388 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO.
2, FEBRUARY 2008
A Model for Idiopathic Intracranial Hypertension and
Associated Pathological ICP Wave-Forms
Scott A. Stevens*, Jesse Stimpson, William D. Lakin, Nimish J. Thakore, and Paul L. Penar
Abstract—Idiopathic intracranial hypertension (IIH) is a syn-
drome of unknown cause characterized by elevated intracranial
pressure (ICP). While imaging often reveals a stenosis of the
transverse sinuses, the role of this feature in IIH has been in
dispute. Many patients with chronic daily headache have been
found to actually be suffering from a milder form of IIH without
papilledema (IIHWOP). These patients often demonstrate hy-
pertensive B-waves and plateau-like waves upon continuous ICP
monitoring. Recently, we presented modeling studies which suggest
that the sinus stenosis and hypertension of IIH are physiological
manifestations of a stable state of elevated pressures that exists
when the transverse sinus is sufficiently collapsible. Many of the
features of IIH were explained by this model but the prevalence
of pathological ICP wave-forms observed in IIHWOP remained
unresolved. The model presented here is a modified version of
a previous model with a semi-collapsible sinus represented by a
refined downstream Starling-like resistor based on experimental
data. The qualitative behavior of this model is presented in terms
of the collapsibility of the transverse sinus. For a sufficiently rigid
sinus, there is a unique stable state of normal pressures. As the
degree of collapsibility increases, there is a Hopf bifurcation, the
normal state becomes unstable, low-frequency, high-amplitude
ICP waves prevail, and small perturbations can lead to hyper-
tensive ICP spikes. As the collapsibility increases further, so does
the duration of the waves, until they are replaced by two stable
states: one of normal pressures and one of elevated pressures. In
this parameter domain, temporary perturbations can now cause
permanent transitions between states. The model presented here
retains the capability of our previous model to elucidate many fea-
tures of IIH and additionally provides insight into the prevalence of
the low-frequency, high-amplitude waves observed in IIHWOP.
Index Terms—B-waves, idiopathic intracranial hypertension
(IIH), intracranial pressure dynamics, mathematical model,
plateau waves.
I. INTRODUCTION
DIOPATHIC intracranial hypertension (IIH), also called
I pseudotumor cerebri and benign intracranial hypertension,
is a syndrome of unknown cause characterized by elevated
Manuscript received December 7, 2006; revised April 24, 2007. Asterisk in-
dicates corresponding author.
*S. A. Stevens was with the School of Science, Penn State Erie, Behrend
College, Erie, PA 16563-0203 USA. He is now with the Division of Informa-
tion Technology and Sciences, Champlain College, Burlington, VT 05402 USA
(e-mail: stevens@champlain.edu).
J. Stimpson is with Penn State Erie, Behrend College, Erie, PA 16563-0203
USA.
W. D. Lakin is with the the Department of Mathematics and Statistics and
the Biomedical Engineering Program, University of Vermont, Burlington, VT
05401 USA.
N. J. Thakore is with Metrohealth, Department of Neurology, Case Western
Reserve University, Cleveland, OH 44106 USA.
P. L. Penar is with the Department of Surgery, Division of Neurosurgery,
College of Medicine, University of Vermont, Burlington, VT 05401 USA.
Digital Object Identifier 10.1109/TBME.2007.900552
0018-9294/$25.00 © 2008 IEEE
intracranial pressure (ICP) without evidence of ventricular di-
latation, mass lesion, cerebrospinal fluid (CSF) abnormality, or
dural sinus thrombosis. It presents with symptoms of headache,
nausea, vomiting, papilledema, and visual obscurations [1].
In many patients suffering from IIH, a stenosis or tapering
of the transverse sinuses is observed by magnetic resonance
venography (MRV) or retrograde catheter venography [2]–[4].
The role played by transverse sinus stenosis in the etiology of
IIH has been a matter of some dispute.
There are also indications that IIH, in a subclinical form
without papilledema (IIHWOP), may be more common than
previously realized. A study by Bono et al. [5] revealed that
6.7% of 724 migraine patients had bilateral transverse sinus
stenosis and 67.8% of these were diagnosed with IIHWOP. A
study by Torbey et al. [6] also suggests that many patients with
chronic daily headache (CDH) may actually be suffering from
IIHWOP. Furthermore, of the ten patients examined in the
Torbey study, all ten exhibited hypertensive B-waves and nine
of the ten exhibited plateau or “near-plateau” waves during
continuous CSF pressure monitoring.
Recently, we reported [7], [8] the results of modeling studies
that suggest the observed sinus stenosis and persistent hyper-
tension of IIH may be a physiological manifestation of a stable
steady-state of elevated pressures predicted by the studies to
exist when the transverse sinus is allowed to be sufficiently col-
lapsible rather than rigid. In these studies, a transition to an el-
evated steady-state took place by means of a saddle-node bi-
furcation in the parameter domain. These results provide ex-
planations for the following phenomenon associated with IIH:
1) the observed sinus stenosis [2]–[4], 2) intermittent occur-
rence of symptoms [5], [9], [10], 3) large pressure gradients
observed across the transverse sinus [4], 4) the long-term re-
lief that has been observed following lumbar puncture [9]–[11],
and 5) the possible persistence of sinus stenosis after the hy-
pertension has been resolved via pharmaceutical treatment and
weight loss [12]. These simulations do not, however, predict the
prevalence of hypertensive ICP spikes (B-waves) or plateaus ob-
served by Torbey in IIHWOP patients. The spontaneous gener-
ation of hypertensive ICP plateau waves due to cerebral blood
flow autoregulation in association with pathologies such as ele-
vated ICP, low CSF compliance, and increased CSF resistance
to absorption has been well documented [13]–[15] and modeled
[16], [17]. However, the mechanism for generation of similar
waveforms in conditions such as IIHWOP is poorly understood.
In many mathematical models of intracranial blood flow, a
Starling resistor is assumed to exist at the location of a col-
lapsible vein [28]–[31] where the normal transmural pressure
is positive. For such a traditional Starling resistor, if transmural
pressure at the location becomes zero or negative, the vessel is
STEVENS et al.: MODEL FOR IDIOPATHIC INTRACRANIAL HYPERTENSION AND ASSOCIATED PATHOLOGICAL ICP WAVE-FORMS 389

assumed to be fully collapsed and all flow is occluded. A tra-

ditional Starling resistor thus cannot be placed at the location
of the transverse sinus where the normal transmural pressure
is negative. Consequently, these vessels have traditionally been
considered rigid and modeled by assuming a constant resistance
to flow. In healthy individuals, this is certainly a valid assump-
tion. However, a stenosis of these vessels has been observed in
IIH patients [2]–[5]. This suggests that these sinus vessels are
not rigid in patients who have, or are at risk for, IIH. Repre-
senting a nonrigid transverse sinus in a mathematical model of
this phenomenon thus requires the development and use of a
nonstandard form of a Starling resistor.
The lumped-parameter model presented here is a refinement
of the model developed by Stevens et al. in [8]. This pre-
vious modeling effort introduced a “downstream” Starling-like
resistor at the location of the transverse sinus [7], [8]. The rep-
resentation for this resistor involved a fluidity expression that
was linearly increasing with downstream transmural pressure.
Thus, flow resistance, which is the inverse of fluidity, while
not strictly linear, was decreasing with downstream transmural
pressure. A variable maximum level of vessel collapse was
also included to accommodate alternative drainage routes and a Fig. 1. Diagram of the lumped-parameter model. Compartment labels are in-
dicated in parentheses. The dark line represents the rigid cranial wall and thin
geometry that may prevent complete collapse. The form of this lines separate compartments. Q represents fluid flow from compartment i to
earlier nonstandard resistor was loosely based on the results compartment j as determined by (1). Arrows indicate the customary direction
of flow. The normal resting pressure of each compartment is represented by
of Heil [18], in which the lumped-resistance to constant flow the number in square brackets. Compliance terms C represent a deformable
through a collapsible tube was shown to decrease with respect membrane between compartments i and j across which volume changes may
to the downstream transmural pressure. The principal differ- be accommodated through (3). The pressures and flows depicted here are from
Stevens et al. [8].
ence between the current model and the previous model is that
the description of the downstream resistor has been modified
to more closely reflect the experimental data provided by Heil
are represented by compartment . Blood then drains through
[18]. This resistor is placed at the location of the transverse
the transverse sinuses and into the jugular veins in compartment
sinuses in the model where the stenosis has been observed
. Brain tissue and the associated intra- and extracellular fluid
clinically. With this modified downstream resistor in place, the
are combined with the CSF to form compartment . The present
model demonstrates the existence of various self-excited ICP
model also contains one extra-cranial compartment, namely the
waveforms not present in our previous modeling studies. The
thoracic space, which is represented by compartment .
present results thus provide insight into the prevalence of the
Pressure in each model compartment is given in mmHg and
low-frequency, high-amplitude waves observed in IIHWOP.
denoted by . For example, will denote the spatially aver-
aged pressure in the brain/CSF compartment. Fluid flow or fil-
II. MODEL tration from compartment to compartment is given in ml/min
The lumped-parameter model utilized in this study is shown and denoted by . For example, represents the flow of
in Fig. 1. This model is similar to the model used by Stevens et blood from the capillaries into the veins. Likewise, rep-
al. [8]. However, for the present study, the representation of the resents fluid filtration from the capillaries and choroid plexuses
Starling-like resistor in the model has been chosen to emulate into the brain and CSF compartment.
the relationship between downstream transmural pressure and The term “intracranial pressure” (ICP) is used clinically to
lumped resistance through a collapsible tube as experimentally describe both the CSF and brain pressures, which are nearly
and computationally determined by Heil [18]. The details of this equal in healthy physiology [20]. ICP will thus be used here to
refinement are presented in Assumption 4 later on. refer to the CSF/brain pressure from both the clinical and current
Only the main components of the lumped-parameter model modeling perspective.
will be described here as a more detailed description of the
model, the parameter identification process, and validation sim- A. Model Assumptions
ulations is available in Stevens et al. [8]. Consistent with math- The following basic assumptions lead to the fully time-depen-
ematical models of this type [17], [19], [20], the intracranial re- dent differential equations that govern the pressure dynamics of
gion is divided by constituent into interacting subunits termed this system:
“compartments.” For this study, the intracranial region is di- Assumption 1: All fluids are considered incompressible and
vided into the four compartments as depicted in Fig. 1. Com- isothermal.
partment consists of the intracranial arteries and capillaries. Assumption 2: The regulation of cerebral blood flow
This vasculature leads into the veins and saggital sinus which and CSF production by the choroid plexus over a full
390 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 2, FEBRUARY 2008

Fig. 2. Panel A: The relationships between transmural pressures and the pressure drop through a collapsible tube with constant flow as presented by Heil [18].
Here, since flow is constant, the pressure drop depicted on the abscissa is a multiple of the lumped-resistance defined by R = (pressure drop=flow). Panel B: A
graph of the resistance term R defined by (2) with m = :4; p = 0:7, and q = 7. Here, R is actually a function of P . The abscissa and ordinate axes
have been switched to facilitate comparison to the downstream branches of the Heil graphs. A maximum resistance (q R  ) represented by the second vertical
portion is imposed to accommodate alternative drainage routes and/or a sinus geometry that may prevent complete collapse [9].

range of intracranial pressures is described by [21]. For the where

pressure ranges considered here, these regulatory mechanisms
remain robust. Hence, for simplicity, constant rates will be as-
sumed for both and . An exception to this assump-
tion is made for simulations of cerebral blood flow disturbances
where is utilized as a time-varying forcing term.
Assumption 3: All other flows ( , and ) are Here, and represent the normal resting values of
related to pressure differences by the hydrodynamic version of the pressure difference and the resistance , respec-
Ohm’s law: tively. The form of this resistor is based on the relationship be-
tween downstream transmural pressure and the pressure drop
(1) for constant fluid flow through a collapsible tube reported by
Heil [18]. The data of Heil and a graph of the present are
depicted in Fig. 2.
where is the flow from compartment to compartment
There are three relevant parameters in the present Starling-
and are the spatially averaged pressures of compartments
like resistance term. First, describes the initial collapsi-
and , respectively, , and is the lumped
bility of the vessel. Second, defines the minimal resistance,
resistance.
, when the vessel is in its maximally open
Assumption 4: A partially collapsible transverse sinus will be
state. Third, defines the maximal resistance, ,
allowed in this study by introducing a downstream Starling-like
when the vessel is in its maximally collapsed state. This max-
resistor into the model between compartments and . Sim-
imum, short of total collapse, is imposed to accommodate
ilar Starling-like resistors have been used previously in lumped-
alternative drainage routes and/or a geometry that may prevent
parameter models to help explain aspects of the etiology, diag-
complete collapse [9]. If the sinus is fully rigid, the parameters
nosis, and treatment of IIH [7], [8]. As in these previous works,
in (2) take the values and . In this case, the
the resistance term in this modeling effort will be an increasing
resistance is constant at .
function of the opposite of the downstream transmural pressure
Assumption 5: The deformation of the membrane between
as described by Lakin et al. [22]. However, un-
adjacent compartments is a function of the change in the pres-
like previous versions of this resistor, the representation here is
sure difference between these compartments. That is
linearly dependent on downstream transmural pressure and in-
volves both a maximum and minimum resistance value. Specif-
ically, this resistance function is defined by (3)

if where denotes the volume of the cup formed at the interface

if of compartments and , and represents the time derivative
if of the pressure difference . Here, denotes the local
(2) compliance [23] between the two compartments.
STEVENS et al.: MODEL FOR IDIOPATHIC INTRACRANIAL HYPERTENSION AND ASSOCIATED PATHOLOGICAL ICP WAVE-FORMS 391

B. Governing Equations (12)

The model’s governing differential equations are obtained by
The aforementioned parameters are given numerical values
imposing conservation of mass in each compartment. As the
as follows:
fluid is considered incompressible, this requirement is equiva-
lent to a conservation of volume equation of the form

flow rate in - flow rate out = rate of volume change.

(4)
Imposing this constraint in compartments and results in
the equations
(5)
and
(6)
respectively. Here, the right-hand sides denote the change in
volume with respect to time. Applying Assumptions 2, 3, and
5, the model’s governing equations now become
(7)
(8)
C. Parameter Identification
Before analyzing changes from a base state due to various
stimuli, it is necessary to characterize this base state in terms
of initial values for the variables and parameters involved in the
model. These starting values have been obtained from available
clinical data and will be referred to throughout this study as base
values.
1) Base Values for Pressures, Flows, and Resistances: Base
values for resistances are determined by solving (1) for
given prescribed base pressures and flows. That is
(9)
where the over-bar indicates the normal base-value for a specific
variable or parameter. The values for those terms on the right-
hand side are given in Fig. 1. The justification for these values
are detailed in Stevens et al. [8], [20].
2) Compliances: The compliances defined in (3) are exten-
sions of the pressure dependent compliance functions previ-
ously developed by Stevens and Lakin [23]. Minor modifica-
tions of the previous compliance terms are used in the current
model. See Stevens et al.[8] for a detailed description of this
modification process and validation trials. For the sake of com-
pleteness, the compliances which result from this process are
given by
D. Model Validation
Two sets of simulations were performed and compared to
measured clinical data to validate the current model. In both
sets, healthy physiology was assumed, and thus the value of
the Starling resistor was considered fixed at its baseline
value of . Under this assumption, the previous model [8]
and the current model are identical. Detailed results of these val-
idation simulations are presented by Stevens et al. [8] and will
not be duplicated here. However, it should be pointed out that
the model accurately portrays clinical observations regarding
the conductance of CSF outflow defined by
the change in CSF absorption
the change in CSF pressure
and the global CSF pressure-volume relationship over the full
spectrum of feasible CSF pressures.
E. Steady-State Equations
The first step in a steady-state analysis of the current model
is to set the time derivative terms in the governing equations (7)
and (8) to zero. The resulting system of algebraic equations for
the steady-state now becomes
(13)
(14)
respectively.
As described in the parameter identification section above,
the resistance terms are calculated in such a way that the normal
pressures and will satisfy the steady-state
equations. Hence, the ordered pair will be referred to
as the base value steady-state solution of the system described
by (13) and (14). It is convenient at this point to perform a trans-
formation of variables
(15)
(16)
so that the origin (0,0) now corresponds to the base value steady-
state, and represents changes from the base value state.
Applying this transformation, and assuming that and
remain constant at their respective base values, (13) and (14)
become

(10)
(11) (17)
392 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 2, FEBRUARY 2008

A. Steady-State Solutions and Stability

Expanding (17) and (18), these become
(18)

In terms of and , the downstream Starling resistor in (2) (24)

now has the representation
(25)
if
if (19) Since the resistance terms and are derived by (9),
if this implies and .
Substituting these into (24), and (25) yields
Algebraic solutions of (17)–(18) describe the steady-state so-
lutions of the time dependent differential equations (7) and (8). (26)

F. Stability Analysis
(27)
When the fully time dependent system of differential (7) and
(8) are expressed in terms of the transformed variables and , Further, in the steady-state when compartmental volumes are
the result can be put into matrix form as constant, and .
Imposing these constraints in (26) and (27) now produces
(20)
(28)
where the compliance matrix is given by
(29)
(21)
Solving the first of these for results in
Here, the functions and are defined in the
steady-state equations (17) and (18). Note that since all of the (30)
local compliances are positive, the compliance matrix is nonsin-
gular, and system (20) can be expressed explicitly as and on this line . Since this corresponds to
, this null-cline will be denoted by and defined by
(22)

where
The stability properties of various steady-state solutions can
now be determined from the eigenvalues of the Jacobian matrix (31)
of partial derivatives, defined by
Solving (29) for and using a similar notation produces
(23)
(32)
where and are defined in (22). The process of determining
stability properties from the eigenvalues of the Jacobian matrix and along this curve . Unlike the expres-
is described in most texts on differential equations and nonlinear sion for , the above relationship is not linear because of the de-
dynamics. See, for example, [24]. pendence of on described in (19). Solving for the points
of intersection of the two null-clines; and , produces the
steady-state solutions. Notice that if , then
III. RESULTS
and the numerator in (32) is zero. Likewise, when
The results presented here are based on explicitly solving (17) . Therefore, the the base value state (0,0) is seen to al-
and (18) for all steady-state solution pairs using the rep- ways produce a solution to the steady-state equations. The al-
resentation (19) for the Starling-like resistor . The algebra gebra involved in finding the other steady-state solutions, the
involved in solving these equations is straightforward but very domains of validity, and the stability properties is straightfor-
tedious and time consuming and will not be described in detail. ward but tedious and time-consuming. Therefore, these details
Additionally, the inherent complexity of the Jacobian matrix in will be omitted.
(23) precludes explicit derivation of its eigenvalues in terms of As seen in Fig. 3, the value of determines the stability
the model parameters. Therefore, these eigenvalues will be de- of the base-value state (0,0) and whether there are additional
termined for numerical values assigned to these parameters. As steady-states present. In terms of the steady-state solution
such, the stability and bifurcation classifications described here there is a unique stable steady state for .
are based on these numerical results. Consequently, for values of less than , there is only the
STEVENS et al.: MODEL FOR IDIOPATHIC INTRACRANIAL HYPERTENSION AND ASSOCIATED PATHOLOGICAL ICP WAVE-FORMS 393

x m
Fig. 3. Bifurcation diagram of the steady-state solution with respect to the rigidity parameter . When m<m x
, the horizontal line at = 0 represents the
unique, globally stable, base-value state where P P
=  . A Hopf bifurcation occurs at m x
resulting in a stable limit cycle of pressures. The pulse pressure of
m
these cycles increases quickly in to the limiting value of x 0x . Atm , a saddle-node bifurcation occurs producing a stable hypotensive state. Soon after,
atm x
, a nontraditional bifurcation occurs and a stable hypertensive state is introduced. The level of the hypotensive state ( ) and the stable hypertensive state
x p q
( ) are determined by the rigidity parameters and , respectively. These are described by (33) and (34).

TABLE I
SUMMARY OF THE BIFURCATION DIAGRAM IN FIG. 3

base-value state, and it is globally stable. A Hopf bifurcation
occurs at resulting in a stable limit cycle of pressures. The
pulse pressure (max–min) of these cycles increases quickly with
to a limiting value of . At , a saddle-node bifurca-
tion occurs producing a stable hypotensive state. Soon after, at
, a nontraditional bifurcation occurs and a stable hyperten-
sive state is introduced. For the stable hypotensive and
hypertensive states persist. Table I gives a list of the relevant
terms, their approximate numeric values, and a brief description
of the stability changes that occur at each level. Simulations in
reflective parameter domains have been performed, and the pre-
dicted results from these simulations are presented next.
The bifurcation diagram presented in Fig. 3 is based solely
on variations in the collapsibility parameter . All other model
parameters were assigned numerical values. It should be noted
that changing the parameters involved in (10)–(12) for com-
partmental compliances can alter the type of bifurcations and
the stability of fixed points. However, there is no change in the
number of fixed points as these are determined by the algebraic
equations where all derivatives are identically zero.
B. Time-Dependent Simulations
In the healthy state when the rigidity parameter falls below
, all solutions of the governing differential equations tend to
the base-state of normal pressures, regardless of initial condi-
tions or transient perturbations. Parameter domains of interest
for IIH are thus those for which . When , there
is the possibility of self-excited ICP waves and transient pertur-
bations that can lead to permanent transitions between states.
Therefore, this will be the focus of the time-dependent simula-
tions.
The system of differential equations can be numerically
solved in terms of either and from (7) and (8) or in
terms of and from (22). These systems are equivalent under
the transformation and . However,
results of the time-dependent simulations will be displayed
here in terms of and for the sake of comparison to actual
CSF and sinus pressures, respectively. The results will further
be organized according to the value of the rigidity parameter
and the type of solutions indicated by the bifurcation diagram
in Fig. 3. As in the steady-state analysis, will be maintained
at its base-state value of .
The numerical integration of the time-dependent equations
was performed independently with the software packages
Mathematica (Wolfram Research, Inc., Mathematica, Version
5.2, Champaign, IL, 2005) and MATLAB (The Mathworks,
Inc., MATLAB, Version 7.2, 2006). In both cases, accuracy
and precision settings were set to a maximum. The time-de-
pendent simulations faithfully reproduce expected results from
394
Fig. 4. This sequence of simulations demonstrates that as m increases from 0.445 (upper left) to 0.47 (lower right) the self-excited wave frequency diminishes
and the plateau period increases. The first graph demonstrates waves that are similar to clinically observed hypertensive B-waves and the later graphs are similar in
wavelength and amplitude to clinically observed A-waves. In all four cases, the rigidity parameters and were given numerical values of 0.7 and 7, respectively.
the steady-state analysis, and it is therefore unlikely that the
instabilities predicted here are due to the numerical integration
process. Rather, they appear to accurately reflect instabilities of
interacting physiological mechanisms included in the model.
1) Low-Frequency Self-Excited Waves—Limit Cycles: When
the rigidity parameter falls between and , there are
no stable steady-state solutions but rather stable limit cycles
of relatively low-frequency pressure oscillations. Fig. 4 depicts
several examples of these self-excited oscillations. These pul-
sations progress from high-frequency spike-like waves when
first crosses (panels A and B), to less frequent plateau-like
waves of increasing period (panels C and D).
2) Cerebral Blood Flow Disturbances: When the rigidity
parameter falls between and and there are no stable
steady-state solutions, the system is very sensitive to pressure
and flow disturbances that may naturally occur during normal
physiology. This is demonstrated by introducing a cerebral
blood flow (CBF) “spike” into the model. The duration of this
spike is 6 s and its magnitude is 10% of the normal level of
CBF. This spike is generated as the first half of a sine wave,
so the flow is continuous, but not differentiable, at the
beginning and end of the spike.
In simulations of normal physiology, where the sinuses are
rigid and , the CBF spike described above produces ap-
proximately a 0.2 mmHg spike in ICP, a response that would
go unnoticed physiologically and clinically. In fact, this spike
is similar in duration and amplitude to the B-waves observed in
healthy individuals [25], [26]. Since the ICP response was so
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 2, FEBRUARY 2008
p q
trivial in these simulations of healthy physiology, graphical de-
pictions of these responses will not be presented here.
Conversely, in simulations of pathophysiology where the
sinus is semi-collapsible with , this small
CBF disturbance can cause a much more drastic ICP spike
or a premature transition to a plateau wave. These responses
are depicted in Fig. 5 with . The low-frequency
plateau waves are caused by the semi-collapsible sinus (see
panel B in Fig. 4). When a mild CBF spike occurs shortly
after a completed wave, a single hypertensive ICP spike results
(panel A of Fig. 4). When this same stimulus occurs closer
to the beginning of the next plateau wave, it causes a plateau
wave to occur sooner than it otherwise would. Therefore,
when , it is quite likely that hypertensive
B-waves and plateau-like waves that could be missed during a
single lumbar pressure measurement would be observed during
long-term ICP monitoring.
3) Two Stable States: Hypotensive and Hypertensive: When
, there are steady-state solutions in addition to (0,0).
The -values of these states are found to be
for (33)
and
if
if
(34)
STEVENS et al.: MODEL FOR IDIOPATHIC INTRACRANIAL HYPERTENSION AND ASSOCIATED PATHOLOGICAL ICP WAVE-FORMS
Fig. 5. Six-second 10% increase in cerebral blood flow (Q ) initiates a hypertensive B-wave (left) when introduced shortly after a plateau wave and initiates
an early plateau wave (right) when introduced later in the cycle. Here, p = 0:7; q = 7, and m = 0:445 as in panel A of Fig. 4.
and the associated -values are found by multiplying these terms
by . Here, is the stable steady-state represented by the solid
horizontal line in Fig. 3 below . The first of the expres-
sions for in (34) is represented by the nearly vertical dashed
curve depicted in Fig. 3. The value of this unstable steady-state
starts hypotensive (below 0) but quickly jumps to the constant,
hypertensive stable steady-state given by the second expression
in (34), represented by the upper solid line in Fig. 3. It should be
observed that the magnitude of the stable elevated state is deter-
mined by the maximal collapsibility parameter . This demon-
strates the importance of this feature in the model. If the resis-
tance to blood flow drainage is allowed to get indefinitely large
, then the stable hypertensive state will likewise be-
come unbounded.
As seen in Table I, there is a very small range of values for
which the hypotensive state is stable and the hypertensive state is
unstable. As such, we will focus on the case where and
there are two stable steady-states. Additionally, the hypotensive
state given by is governed by the minimum resistance
parameter and has a lower bound of
when . Numerically, this is approximately mmHg.
As such, this hypotensive state could well go unnoticed phys-
iologically because the actual ICP would only be 2 mmHg
below normal. When , the numerical value of is
about mmHg, which makes the ICP at this value about
10.4 mmHg. Again, clinically this would not be considered hy-
potensive. It is defined as the “hypotensive state” here because
mathematically ICP falls below the base-value state.
When and there is a stable hypertensive state, var-
ious stimuli can trigger permanent transitions to it and from it.
Two such transitions will be demonstrated here. The first, de-
picted in Fig. 6, displays the predicted CSF and sinus pressure
responses to a temporary increase in cerebral blood flow. The
simulation starts from the stable hypotensive state and at
minutes a 20% increase in cerebral blood flow is instigated that
lasts for 30 s. As seen in Fig. 6, this causes a rapid jump in both
pressures and an eventual permanent transition to the hyperten-
sive steady-state. It should be observed here that the transition
395
Fig. 6. A 30–second, 20% increase in cerebral blood flow at t = 2 minutes
initiates a permanent transition from the hypotensive to the hypertensive state.
For this simulation, p = 0:7; q = 7, and m = 0:5. The dashed line represents
the sinus pressure (P ) and the solid curve represents the CSF pressure (P ).
Notice, the CSF pressure overshoots the steady-state and slowly comes back
down to it. The sinus pressure response displays a nearly instantaneous transition
to the hypertensive steady-state.
of the sinus pressure occurs almost immediately while the CSF
pressure requires some time to settle back to the steady-state.
The second simulation of a transition between steady-states
begins at the elevated state, and a temporary CSF withdrawal
initiates a permanent transition to the hypotensive state. As in
the previous simulation, this transition could be instigated with a
temporary cerebral blood flow decrease. However, a CSF with-
drawal was chosen here because it illustrates a possible diag-
nostic technique. To perform this simulation, a CSF withdrawal
term is subtracted from the left side of (7). For this example, the
withdrawal term is defined to begin at , drain fluid at a
rate of 3 ml/min, and terminate after 3 min, thus draining a total
of 9 ml of fluid from the CSF compartment. The results of this
simulation are depicted in Panel A of Fig. 7. Here, as should be
expected, there is a rapid drop in both pressures and an even-
tual permanent transition to the hypotensive state. Again, the
sinus pressure response is much faster than the CSF pressure
response.
For the sake of comparison, panel B of Fig. 7 demonstrates
the same withdrawal simulation performed on the model when
the sinus is completely rigid and the initial CSF
396
Fig. 7. At t = 10, a CSF withdrawal is simulated at a rate of 3 ml/min for 3 min. The solid curve is CSF pressure (P ) and the dashed curve is sinus pressure
(P ). Left: The initial state is the elevated state associated with a maximally collapsed sinus. Here, m = 0:5; p = 0:7, and q = 7. Right: The sinus is totally rigid
(m = 0) and the initial elevated state is due to impaired CSF absorption.
elevated state is due to a simulated blockage in CSF absorption.
In this case, the values for and are increased to those
values which result in a CSF pressure equal to the hypertensive
state in the previous simulation. Here, the rapid decline in CSF
pressure is followed by a logistic return to the initial state,
in agreement with the results produced in the classic work of
Marmarou et al. [27].
IV. SUMMARY AND CONCLUSION
The present refinement in the way a compressible transverse
sinus is represented in a mathematical model of lumped-param-
eter type expands our previous ability to explain many of the
various phenomenon associated with IIH and IIHWOP. These
included sinus stenosis, intermittent occurrence of symptoms,
large pressure gradients across the transverse sinuses, and long-
term relief after a single CSF withdrawal. In addition, the cur-
rent model provides new insight into the existence of low-fre-
quency high-amplitude ICP waveforms observed by Torbey [6]
in chronic daily headache patients diagnosed with IIHWOP.
In the current modeling effort, additional realism is intro-
duced by basing the form of the downstream Starling-like re-
sistor more closely on the results presented by Heil [18]. Specif-
ically, the Starling-like resistor is formulated directly in terms of
resistance rather than fluidity. This resistance term now remains
constant until a certain buckling pressure is reached. Beyond
this point, it becomes linearly decreasing with respect to down-
stream transmural pressure until the maximum level of vessel
collapse is reached. At this point, in common with the previous
representation, resistance is maintained as a constant at its max-
imum value.
In the representation (2) of the refined Starling-like resistor,
the parameter describes the initial collapsibility of the vessel
at the normal state. If the transverse sinus is fully rigid,
, and the flow resistance parameter is constant at . with respect to a diagnosis of IIHWOP in migraine patients.
In this case, the governing model equations, as calibrated in
Section II-C, have exactly one asymptotically stable steady-state
solution of normal pressures and simulations reproduce mea-
sured clinical results for both the conductance of CSF outflow
and the bulk CSF pressure-volume relationship. Analysis of the
governing equations provides a sequence of six critical values
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 2, FEBRUARY 2008
of that describe the behavior of the dynamical system. Three
of these critical values, ordered , charac-
terize the predicted behavior of the physiological system when
the transverse sinus is not required to be rigid.
Solution of the model equations indicates that full rigidity of
the transverse sinus is not required to preclude the appearance
of pathological features. For , the unique stable steady-
state of normal pressures persists until reaches the first critical
value . At this point, there is a Hopf bifurcation, and the
normal pressure state becomes unstable. As increases further
into the range , self-excited oscillations are
present in the form of low-frequency, large-amplitude waves.
Simulations indicate that the duration of a single wave can vary
between 2 and 15 min. In fact, as tends to the second critical
value the duration of these waves increases, and there is no
finite upper-bound on the duration of these plateau-like waves.
The results of Torbey et al. [6] demonstrate both short ICP
spikes and long plateau hypertensive waves in nine of the ten
IIHWOP patients studied. Model simulations reproduce this
observation in the range when a perturbation
is introduced via one of the forcing terms. When a small
disturbance in cerebral blood flow similar to that observed in
healthy individuals [25], [26] is introduced into the model, this
perturbation initiates either an isolated short-term hypertensive
ICP spike, or a premature transition to a full-length plateau-like
wave. The model simulations therefore suggest that with a
sufficiently collapsible sinus and the presence of even a small
amount of variation, plateau-like waves can be expected to co-
incide with intermittent hypertensive spikes (B-waves), which
will not be observed to occur in a truly periodic fashion. This is
in agreement with the Torbey’s observations.
The intermittent nature of these hypertensive spikes and
waves when has potential clinical implications
For example, in the data presented by Bono et al. [5], the
percentage of his migraine patients who displayed bilateral
transverse sinus stenosis upon MRV, as well as the percent of
these patients diagnosed with IIHWOP upon lumbar puncture,
may somewhat underestimate the actual values for patients in
his study. Since model simulations predict that instances of
STEVENS et al.: MODEL FOR IDIOPATHIC INTRACRANIAL HYPERTENSION AND ASSOCIATED PATHOLOGICAL ICP WAVE-FORMS
hypertensive CSF pressure and maximal sinus collapse occur
only intermittently, a transverse sinus stenosis or evidence of
IIHWOP may not have been present during the observation
period.
As the degree of sinus collapsibility increases and crosses
, the self-excited oscillations vanish in the narrow range
, and there is again only a unique stable
steady-state solution. The pressure in this state is slightly below
normal and would not be clinically recognized as hypotensive.
It is termed a “hypotensive” state here for the sake of differenti-
ating it from the normal and the hypertensive states. Under these
conditions, all transient perturbations eventually decay, and all
solutions tend to the hypotensive state. However, small distur-
bances in cerebral blood flow will generate long-duration hyper-
tensive plateau wave. For this reason, and because the range of
associated with a unique hypotensive steady state is so small,
this state is unlikely to be clinically observed. However, from
a dynamical systems perspective, this range of does play a
critical role in that it provides a transition between two very dif-
ferent qualitative states of the system. For , there are
self-excited oscillations, and for there are permanent
transitions to one of two stable steady-states.
At the final stage of sinus collapsibility , a new stable
hypertensive steady-state is introduced in addition to the stable
hypotensive state described above. The qualitative behavior of
the system in this parameter domain thus involves one stable
steady-state of elevated pressures and one of near-normal pres-
sures. Furthermore, perturbations or transient stimuli can cause
rapid and permanent transitions between these two stable states.
The ability of an increase in cerebral blood flow to cause a
rapid and permanent transition to the elevated pressure state
when provides one possible link between obesity and
IIH. Acute and chronic hypercapnia are prevalent in obese pa-
tients with obstructive sleep apnea [33], [34], and it has been
observed that cerebral blood flow velocity can increase by 19
to 219 percent during apneic episodes [35]. Since sleep apnea
is associated with obese patients [36], it may be that even if the
healthy state is achieved in these patients it will not persist be-
cause apneic episodes provide an instigating event for a perma-
nent transition to the elevated state.
Conversely, the current simulations indicate that an appro-
priate perturbation can also cause a transition from the elevated
to the normal pressure state when , and this result
has clinical implications for diagnosing IIH as the cause of cer-
tain cases of intracranial hypertension. If a CSF withdrawal pro-
duces a rapid and permanent transition to a new state, then such
a result would suggest the existence of a collapsible sinus and
a two stable steady-state system reflective of IIH. However, if
the post-withdrawal CSF pressure appears to be returning to its
original elevated state, this indicates some other cause for the
initial hypertension. The results of King et al. [4] suggest that
the former of these responses occurred in IIH patients where
drastic changes in sinus pressure were observed following CSF
withdrawal.
ACKNOWLEDGMENT
Partial support for this research (SAS, JS, and WDL) has been
provided by NASA under cooperative agreement NCC5 - 581.
397
REFERENCES
[1] D. Binder, J. Horton, M. Lawton, and M. McDermott, “Idiopathic in-
tracranial hypertension,” Neurosurgery, vol. 54, no. 3, pp. 538–540,
2004.
[2] R. Farb, I. Vanek, J. Scott, D. Mikulis, R. Willinsky, G. Tomlinson,
and K. terBrugge, “Idiopathic intracranial hypertension: The preva-
lence and morphology of sinovenous stenosis,” Neurology, vol. 60, pp.
1418–1424, 2003.
[3] J. Higgins and J. Pickard, “Lateral sinus stenosis in idiopathic intracra-
nial hypertension resolving after CSF diversion,” Neurology, vol. 62,
pp. 1907–1908, 2004.
[4] J. King, P. Mitchell, K. Thomson, and B. Tress, “Manometry combined
with cervical puncture in idiopathic intracranial hypertension,” Neu-
rology, vol. 58, pp. 26–30, 2002.
[5] F. Bono, D. Messina, C. Giliberto, D. Critiano, G. Broussard, F. Fera,
F. Condino, A. Lavano, and A. Quattrone, “Bilateral transverse sinus
stenosis predicts IIH without papilledema in patients with migraine,”
Neurology, vol. 67, pp. 419–423, 2006.
[6] M. Torbey, R. Geocadin, A. Razumovsky, D. Rigamonti, and M.
Williams, “Utility of CSF pressure monitoring to identify idiopathic
intracranial hypertension without papilledema in patients with chronic
daily headache,” Cephalalgia, vol. 24, pp. 495–502, 2004.
[7] S. Stevens, W. Lakin, N. Thakore, P. Penar, and B. Tranmer, “A Mod-
eling Study of Idiopathic Intracranial Hypertension,” in Modelling in
Medicine and Biology VI, M. Ursino, Ed. Southampton, U.K.: WIT-
Press, 2005.
[8] S. Stevens, M. Previte, W. Lakin, N. Thakore, P. Penar, and B.
Hamschin, “Idiopathic intracranial hypertension and transverse sinus
stenosis: A modeling study,” Math. Med. Biol., vol. 24, no. 1, pp.
85–109, 2006.
[9] R. De Simone, E. Marano, C. Fiorillo, F. Briganti, F. Di Salle, A. Volpe,
and V. Bonavita, “Sudden re-opening of collapsed transverse sinuses
and longstanding clinical remission after a single lumbar puncture in
a case of idiopathic intracranial hypertension. pathogenetic implica-
tions.,” Neurol. Sci., vol. 25, no. 6, pp. 342–344, 2005.
[10] D. Friedman and E. Rausch, “Headache diagnoses in patients with
treated idiopathic intracranial hypertension,” Neurology, vol. 58, no.
10, pp. 1551–1553, 2002.
[11] D. Soler, T. Cox, P. Bullock, D. Calver, and R. Robinson, “Diagnosis
and management of benign intracranial hypertension,” Arch. Dis.
Child., vol. 78, no. 1, pp. 89–94, Jan. 1998.
[12] F. Bono, C. Gilberto, C. Mastrandrea, D. Cristiano, A. Lavano, F. Fera,
and A. Quattrone, “Transverse sinus stenoses persist after normaliza-
tion of the CSF pressure in IIH,” Neurology, vol. 65, pp. 1090–1093,
2005.
[13] M. Hayashi, Y. Handa, H. Kobayashi, H. Kawano, H. Ishii, and S. Hi-
rose, “Plateau-wave phenomenon (i),” Brain, vol. 114, pp. 2681–2691,
1991.
[14] M. Hayashi, H. Kobayashi, Y. Handa, H. Kawano, S. Hirose, and H.
Ishii, “Plateau-wave phenomenon (ii),” Brain, vol. 114, pp. 2693–2699,
1991.
[15] M. Rosner and D. Becker, “Origin and evolution of plateau waves,” J.
Neurosurg., vol. 60, no. 1, pp. 312–324, 1984.
[16] M. Ursino, “A mathematical model of the relationship between cere-
bral blood volume and intracranial pressure changes: The generation
of plateau waves.,” Ann. Biomed. Eng., vol. 19, no. 1, pp. 15–42, 1991.
[17] M. Ursino and C. Lodi, “A simple mathematical model of the inter-
action between intracranial pressure and cerebral hemodynamics,” J.
Appl. Physiol., vol. 82, no. 4, pp. 1256–1259, 1997.
[18] M. Heil, “Stokes flow in collapsible tubes: Computation and experi-
ment,” J. Fluid Mech., vol. 353, pp. 285–312, 1997.
[19] S. Sorek, J. Bear, and Z. Karni, “A non-steady compartmental flow
model of the cerebrovascular system,” J. Biomech., vol. 21, pp.
695–704, 1988.
[20] S. Stevens, W. Lakin, and P. Penar, “Modeling steady-state intracra-
nial pressures in supine, head-down tilt, and microgravity conditions,”
Aviat. Space Environ. Med., vol. 76, pp. 329–338, 2005.
[21] W. Lakin, S. Stevens, B. Tranmer, and P. Penar, “A whole-body math-
ematical model for intracranial pressure dynamics,” J. Math. Biol., vol.
46, pp. 347–383, 2003.
[22] W. Lakin, S. Stevens, and N. Thakore, “On the pressure dependence
of a starling-like resistor,” Int. J. Pure Appl. Math., vol. 32, no. 1, pp.
23–32, 2006.
[23] S. Stevens and W. Lakin, “Local compliance effects on the global CSF
pressure-volume relationship in models of intracranial pressure dy-
namics,” Math. Comput. Modelling Dynamical Syst., vol. 6, no. 4, pp.
445–465, 2001.
398 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 2, FEBRUARY 2008

[24] S. Strogatz, Nonlinear Dynamics and Chaos: With Applications [35] J. Klingelhofer, G. Hajak, D. Sander, M. Schulz-Varszegi, E. Ruther,
to Physics, Biology, Chemistry, and Engineering. Reading, MA: and C. B. , “Assessment of intracranial hemodynamics in sleep apnea
Addison-Wesley, 1994. syndrome,” Stroke, vol. 23, no. 10, pp. 1427–1433, 1992.
[25] D. Mautner-Huppert, R. Haberl, U. Dirnaql, A. Villringer, P. [36] N. Cherniak, “Sleep apnea and its causes,” J. Clin. Invest., vol. 73, pp.
Schmiedik, and K. Einhaupl, “B-waves in healthy persons,” Neurol. 1501–1506, 1984.
Res., vol. 11, no. 4, pp. 194–196, 1989.
[26] M. Edsbagge, M. Tisell, L. Jacobsson, and C. Wikkelso, “Spinal csf ab-
sorption in healthy individuals,” Amer. J. Physiol. Regul. Integr. Comp.
Physiol., vol. 287, pp. R1450–R1455, 2004.
[27] A. Marmarou, K. Shulman, and R. Rosende, “A nonlinear analysis of
the cerebrospinal fluid system and intracranial pressure dynamics,” J. Scott A. Stevens, photograph and biography not provide at the time of
Neurosurg., vol. 48, pp. 332–344, 1978. publication.
[28] M. Ursino, “A mathematical study of human intracranial hydrody-
namics. Part 1—the cerebrospinal fluid pulse pressure.,” Ann. Biomed.
Eng., vol. 16, pp. 379–401, 1988.
[29] M. Czosnyka, S. Piechnik, S. Richards, P. Kirkpatrick, P. Smielewski,
and J. Pickard, “Contribution of mathematical modelling to the inter- Jesse Stimpson, photograph and biography not provide at the time of
pretation of bedside tests of cerebrovascular autoregulation,” J. Neurol. publication.
Neurosurg. Psychiatry, vol. 63, pp. 721–731, 1997.
[30] S. Piechnik, M. Czosnyka, H. Richards, P. Whitfield, and J. Pickard,
“Cerebral venous blood outflow: A theoretical model based on labora-
tory simulation.,” Neurosurgery, vol. 49, pp. 1214–1223, 2001. William D. Lakin, photograph and biography not provide at the time of
[31] T. Pedley, B. Brook, and R. Seymour, “Blood pressure and flow rate in publication.
the giraffe jugular vein,” Philosophical Trans.: Biol. Sci., vol. 351, no.
1342, pp. 855–866, 1996.
[32] T. Pedley and X. Luo, “Modelling flow and oscillations in collapsible
tubes,” Theoret. Comput. Fluid Dynamics, vol. 10, pp. 277–294,
1998. Nimish J. Thakore, photograph and biography not provide at the time of
[33] K. Berger, I. Ayappa, I. Sorkin, R. Norman, D. Rapoport, and R. publication.
Goldring, “Co(2) homeostasis during periodic breathing in obstructive
sleep apnea,” J. Appl. Physiol., vol. 88, no. 1, pp. 257–264, 2000.
[34] A. Lee, K. Golnik, R. Kardon, M. Wall, E. Eggenberger, and S. Yedally,
“Sleep apnea and intracranial hypertension in men,” Opthalmology,
vol. 109, no. 3, p. 482485, 2002. Paul L. Penar, photograph and biography not provide at the time of publication.