is small compared with nominal brain diameter, also variation
of model distance between brain and skull was considered as second-order effect. The present
brain model intends to describe a methodology to predict brain shift.
3.3. Boundary Conditions
For the simulation of the brain shift effects, we consider two boundary conditions.
Fix of spinal cord: in order to limit movement of the brain and allow greater defor-
mations only in the area affected by the change in pressure.
Pressure variation in area of operation: intracranial pressure caused by brain, blood,
and CSF is approximately 770mmHg. The atmospheric pressure is 760mmHg. Upon opening
the skull, there is trickle of CSF and blood, mainly affecting the operation area and leaving it
exposed to atmospheric pressure. This condition results in a negative pressure in the opening
area equivalent to 10mmHg or 1333Pa.
Figure 6 shows the application of the pressure boundary conditions for the different
craniotomies. Figure 6 shows the skull with the brain inside, and indicated with red color the
region where the pressure boundary condition is applied in each case.
Journal of Applied Mathematics 11
J: craneotomia lateral
Time: 6 s
09-11-2010 13 : 07
Pressure: −1333 Pa
Image to file...
Time: 6 s
09-11-2010 12 : 56
Pressure: −1333 Pa
I: craneotomia delantera
Time: 6 s
09-11-2010 13 : 21
Pressure: −1333 Pa
K: craneotomia trasera
Figure 6: Boundary condition for parietal, frontal, and posterior craniotomy.
The region size to apply the pressure boundary condition in the three brain crani-
otomies was very difficult to choose. The first idea was to apply the pressure on the same
area for the three cases; however, personal communications with neuroradiologists from the
Instituto de Neurocirugia Asenjo, that help us in this project, indicated that the affected area
is different for the three investigated craniotomies. The red areas showed in Figure 6 show
the chosen areas to apply the pressure boundary condition. The areas were not the same for
the tree cases, because the goal was to try to reproduce clinical results.
For the brain shift, the small distance between skull and the brain is the most relevant
parameter that induces pressure differences on brain surface. During a craniotomy, CSF
moves outside the skull and CSF flow does not produces pressure difference on brain surface.
3.4. Computational Method
The model was solved by a commercial finite element package ANSYS v12.1. The finite
element method ?FEM? is used to solve the governing equations. The FEM discretizes the
computational domain into finite elements that are interconnected by element nodal points.
We have used the static structural formulation with a maximum time of 6s. Incompressible
material behavior may lead to some difficulties in numerical simulation, such as volumetric
locking, inaccuracy of solution, checkerboard pattern of stress distributions, or divergence.
We used the mixed u-P elements available in ANSYS to overcome these problems.
The unstructured grids were composed of tetrahedral SOLID187 with 10 nodes
available in ANSYS. Figure 7 shows the details of the three different grids used for the
parietal, posterior, and frontal craniotomies. For the parietal craniotomy, the grid was more
refined in the middle brain region. For the frontal craniotomy, the grid was refined nearer
than the frontal region of the brain.
12Journal of Applied Mathematics
Figure 7: Isometric and superior views of the different computational grids of the brain model used for ?a?
parietal, ?b? posterior, and ?c? frontal craniotomy.
Table 3: Comparison of maximum displacement, strain, and stress for three different grid sizes.
The three grids used are similar, and the variations of element size in the brain depend
also on model construction, see Figure 5.
Grid independence study was performed for three grid sizes; maximum displacement,
equivalent strain, and equivalent stress were compared in Table 3. For the comparison, we
have used the frontal craniotomy with the elastic brain. The differences between the results
are very small. Therefore, the middle grid size was used to perform all the computational
simulations. This test ensures that the grid density does not affect the expected results about
Journal of Applied Mathematics 13
Table 4: Comparison of maximum displacement and effective stress using seven brain models for a frontal
pressure boundary condition without skull.
Maximum displacement ?mm? 8.923
Maximum effective stress ?Pa?37045
4. Results and Discussion
The predictions of maximum displacement and effective stress of brain under a pressure
boundary condition similar to frontal craniotomy but without skull were compared for
elastic, neo-Hookean, Ogden first-order, Ogden second-order, Ogden third-order, Mooney-
Rivlin with two-parameter, and Mooney Rivlin with five-parameter models. The results are
showed in Table 4. The low-order neo-Hookean model, the first-order Ogden model, and the
Mooney-Rivlin model with two parameters predict similar displacement and effective stress.
however, the prediction of maximum effective stress is 66% lower than the prediction of the
second-order Ogden model.
The predictions of hyperelastic second-order Ogden and third-order Ogden models
are similar. The Mooney-Rivlin model with five parameters predicts lower brain displace-
ment compared with all models included in the elastic model and therefore is discarded. With
these considerations, the hyperelastic second-order Ogden model is selected as adequate
solid model for comparison with an elastic model for prediction of brain shift phenomena.
In this work, we have considered three brain craniotomies, a parietal, a posterior, and
a frontal case as showed in Figure 6, and the maximum displacement and effective stress
were investigated. The solid models of brain are the elastic and the hyperelastic second-order
Figure 8 shows in logarithmic scale the distribution of the effective stress for the
parietal craniotomy, and an isometric and an inferior views of the brain are showed. The
effective stress on brain surface shows large areas with values around 1000Pa. The maximum
effective stress is 53286Pa. The maximum is on the brain base near the spinal cord, where the
model is fixed.
Figure 9 shows in logarithmic scale the distribution of the effective stress for the
posterior craniotomy, and an isometric and an inferior views of the brain are showed. The
effective stress on brain surface shows large areas with values around 500Pa. The maximum
effectivestress isnow38073Pa.Themaximum isonthebrainbase nearthespinal cord,where
the model is fixed. Areas on the brain base are under relatively high stress compared with the
rest of brain.
Figure 10 shows in logarithmic scale the distribution of the effective stress for the
frontal craniotomy, and an isometric and a inferior views of the brain are showed. The
effective stress on brain surface shows large areas with values around 200Pa. The maximum
effective stress is now only 10049Pa. The maximum is on the brain base near the spinal cord,
where the model is fixed. Areas on the brain base are under relatively high stress compared
with the rest of brain.
A comparison between the three craniotomies shows that the parietal produces higher
effective stress on brain than the posterior and frontal interventions. High stress values are
distributed principally on brain base.
14 Journal of Applied Mathematics
Type: equivalent (von-Mises) stress
11-11-2010 14 : 34
J: craneotomia lateral
Type: equivalent (von-Mises) stress
11-11-2010 14 : 27
J: craneotomia lateral
Figure 8: Effective stress for parietal craniotomy, hyperelastic second-order Ogden model.
Comparing the maximum effective stress for the frontal craniotomy with the data of
Table 4 can be concluded that the effect of the skull is very important, and the stress for
the craniotomy is considerably lower as the value obtained only with a pressure boundary
condition. This indicates that modeling of brain shift must consider the skull to obtain more
realistic values. Ji et al. ?5? reported the relevance of brain-skull contact in the determination
of brain shift compensation.
The maximum effective stress is high compared with the values of the stress strain
curve showed in Figure 1; therefore, the use of hyperelastic model for the brain is relevant for
better prediction of brain shift.
The distribution of brain displacement for parietal craniotomy calculated with the
hyperelastic second-order Ogden model is showed in Figure 11. Figure 11 shows the
displacement of brain surface and the displacement in a plane through the craniotomy. The
upper surface shows displacements around 7mm. The brain area with large displacement is
important. Also, in the brain center, the displacements are around 3mm. Figure 11 shows the
relevance of brain shift for parietal craniotomy.
Figure 11 shows that a part of the brain is displaced out of the skull in the craniotomy
area. The zones with low displacement are near the spinal cord, due to the fix condition in
this area. From a neurological point of view, this result is realistic. For the brain, the zone
with maximum stress does not coincide with the location of maximum brain shift or brain
Journal of Applied Mathematics 15
Type: equivalent (von-Mises) stress
11-11-2010 15 : 01
K: craneotomia trasera
Type: equivalent (von-Mises) stress
11-11-2010 15 : 01
K: craneotomia trasera
Figure 9: Effective stress for posterior craniotomy, hyperelastic second-order Ogden model.
The distribution of brain displacement for the posterior craniotomy calculated with
the hyperelastic second-order Ogden model is showed in Figure 12. The figure shows the
displacement of brain surface and the displacement in a middle brain plane. The frontal
region shows displacements around 12mm. The brain area with large displacement is
important. Also, in the brain center, the displacements are around 5mm. Figure 12 shows
the relevance of brain shift for posterior craniotomy. The zones with low displacement are
near the spinal cord, due to the fix condition in this area.
Finally, the distribution of brain displacement for the frontal craniotomy calculated
with the hyperelastic second-order Ogden model is showed in Figure 13. The figure shows
the displacement of brain surface and the displacement in a middle brain plane. The
superior region shows displacements around 4mm. The brain area with large displacement
is important. Also, in the brain center, the displacements are around 2mm. Figure 13 shows
the relevance of brain shift for frontal craniotomy. The zones with low displacement are near
the spinal cord, due the fix condition in this area. For the brain, the zone with maximum
stress does not coincide with the location of maximum brain shift or brain displacement and