Hi ngh ton quc v iu khin v T ng ho - VCCA-2011

VCCA-2011
A Study on Cerebral Perfusion Pressure Control
Mt nghin cu iu khin p lc ti mu no
Nguyn Ch Ngn
Trng i Hc Cn Th, e-Mail: ncngon@ctu.edu.vn

Abstract
This study presents a solution of raising the mean
arterial pressure (MAP) by automatic control of
Noradrenaline (NA) infusion for a purpose of
regulating the cerebral perfusion pressure (CPP) at an
adequate level compensating the increase of the
intracranial pressure (ICP) often caused by traumatic
brain. In general, the Datex AS/3 and Codman ICP
express monitoring devices are used to measure the
current MAP and ICP of the patient. The CPP is
calculated from these values, and considered as a
feedback of a PID controller. The NA infusion rate is
generated to maintain the CPP at a reference level due
to the increase of ICP. The control system was tried
on regulating the CPP around 70 mmHg on 10 pigs
under anaesthesia condition and manual increasement
of ICP by an intraventricular balloon catheter. The
system response achieves the settling time 51.30 min
and the overshoot 41.72 mmHg. Results indicated
the feasibility of the control system.
Tm tt
Nghin cu ny trnh by mt gii php nng cao
huyt p ng mch trung bnh MAP bng cch t
ng iu khin vic tim Noradrenaline, nhm mc
tiu duy tr p lc ti mu no CPP mt gi tr
thch hp b tr li vic tng p lc ni s ICP,
thng gy ra do tn thng no. Mt cch tng qut,
thit b Datex AS/3 v Codman ICP Express c s
dng o gi tr MAP v ICP hin ti ca i tng.
CPP c tnh ton t hai gi tr ny v c hi tip
v b iu khin PID. B PID s quyt nh tc
tim Noradrenaline cho my bm t ng, nhm duy
tr CPP gi tr mong mun, b tr li s gia tng
ICP. H thng iu khin c thc nghim n
nh CPP xung quanh gi tr 70 mmHg trn 10 con
heo, trong iu kin gy m kt hp vi vic tng ICP
nhn to thng qua s thay i p lc bng hi ni
no tht. p ng ca h thng c thi gian xc lp
t 51.30 pht v vt l ch 41.72 mmHg. Kt
qu thc nghim kim chng c kh nng iu
khin p lc ti mu no ca h thng.
Ch vit tt
MAP:
CPP:
ICP:
Mean arterial pressure
Cerebral perfusion pressure
Intracranial pressure
ANU: Non-renal effect of angiotensin
AMM: Muscle vasoconstriction
ARM: Vasoconstrictor effect of all autoregulation types
AUM: Sympathetic vasoconstrictor effect on arteries
PAM: Effect of arterial pressure in distending arteries
RAM: Basic vascular resistance of muscles
RAR: Basic resistance of non-muscle, non-renal arteries
RVS: Venous resistance
VIM: Blood viscosity effect on resistance
RSM: Vascular resistance in muscle tissues
RSN: Vascular resistance in non-muscle, non-renal tissues
REM: Effect of SNP on RSM
REN: Effect of SNP on RSN
F: Low-pass filter
1. Introduction
The cerebral perfusion pressure (CPP) is defined as
the mean arterial pressure (MAP) minus the
intracranial pressure (ICP), i.e. CPP=MAP-ICP,
which presents the pressure gradient driving cerebral
blood flow (CBF) and hence oxygen and metabolite
delivery [1]. An intractable increase in ICP leading to
a progressive decrease in CPP and CBF is the
dominant cause of death in patients with severe brain
trauma [2]. When the cerebral autoregulation is
compromised, manually-controlled therapy of MAP
has many disadvantages regarding precision and
response time [3]. Therefore maintenance of an
adequate CPP is vital to prevent secondary brain
damage when anaesthetising patients who may have
raised ICP, which is a common problem in
neurosurgical and neurological practice [4-5].
In order to develop a CPP control system for healing
the traumatic brain injury patients from traffic
accidents, especially in the traffic conditions in
Vietnam, the aim of the study is leading to develop a
closed-loop system that can regulate the CPP around
70 mmHg through raising the MAP by automatically
controlling noradrenaline (NA) infusion. The study
was tested on 10 pigs at the Anesthesiology and
Intensive Care, University of Rostock, Germany.
2. Materials and Methods
2.1 Overview of the control system
2.1.1 Clinical protocol
Under approving of the local Ethics Committee on
animal research, the study was performed on 10
anesthetized pigs. After induction of anaesthesia
(midazolam, propofol, fentanyl, rocuronium) and
intubation of the trachea, a central venous catheter
and an arterial line via femoral artery were placed.
792
Hi ngh ton quc v iu khin v T ng ho - VCCA-2011

VCCA-2011
Subsequently an ICP measuring probe (i.e., Codman
ICP express microsensor) was placed. To artificially
increase the ICP, an intraventricular balloon catheter
was placed via a borehole trepanation. The correct
positioning of the balloon was judged by CSF
aspiration and x-ray. This technique allowed
adjustment of ICP to an arbitrary level without major
brain damage. After manually increasing the ICP, NA
was applied by a closed-loop controlled infusion to
maintain the CPP at a desired level as fast as possible.
2.1.2 Control structure
The control structure is explained based on a closed-
loop control system performed for an anaesthetized
pig (Fig.1). After measuring the current MAP and ICP
of the pig by the Datex AS/3 and Codman ICP
express monitoring systems, the current CPP is
calculated and sent back to the input. Based on the
difference between this feedback value and the
setpoint, the controller generates a next infusion rate
to deliver NA into the veins for raising the MAP due
to the artificial increase of ICP. This procedure is
repeated for the next sampling time to maintain the
CPP at a desired level.

- CPP
ref.

CPP
Patient
IV Fluid
Control
PC
Datex
AS/3
&
Codman
ICP Exp.
PC
Inf.
Pump
Graseby
3400
Monitoring
System

Super-
visor
PID
NA
Manager

Fig. 1 Overview of the control system
2.2 Model of blood pressure response to NA
2.2.1 Blood pressure response
To develop the blood pressure response model, the
MAP of 2 pigs in 8 experiments on raising MAP by
different constant infusion rates of NA were
measured. In each experiment, after a short delay time
from starting infusion, the MAP is rapidly increased
to maximum effect of NA. The steady-state is
maintained in some minutes then MAP is decreased
by multiple feedbacks of BP autoregulatory
mechanisms [6]. When stopping the infusion pump,
the MAP returns to normal value in a short time. The
reaction of body is strong nonlinear when increasing
BP to a high level.
2.2.2 Guytons model
Guytons model consists of 18 modules containing
about 600 physiological parameters and variables
which was developed to simulate human circulation.
In this system, we reconstructed Guytons model in
MATLAB/Simulink. It was also modified to model
the effects of NA on the circulatory dynamics (see
Appendix).
2.2.3 Modeling method
NA works by stimulating receptors (adrenoceptors)
which are found all over the body. When injected into
a vein, NA acts mostly on a type of adrenoceptor
known as an alpha receptor. These are present in the
muscle within the walls of blood vessels of the
extremities. By stimulating these alpha receptors, NA
causes the muscle to contract, resulting in narrowing
of the blood vessels. Therefore, the idea to model the
effects of NA on the circulation is that under
administered this hypertensive agent, the vascular
resistances in muscle tissues (RSM) and in non-
muscle, non-renal tissues (RSN) are increased.

RSM
RSN
RSMG
RAR
ARM
VIM
PAM
AUM
RVS
RAM
ANU
AMM
1.79
VIM
NEM
F
F
NEN
RSNG

Fig. 2 Modeling the effects of NA on vascular resistances
Let NEM and NEN be the coefficients of the effects of
NA on RSM and RSN, respectively. The module
expressed in Fig.2 and (1)-(2) presents the effects of
NEM and NEN on the vascular resistances. Where,
RSM
G
and RSN
G
denote the original vascular
resistances of Guytons model. F is a low-pass filter;
and all other notations in this module can be found in
the appendix.
79 . 1 RVS VIM
PAM ARM AUM ANU RAR VIM RSN
PAM AMM AUM ANU RAM VIM RSM
G
G

+ =
=
(1)
( )
( ) NEN RSN F RSN
NEM RSM F RSM
G
G
=
=
(2)

1 s
1
2
+

1 s
1
1
+

x
G
s T
e
1

s T
e
2

NA NEM

1 c
NEN
z

Fig. 3 NA effective module
NEM and NEN are calculated by (3), which can be
implemented in Fig.3, called NAg effective module.
In (3), NA presents the noradrenaline infusion rate. G
and express the patient sensitivity and the reaction
of the body, respectively. T
1
and T
2
are the transport
delays. t
1
and t
2
present the response time constants. c
793
Hi ngh ton quc v iu khin v T ng ho - VCCA-2011
VCCA-2011

is the different effect of NA on muscle tissues and
non-muscle, non-renal tissues.
The operation of NA effective module can be simply
described as follows: When the pump is off, NA
infusion rate is zero, so z=x=0, hence NEM and NEN
are standardized, it means NEM=NEN=1. Nothing is
changed in BP. When the infusion rate is increased, x
and z are raised by (3), so NEM and NEN are all
increased, causing an increases of RSM and RSN by
(2) leading to an increase of BP in the circulatory
dynamics by Guytons model.
z 1 NEN ; z 1 NEM
0 x , 0
0 x , x
z
1 s
e
1 s
e
G
NA
x
2
s T
1
s T
2 1
c
t

t
+ = + =

<
>
=
|
|
.
|

\
|
+
+
+
=

(3)
All parameters of (3) are given in Table 1, which were
estimated from measured data collected from 8
experiments on 2 pigs. The average values in Table 1
are used as default parameters for the NA effective
module. A simulation result of the model output is
illustrated in Fig.4 which indicates that this may be a
suitable model for CPP control system development.

Fig. 4 Model output and measured data on pig
2.3 Controller design
The csystem is designed based on an auto-tuning PID
controller. The output CPP is calculated from MAP
and ICP measured by the monitoring systems:
ICP MAP CPP = (4)
The controller aims to control NA infusion rate for
increasing MAP to maintain the CPP at a reference
level. A discrete PID controller with a form of
| |
| | 2) - e(k 1) - 2e(k - e(k)
T
K
k e T K 1) - e(k - e(k) K k u u(k)
s
d
s i p
+ +
+ + = ) ( ) 1 (
(5)
is applied [7]. Where K
p
, K
d
and K
i
are proportional,
derivative and integral gains; T
s
is the sampling
interval. For tuning the controller, this system is
applied by both online and off-line tuning methods.
2.3.1 Off-line tuning method
At starting time, an identification period is applied by
using a step infusion of NA to estimate the behaviors
of the body. The change of blood pressure under
administered NA is measured and considered as the
response of controlled object. When MAP is
increased around 10 mmHg, the step infusion is
stopped. Then the measured data is used for an early
estimation method to calculate the delay time L and
the early slope R of the output response curve [8]. The
values of R and L are used for tuning the controller
similar to Ziegler-Nichols method. The PID gains K
p
,
K
d
and K
i
are calculated by (6). Where, T
d
and T
i

present the derivative and integral time constants,
respectively.
Tab. 1 Parameters of the NA effective module
Parameters Average Min Max Unit
G 0.1 0.02 0.4 (ml/hr)
-1

-0.4 -0.1 -1.0 -
t
1
40 30 75 sec
T
1
55 20 120 sec
t
2
7 4 12 min
T
2
4 3 7 min
c 1.2 1.0 2.6 -
d i
d
p
4T T
L T
RL
1.2
K
=
=
=
;
i
p
i
d p d
T
K
K
T K K
=
=
(6)
The coefficient , in (6) is determined due to the value
of the delay time L, which is estimated from
simulation and presented in Table 2.
Tab. 2 Calculating , due to the delay time L
L(sec) [min - 30] (30 - 60] (60 - 90] (90 - max]
, 0.6 0.4 0.2 0.1
2.3.2 Online tuning method
During activities, a supervising algorithm is also used
for online tuning the PID controller to make it acting
stronger or weaker due to the behaviors of the body
(called supervisor in Fig.1). During anaesthesia, the
blood pressure typically exhibits a 2 to 4 mmHg
amplitude fluctuation effected by mechanical
ventilation [9]. Therefore, instead of using an exactly
reference level, a setpoint-bank is defined by a range
of 5 mmHg upper and lower the setpoint. The aim of
the supervising algorithm is to observe the system
output, and to tune the PID controller due to the body
reactions and the effects of clinical activities to
maintain the CPP within the setpoint-bank during
control action. The algorithm can be summarized as
follows:
- If the current CPP is within the setpoint-bank,
no change is made in the PID.
794
Hi ngh ton quc v iu khin v T ng ho - VCCA-2011

VCCA-2011
- When the current CPP is below the upper
setpoint-bank, K
p
and K
i
are increased 5% of
their current values; K
d
is decreased 5% of its
current value.
- For CPP is within the CPP

maxover
and the upper
setpoint-bank, that means the overshoot appears
in an acceptable range, K
p
and K
i
are decreased
5%, K
d
is increased 5%.
- If CPP is above the CPP
maxover
, it means an
unsafe overshoot appears, the infusion pump will
be immediately turned off and a message will be
delivered to the operator.
3. Results
This system was tested on 10 anaesthetized pigs at
animal laboratory. A selected result is demonstrated
in Fig.5. In this experiment, the CPP was tried to
maintain at 705 mmHg. The initial CPP was around
60 mmHg. It was increased to the setpoint within 5
minutes and maintained at setpoint-bank during the
operation. At the 46
th
minute, the ICP was manually
increased from 24 mmHg up to 40 mmHg leading to a
decrease of 16 mmHg in the CPP by (4). The NA
infusion rate was automatically increased to for
raising the MAP to drive the CPP back to the
setpoint-bank. That indicated a feasibility of
controlling the CPP for raised ICP patients.

Fig. 5 A result of CPP control on a pig
Table 3 presents the specifications of the CPP control
system which were estimated from experimental
results on 10 pigs. The specifications indicate a fast
response of the control system.
The animal experiment with ICP-measurement probe
and intraventricular balloon catheter to perform the
artificial increase of controlled ICP seems to be

CPPmaxover is defined as an acceptable maximum overshoot.

useful. It makes a closed-loop control system for
regulating the CPP of raised ICP patients.
Tab. 3 Characteristics of the CPP controller
Specifications Value Unit
Settling time

5 1.30 min
Rising time 2 1.06 min
Overshoot 4 1.72 mmHg
Steady-state error 3 2.14 mmHg

: included the identifying period of estimating R and L.

4. Conclusions
The study presented a method of controlling MAP by
NA infusion to maintain the CPP of raised ICP pigs.
Guyton-based physiological model was modified and
used for tuning and simulating the control system.
The auto-tuning PID controller with both online and
off-line tuning methods was designed. The closed-
loop system that can artificially increase the ICP was
performed on pigs to test the designed system.
Experimental results indicated that raising MAP by
controlling NA infusion rate may be a suitable
method to maintain the CPP at an adequate level for
raised ICP patients.
Acknowledgments
The author is grateful to Prof. Dr. Lampe, Dr.
Simanski, and Dr. Khler, University of Rostock, for
many useful discussions. He also thanks to Dr. Janda
and Dr. Bajorat, Anesthesiology and Intensive Care,
University of Rostock for clinical experiments.
References
[1] Juul N., G.F. Morris, S.B. Marshall, and L.F.
Marshall: Intracranial hypertension and cerebral
perfusion pressure: influence on neurological
deterioration and outcome in severe head injury, J
Neurosurg, vol. 92, pp16, 2000
[2] Nordstrm C.-H.: Assessment of critical thresholds
for cerebral perfusion pressure by performing bedside
monitoring of cerebral energy metabolism, Neurosurg
Focus, vol. 15, 15-E5, 2003
[3] Bajorat, J, M. Janda, C-N. Nguyen, B. Pohl, G. N.-
Schomburg: Closed Loop Control of Cerebral
Perfusion Pressure in an Acute Porcine Model,
Anesthesiology 105: A977, 2006
[4] Walters, FJM.: Intracranial Pressure and Cerebral
Blood Flow, Physiology, no. 8, pp. 18-23, 1998
[5] Dunn L.T.: Raised Intracranial Pressure, J Neurol.
Neurosurg. Psychiatry, vol. 73, pp. i23-i27, 2002
[6] Guyton, A.C., T.G. Coleman and J.P. Montani:
Annotation of large circulatory model, University of
Mississippi Medical Center, 1993
[7] Bobd V., J. Bhm, J. Fessel and J. Machcek:
Digital Self-tuning Controllers: Algogrithms,
Implementation and Application. Springer, 1
st
Ed.,
318p., 2005. ISBN-13: 978-1852339807
[8] Nguyen, C.-N., O. Simanski, R. Khler, A. Schubert,
B. Lampe: An online Fuzzy Gain scheduling for
795
Hi ngh ton quc v iu khin v T ng ho - VCCA-2011
VCCA-2011

blood pressure regulation. Proc. 16th IFAC World
Congr., Prague, CZ, 4-8 July 2005, Th-A19-TO/3
[9] Slate, J.B.: Model-based design of a controller for
infusing sodium nitroprusside during postsurgical
hypertension. PhD thesis, University of Wisconsin-
Madison, 1980
[10] Chi-Ngon Nguyen, Ralf Khler, Olaf Simanski,
Agnes Schubert, and Bernhard Lampe: The Benefits
of using Guytons model in Hypotensive Control
System, Journal of Computer Methods and Programs
in Biomedicine, Vol. 89, No. 2, pp. 153-161, 2008,
ISSN 0169-2607, doi:10.1016/j.cmpb.2007.03.005.

Appendix Modified Guytons model was developed in the study

Fig. 6 Circulatory dynamics [6]
The basic hemodynamic of the circulation can be
summarized in Fig.6. It presents the pathway of blood
flow (BF) around the circuit from arteries, to veins, to
heart, to lungs and back to heart again. This module is
divided into five different volume segments included
the aorta, the veins, the right atrium, the pulmonary
arteries, and combination of pulmonary veins and left
atrium. In each segment, the BF from this respective
segment to the next segment (Q
out
) is calculated
through dividing pressure difference (AP
x
) by
resistance (R
x
); the volume is integrated with respect
to time (V
x
); and volume of filling that segment
(VEX
x
) determines its pressure (P
x
), with C
x
is the
compliance of the segment (i.e., the slope of the
pressure-volume family curves, C=V/P). The
circulatory dynamics module also calculates the
resistances and their effects by various factors. The
computing method can be briefly explained by
following equations [6]:
flow output
R
P
Q
x
x
out
;
A
=

volume of change Q Q V
out in x
; = A

volume dt V V
x x
;
}
A =

excess in volume V V VEX
x x x
; ) 0 ( =

pressure
C
VEX
P
x
x
x
; =

796
Hi ngh ton quc v iu khin v T ng ho - VCCA-2011

VCCA-2011

Translating Guytons model into Simulink has many
advantages [10]. During simulations, the control
system designer can easily adjust some parameters
and interrupt the model computations to analyze some
variables. Moreover, it is advantageously to view any
signals by Simulink scopes. The Guytons model in
MATLAB/Simulink version was developed in Fig.7.
In this model, the circulatory dynamic module was
modified to insert two inputs of REN and REM for
modelling the effects of NA on arterial blood pressure
(see Fig.2 and Fig.3).

Fig. 7 Simulink version of Guytons model developed in the study [10]

Nguyen Chi Ngon received his
B.Sc. and M.Sc. degrees in E.E.
from Can Tho University and
National University, Ho Chi Minh
City University of Technology, in
1996 and 2001, respectively. The
degree of Ph.D. was awarded by the
University of Rostock, Germany, in
2007. Since 1996, he has worked at the Can Tho
University. Currently, he is a senior lecturer at
Department of Automation Technology. He is
working as positions of Chief of Process Control Lab,
Director of Electrical and Electronic Center, and Vice
Dean of College of Engineering Technology at the
Cantho University. His research interests are
intelligent control, medical control, pattern
recognition, classifications, speech recognition and
computer vision.
797