In-vitro characterization of an im

mplantable thermal flow sensor for hydrrocephalus

J. Burger1, T. Bork1, A. Hogg2, M. Lempen2, D. Mueller2, D. Joss3, T. Bardyn4, P. Bu
uechler4,
H.. Keppner3 and Y. Tardy1
1
Codman NNeuro Sciences sàrl, Le Locle, Switzerland
2
Berne Univeersity of Applied Sciences, Biel, Switzerland
3
Neode, Haute Ecolee ARC Ingénierie, La Chaux-de-Fonds, Switzerland
4
Institute for Surgical Technologiees and Biomechanics ISTB, University of Bern, Bern, Switzerland

Abstract— An implantable thermal flow sensor for the

treatment of hydrocephalus has been developed. The sensor
uses passive telemetry at 13.56 MHz for power supply and II. MATERIALS AND METHODSS
read out of the measured flow rate. The in vitro p performance
of the sensor has been characterized using artificiaal Cerebros- A. Sensor description
pinal Fluid (CSF) with increased protein concen ntration and
artificial CSF with 10% fresh blood. No drift coould be ob- The implant consists of three main partss: a thermal flow
served in the flow sensor measurement which cou uld be asso- sensor, a microcontroller and a RF sectio on (Fig. 1). The
ciated to a deposition of proteins at the sensitive su
urface walls
of the packaged flow sensor.
flow sensor (Sensirion AG, Stäfa, Switzerrland) is realized
on a CMOS silicon chip that contains the sensor
s as well as
Keywords— implant, flow sensor, hydrocephalus, protein ad- all necessary electronics to treat the senso
or signal (ADC,
sorption linearization, amplification, temperature compensation,
calibration memory). The sensors have been n calibrated with
a specification requirement of +-10% for a flow range be-
I. INTRODUCTION tween 2 ml/h and 40 ml/h.
Hydrocephalus is one of the most common congenital antenna
disorders of the central nervous system. Ceerebrospinal power supply
microcontro
oller
Fluid (CSF) is produced by secretory cells of tthe choroid RF components
plexus at a rate of approximately 20 ml per hourr, circulates
through the ventricular system and is finally abbsorbed into
CMOS
the venous system via the superior sagittal sinus. nsor
flow sen
fluid channel
Hydrocephalus occurs when there is an imbbalance be- (0.9x0.9mm2)
tween the amount of CSF that is produced andd the rate at load modulator
which it is absorbed. The occurrence of hydroccephalus in
newborn children is about 1 out of 500 – 2’000 [1, 2]. The
treatment of hydrocephalus requires the implanntation of a
15mm

shunt for fluid drainage. Per year, more than 1000’000 shunt
valves are implanted worldwide. However, mallfunctioning
of the shunt catheter cannot be excluded which might lead
to severe complications especially for children [3].
In order to allow a better management of hyddrocephalus 12mm
shunted patients, a thermal flow sensor which w will be im-
planted close to the shunt valve has been deveeloped. The Fig.1: Components of the implantable thermal flow w sensor. The height
sensor will allow an improved long-term monitooring of the of the encapsulated implant is 4.5 mm. Bottom left: Bottom
B view of the
functioning of the valve as well as a better underrstanding of packaged sensor showing the rectangular floow channel.
Bottom right: Top view of the packaged flow
f sensor
the flow of CSF. Whereas microsystems including pressure
sensors have been developed for intracranial presssure moni-
One of the key features of the sensor is the thermal flow
toring [4-6], flow sensing of CSF is currently done using
measurement which is done through the PyrexP glass sub-
MRI methods [7].
strate of 100 um thickness. Using this co onfiguration, the
thermal sensors and the heaters are not dirrectly exposed to

O. Dössel and W.C. Schlegel (Eds.): WC 2009, IFMBE Proceedings 25/VIII, pp. 265–268, 2009.
www.springerlink.com
266 J. Burger et al.

the body fluids. Therefore, no coating or passivation of the and added immediately to the warmed artificial CSF solu-
sensors and heaters is needed and electrical feed-throughs to tion in order to avoid clotting. As the mixture tends to de-
the fluidic chamber are avoided. In that way, a highly reli- grade after about 24hrs at 37ºC due to break up of red blood
able and long-term biocompatible implant could be realized cells, the solution was changed daily.
which is especially important as the sensor will be inside the
body for up to 10 years or even longer [8]. Table 1 Composition of artificial Cerebrospinal fluid
In order to provide the implant with energy and to read
out the measured flow rate, passive telemetry is used at a Conc. in CSF
Chemical Name Formula
RF frequency of 13.56 MHz [9]. The RF section of the [g/l]
implant is responsible for extracting energy for the implant, Sodium chloride NaCl 8.660
for transmitting the level of RF induced voltage to the mi- Potassium chloride KCl 0.224
crocontroller as well as for modulating the FSK signal for Calcium chloride dihydrate CaCl2-2H20 0.206
communication with the external reading unit. Magnesium chloride hexahydrate MgCl2-6H20 0.163
Sodium phosphate NaH2PO4-
0.027
B. In-vitro characterization of the implantable flow sensor monobasic monohydrate H20
Di-Sodium hydrogen Na2HPO4-
0.214
Thermal flow sensors with sensing and heating elements phosphate heptahydrate, acs 7H20
attached to the walls of the fluid channel are sensitive to Sodium Azide 0.1 M Solution N3Na 0.77 ml
modifications of the sensing surface area. A potential pro- D-(+)-Glucose C6H12O6 0.600
tein or blood component adsorption especially during the Urea CH4N2O 0.201
first days after implantation with a risk of elevated protein Albumin (from human serum) 2.000
concentration and blood within the CSF could have an in- Gamma-globulin
0.015
fluence on the flow sensor response. from human blood
In order to investigate the sensor performance under
worst case situations, the flow rate of the sensor was meas- C. Fluidic test setup
ured using elevated protein concentration and blood content
in CSF. The fluidic circuit for characterization of the implantable
The composition of normal CSF is close to that of ul- flow sensors consists of a tubing pump which pulls the test
trafiltrate of plasma and has a protein concentration of less liquids artificial CSF, artificial CSF with 10% fresh blood
than 0.5% of plasma [10]. The protein concentration is low and sterile water as reference solution through three parallel
with about 0.2 g/l. In order to accelerate the protein absorp- fluidic circuits (Fig. 2). The test setup was kept in an oven
tion, a protein concentration increased by a factor of 10 to at 37ºC. Once a day, the flow sensor measurement was
reach 2 g/l was used for the artificial CSF in these tests.
Due to patients who have hydrocephalus as a result of a Powersupply
3.3V Oven at 37°C

haemorrhagic event such as periventricular haemorrhage in NI I2C - USB

interface
the premature neonate, or subarachnoid haemorrhage in the LabView VI
Sensor Sensor Sensor
adult, CSF containing a profile of proteins different from Computer
1 2 3
Fluidic Path for Scale
that of “normal” CSF must be taken into consideration as Measurement

well. Fluidic Path During

Cycling
Peristaltic
For the testing, the following solutions were used: Pump

- artificial CSF with 2 g/l protein (Albumin from hu- IN

1
IN
2
IN
3
man serum) OUT IN Prime

- artificial CSF with 2 g/l protein (Albumin from hu- Syringe

man serum) and 10% fresh blood (priming fluidic
path to scale)
Magnetic
Scale Stirrer
- sterile water with 0.65% Sodium azide as reference
solution
In order to simulate a potential protein deposition that Fig. 2: Test setup for in-vitro characterisation of the implantable flow
might affect the sensor function after e.g. a subarachnoid sensor. Only one fluidic circuit is shown
haemorrhage, fresh human blood was drawn from a vein

IFMBE Proceedings Vol. 25

In-vitro Characterization of an Implantable Thermal Flow Sensor for Hydrocephalus 267

20.0 100
Relative flow rate (difference between scale and
Sensor 11
Adding Blood to Removing Blood 50ml/h
Sensor 31, 32, 33 from Sensor 31, Sensor 12
15.0 32, 33 Sensor 13
20ml/h 20ml/h
Sensor 21
Sensor 22
10.0 10ml/h

Flow rate [ ml/h]]

10 Sensor 23
5ml/h Sensor 31
5.0
Sensor 32
sensor measurement) [ %]]

2.5ml/h
Sensor 33
0.0 Tested Flow Rate
DI H2O with 0.65% Sodium
1 Azide + Detergent
-5.0
Artificial CSF with 2g/l
Protein

-10.0 Artificial CSF with 2g/l

Protein + 10% fresh blood
during 2 days
-15.0 0.1

0 2 4 6 8 10 12 14 16 18 20 22 24 26 time [days]
Fig. 3: Relative flow rate of the implantable thermal flow sensor as a function of time. In test circuit 3 (sensors 31, 32, 33),
10% fresh blood was temporarily added.

compared to a reference scale measurements (gravimetric surement. Within those 26 days the flow rate has been va-
method). During a period of 180 s, the flow sensor reading ried from the nominal flow of 20 ml/h down to 2.5 ml/h and
was taken at a frequency of 10 Hz and averaged. The scale up to 50 ml/h in order to characterize the relative flow rate
reading was taken in parallel and the flow was calculated by of the flow sensors at different flow rates within the operat-
taking the start reading and the end reading of the balance. ing range. It can be seen that there is no apparent drift in
A relative flow rate in % was calculated by taking any of the 3 fluidic circuits, especially when only compar-
ing the beginning and the end of the testing where the 20
low rate low sensor low rate s ale
ml/day flow rate was tested.
low rate s ale The average of all relative flow rates (261 values) shown
In order to simulate a continuous use of the flow sensor in Fig. 3 is offset by –5.4% for a test temperature of 37°C.
during the test period of 26 days, the heater of the flow As the flow sensors have been calibrated at a temperature of
sensor was periodically turned on for 5 minutes and then 22ºC, the influence of temperature on the difference be-
turned off for 10 minutes in between the scale measure- tween the scale measurements and the flow sensor mea-
ments. The power consumption of the CMOS flow sensor surements was evaluated.
was measured to be 16.3 mW at a supply voltage of 3.3 V. For that purpose, 3 sensors have been characterized at 2
The power consumption of the ultra-low power microcon- different flow rates of 5 ml/h and 10 ml/h at temperatures
troller was 2.87 mW. 22ºC and 37ºC. The temperature offset of the relative sensor
flow rates at 37°C is about 6.3% compared to the relative
sensor flow rates at 22°C.
III. RESULTS A temperature compensation of the offset of –5.4% at
37°C with the temperature offset of +6.3% leads to a com-
In Fig. 3 the relative flow rate for all sensors for a test pensated average of all relative flow rates of +0.9%.
period of 26 days is shown as the difference between the The reference liquid containing sensors (brown in Fig. 3)
flow sensor measurement and the comparative scale mea- as well as the CSF/protein containing sensors (yellow in

IFMBE Proceedings Vol. 25

268
Fig. 3) are maintaining essentially the same difference be-
tween sensor and scale measurements throughout the test.
Only the sensors of the fluidic circuit that had temporari-
ly 10% fresh blood added to it (red in Fig. 3) show an in-
crease of relative flow rate of ~20% to ~25% during the
time the blood is present. If blood components were depo-
sited on the channel wall, this additional layer would de-
crease the heat that is transferred to the fluid which would
result in a decreased measured flow rate of the flow sensor.
The observed increase of flow rate can be explained by
the settling of the heavier red blood cells on the bottom of
the flow channel due to the low flow rate used. This sedi-
mentation could be observed visually in the channel tubing.
The viscosity of this sediment becomes more elevated re-
sulting in a lower flow rate near the bottom of the flow
channel. Since the overall volume of fluid passing through
the channel is kept constant by the peristaltic pump, the
flow rate at the top of the flow channel, where only the
lower viscous CSF-plasma mix is flowing, increases accor-
dingly.
Comparing the relative flow rates before and after the
blood was added/removed even the flow sensors in the
fluidic circuit with artificial CSF and 10% blood don’t ex-
hibit a drift. Furthermore when looking at Fig 3 one can see
that there is no major influence on the relative flow rate
when tested between 2.5 ml/day and 50 ml/day. All meas-
ured differences between sensor and scale measurements for
all sensors remain within 0% to –10% except sensor 32 with
~-12% during the 50ml/day measurement.
This deviation might be explained by a misaligned flui-
dic connection it was found to be very important to have a
laminar flow at the sensor location which is depending on
the fluidic connection and the flow rate used.
IV. CONCLUSIONS
No drift could be observed in the flow sensor measure-
ments which could be associated to a deposition of proteins
even though the tested protein level was about 10 times
higher than the normal protein level in human beings. Even
if a protein layer has been deposited on the flow channel
wall during the 26 days of exposure, its thickness is not
sufficient to have an influence on the sensor reading.
Blood which is not homogeneously mixed in the fluid
can temporarily disturb the flow sensor reading. However in
a real life situation this is not relevant as the patient is mov-
ing and the fluid remains mixed. Blood does not create any
permanent sensor changes due to deposition of blood com-
ponents on the flow channel wall.
IFMBE Proceedings Vol. 25
J. Burger et al.
The flow sensor specification requirement of +-10% for a
flow range between 2 ml/h and 40 ml/h. could be confirmed
at test conditions of 37°C.
ACKNOWLEDGEMENT
This research was supported by the Swiss innovation pro-
motion agency (grant CTI #8558.1 LSPP-LS).
The authors would like to thank J. Sounggalon Siringoringo
for the three dimensional view of the implant and Dr. Roger
Bayston for the recommendations concerning the test liq-
uids and their preparation.
REFERENCES
1. Murshid WR, Jarallah JS, Dad MI (2000), Epidemiology of infantile
hydrocephalus Saudi Arabia: birth prevalence and associated factors,
Pediatr Neurosurg 2000;32:119–123.
2. Persson E-K; Anderson S, Wiklund L-M, Uvebrant P, Hydrocephalus
in children born in 1999-2002: epidemiology, outcome and ophthal-
mological findings. Child's Nervous System, 23 (10), pp. 1111-1118
3. Oikonomou J, Aschoff A, Hashemi B, Kunze S (1999), „New
Valves - New Dangers? 22 Valves (38 Probes) Designed in the
Nineties in Ultralong-Term Test (365 Days)“, European Journal of
Pediatric Surgery 9, Suppl. I, pp. 23-26
4. Ko WH, Leung A M, Cheng E and Lorig R J (1981), Intracranial
pressure telemetry system: I Hardware development , Biotelem. Pa-
tient Monit. 8 (3), pp. 131-150
5. Munshi I, Lathrop D, Madsen JR, Frim DM (1998), “Intraventricular
Pressure Dynamics in Patients with Ventriculopleural Shunts: A
Telemetric Study” , Pediatr Neurosurg 28, pp. 67-69
6. Ginggen, A.; Tardy, Y.; Crivelli, R.; Bork, T and Renaud, P (2008),
A Telemetric Pressure Sensor System for Biomedical Applications,
Biomedical Engineering, IEEE Trans Biomed Eng 55 (4) pp. 1374 –
1381
7. Linninger, A.A.; Xenos, M.; Zhu, D.C.; Somayaji, M.R.; Kondapalli,
S.; Penn, R.D. (2007), Cerebrospinal Fluid Flow in the Normal and
Hydrocephalic Human Brain, IEEE Trans Biomed Eng 54 (2), pp.
291 - 302
8. T. Bork (2007), "Thermal Flow Sensor Having Streamlined Packag-
ing ", US Patent No. 7,181,963 B2
9. Neukomm P A, Furrer D (2003), “Design Rule for a Weakly Cou-
pled Inductive Link avoiding the “Black Hole” in Passive Telemetry
Data Transmission for an Implantable Thermal Flow Sensor”, Pro-
ceedings of the 17th Conference of the International Society on Bio-
telemetry ISOB 17, Sept 1-5, 2003, Brisbane,
10. Kuschinsky W. (1996), Blood-Brain Barrier and the Production of
Cerebrospinal Fluid , In: Comprehensive Human Physiology (Edited
by: Greger R, Windhorst U), Volume 1, Berlin, Springer, pp. 545-
558
Author: Juergen Burger
Institute: Codman Neuro Sciences sàrl
Street: Girardet 29
City: CH-2400 Le Locle
Country: Switzerland
Email: jburger@its.jnj.com