[ Pobierz całość w formacie PDF ]
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
TBME-00056-2009.R1
1
A Technique to Measure Eyelid Pressure using
Piezoresistive Sensors
Alyra J. Shaw*, Brett A. Davis, Michael J. Collins, Leo G. Carney.
Pressure from the eyelids can alter the corneal surface
topography. The cornea is responsible for most of the
focusing power of the eye so any changes to its surface shape
can influence vision. Abnormal eyelids (due to disease) can
increase or alter the pressure on the cornea [4-7]. However
pressure from healthy eyelids, when the upper and lower
eyelids move closer to the centre of the cornea during reading
can also cause temporary corneal distortion [8-13].
There have been attempts to measure eyelid pressure using
modified contact lenses as manometers filled with either air or
water. The apparatus designed by Miller [14] had a water-
filled rubber balloon on the inner side of a hard contact lens
and a catheter connected to a pressure transducer on the outer
side. However a second contact lens had to be worn to protect
the cornea so the total thickness of the system was 2.5 mm.
This thickness greatly alters the normal relationship between
the eyelids and the eye. The use of water has also been
criticised as there was no way to ensure that it was gas-free,
since any air present would inflate the readings [15]. Lydon
and Tait [15] used a hard contact lens with a silicone
elastomer contact lens over the top to create a small chamber
which was filled with gas. While the quantification of the
eyelid pressure was not published they concluded that these
pressures were „small‟. An apparatus similar to Miller‟s
system developed by Shikura
et al.
[16] recorded similar
results for normal lid closure and tight eyelid squeezes. The
measured eyelid pressure from these studies was between 1.7
and 51 mmHg [14-16], however the thickness and complexity
of the systems makes the reliability of the results uncertain.
An effective system to measure eyelid pressure must be
thin, so that there is minimal alteration to the eyelid-cornea
relationship, and with high sensitivity to quantify small
localised pressures. The system must non-toxic and
waterproof so that the tears do not influence measurements. It
must also be able to conform to the surface of the eye and not
be affected by blinking or eye movement. To measure the
pressure between the eyelids and the cornea we used
multiplexed array piezoresistive tactile pressure sensors (I-
scan, Tekscan Inc. Boston, MA, USA). These sensors are
relatively thin (170 µm) , available in a suitable pressure range
to measure eyelid pressure (rated at 5 psi), measurements can
be taken at up to 9.8 Hz and can be trimmed to suitable
dimensions to be placed on the eye.
Various properties of Tekscan sensors used in biomedical
applications have been previously considered [17-20]. For
eyelid pressure measurement, the influence of drift and
Abstract
— Novel procedures were developed to use a thin (0.17
mm) tactile piezoresistive pressure sensor mounted on a rigid
contact lens to measure upper eyelid pressure.
A hydrostatic calibration system was constructed and the
influence of conditioning (prestressing), drift (continued
increasing response with a static load) and temperature
variations on the response of the sensor were examined. To
optimally position the sensor-contact lens combination under the
upper eyelid margin, an
in vivo
measurement apparatus was
constructed. Calibration gave a linear relationship between raw
sensor output and actual pressure units, for loads between 1 and
10 mmHg (R²=0.96). Conditioning the sensor prior to use
regulated the measurement response and sensor output stabilised
about 10 seconds after loading. While sensor output drifts
slightly over several hours, it was not significant over the
measurement time of 1 minute used for eyelid pressure. The
error associated with calibrating at room temperature but
measuring at ocular surface temperature led to a very small
overestimation of pressure.
Eyelid pressure readings were observed when the upper eyelid
was placed on the sensor and removed during a recording. When
the eyelid pressure was increased by pulling the lids tighter
against the eye, the readings from the sensor significantly
increased.
Index Terms
— Eyelid pressure, Measurement, Piezoresistive
Sensors
I.
I
NTRODUCTION
he eyelids act as an anterior physical barrier for the eye.
Blinking of the eyelids is a protective mechanism which
can occur in response to external stimuli such as a sudden loud
noise or a flash of light, with the closing mechanism taking
less than 100 ms [1, 2]. The eyelids also maintain the health
of the eye by replenishing the tear film over the cornea during
normal involuntary blinking. In this role, the eyelids have
been likened to windscreen wipers [3], with the inner edge of
the eyelids serving to spread the tears with each blink.
T
Manuscript received January 20, 2009.
* A.J. Shaw is with the Contact Lens and Visual Optics Laboratory, School
of Optometry, Queensland University of Technology, Brisbane, QLD, 4001,
Australia. (email: aj.shaw@qut.edu.au).
B.A. Davis, M.J. Collins and L.G Carney are with the Contact Lens and
Visual Optics Laboratory, School of Optometry, Queensland University of
Technology, Brisbane, QLD, 4001, Australia. (email:
.
Copyright © 2008 IEEE. Personal use of this material is permitted.
However, permission to use this material for any other purposes must be
obtained from the IEEE by sending an email to pubs-permission@ieee.org.
Authorized licensed use limited to: Politechnika Slaska. Downloaded on May 29, 2009 at 04:19 from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
TBME-00056-2009.R1
2
temperature particularly need to be considered. Drift or creep
is the change in sensor output with a static load, and is thought
to be due to the piezoresistive ink [17]. To improve the drift
response of a sensor it is recommended that it be conditioned
or prestressed prior to use. The sensor manufacturer states
that I-scan sensors are temperature sensitive and typically
measurements vary by up to 0.45% per degree Celsius. The
best control method for temperature is to calibrate the sensor
at the same temperature as the measurement.
The quantification of eyelid pressure will provide a better
understanding of its role in tear film spreading, along with
corneal and contact lens biomechanics.
A novel method to
measure upper eyelid pressure using a thin (0.17 mm) tactile
piezoresistive pressure sensor (Tekscan Pty Ltd, I-scan,
#4201) attached to a contact lens is described in this paper. To
be able to use the sensor to measure eyelid pressure, it was
necessary to design a contact lens to which the sensor could be
attached. An apparatus was constructed to calibrate the sensor
output. To understand the output response of the sensor, the
property drift and the influence of temperature were examined.
A measurement apparatus was developed to accurately and
safely place the sensor-contact lens combination in the eye to
measure upper eyelid pressure.
disinfected in the same way as other ocular instruments (using
mediswabs, 70% isopropyl alcohol). Although applying the
tape altered the sensitivity of the sensor, it was taken into
account by calibrating with the tape in place.
So that the sensor and contact lens would remain stable on
the eye (without rotating or flexing, which can cause false
readings), a support beam was attached to the centre of the
contact lens using a non-toxic glue. A flat area was filed on
the contact lens periphery so that the sensor could be mounted
flat from the support beam onto the contact lens (Figure 1).
The tail of the sensor was attached to the support beam with
double sided tape, while the active part of the sensor was
adhered to the contact lens surface using Histoacryl, medical
cyanoacrylate glue, which has FDA approval, and is
commonly used to seal skin and corneal lacerations [24]. The
active part of the sensor was positioned 3 to 6 mm from
contact lens centre, which is the approximate position of the
upper eyelid relative to the corneal centre for primary
horizontal gaze through to 40˚ downward eye gaze [25].
II.
METHODS
A.
Sensor-contact lens combination
It has been reported that curving Tekscan sensors can cause
an offset and decreased sensitivity of the sensor‟s output [19].
Therefore for the output to result only from the applied
pressure to the sensor and not curvature changes, the sensor
must be fastened in a set shape on a non-flexible surface. It
was therefore attached to a specially designed rigid contact
lens (Capricornia Contact Lens Pty Ltd, Brisbane, Australia)
(Figure 1). A generic back surface shape was used based on
the average of 100 healthy young subjects with a radius = 7.8
mm and prolate eccentricity Q = -0.25 [21]. A lens diameter
of 15 mm was found to be large enough so that both eyelids
maintained their position on the contact lens, increasing the
stability of the lens on the eye. As the contact lens diameter
was larger than the average corneal size (11.7 mm, [22]),
peripheral curves were included in the contact lens design so
that the lens back surface more closely matched the flatter
sclera. The contact lens had a centre thickness of 0.5 mm
which is well above the 0.13 mm critical minimum thickness
to avoid flexure of a rigid (Perspex) contact lens [23]. The
front surface of the contact lens was manufactured with a flat
central area of 6 mm diameter, with a normal peripheral shape
so that the eyelids could still easily slide over the contact lens
surface.
To attach the sensor to the contact lens it first needed to be
trimmed to the appropriate size of 9 cells (3x3). Trimming the
sensor meant that it was no longer sealed from the tear film.
So a layer of „aqua film‟ medical tape, which is water and
bacteria proof, very thin and flexible, was placed around the
entire sensor. Additionally this medical tape can be
Figure 1: Flow chart of constructing the sensor-contact lens combination.
Process of 1) grinding plastic support beam, 2) gluing support beam to contact
lens using epoxy, 3) grinding and polishing a peripheral flat area on lens, 4)
trimming the sensor, 5) covering the sensor with aqua-film tape, 6) gluing the
active part of the sensor to the peripheral flat lens area and taping the tail to
the support beam.
B.
Calibration apparatus
A novel calibration apparatus was designed since
commercially available systems could not apply sufficiently
low pressures and did not allow the thicker sensor-contact lens
combination to be inserted under the pressure applying plate.
A hydrostatic pressure calibration system was designed using
a column of water placed on the sensor (Figure 2). The
sensor-contact lens combination was held with the sensor
perpendicular to the water column and at the same height as
the base tube. By lowering the water column onto the base
tube, the water column acted on the sensor via the plastic
membrane at the bottom end of the water column tube. Using
the density of water = 997.296 kg/m
3
@ 24°C and acceleration
to be 9.79 m/s² @ 27° latitude, a 7.04 cm column of water
applied 5.17 mmHg to the underlying surface. This was
confirmed by placing the water column tube on a balance and
using the increase in mass from the water tube and the area
Authorized licensed use limited to: Politechnika Slaska. Downloaded on May 29, 2009 at 04:19 from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
TBME-00056-2009.R1
3
loaded from the sensor software to independently determine
the pressure applied. An increase in mass on the balance of 14
grams loaded over 2.0 cm² of the sensor is comparable 5.15
mmHg.
The same hydrostatic pressure calibration system was used
for conditioning or prestressing the sensor. The influence of
conditioning was examined by applying a measurement load
of 7.8 mmHg on three occasions. The sensor was then
conditioned with loads of 26 mmHg applied four times, each
for 1 minute with 30 seconds break between loads. The
measurement load of 7.8 mmHg was then once again applied
three times. The importance of the magnitude of the
conditioning load and the length of time between conditioning
and measurement were also examined with conditioning loads
of 10.3, 25.9, and 51.7 mmHg and breaks of 10, 30 and 60
minutes between conditioning and measurement.
Each cell of the pressure sensor was individually calibrated
as each varies significantly in both offset and sensitivity. This
means that the raw score equivalent for a certain pressure can
vary significantly from cell to cell. Raw scores can also vary
from day to day, so calibration is required prior to every
measurement. Custom calibrations for other Tekscan sensors
typically have used 3 or 10 point polynomial fits, which were
shown to be more accurate than the linear or the power law
options used in the Tekscan software [26]. The calibration
process involved randomly applying loads of 1, 2, 2.5, 3, 3.5,
4, 5, 6, 8 and 10 mmHg respectively on two occasions. The
raw score data was averaged for each load between 10 and 30
seconds after loading and the best fit polynomial calibration
was calculated.
within 10% of this average was determined.
The influence of temperature has been previously managed
by keeping the room temperature close to the measurement
temperature (for example at the skin surface) [27]. By placing
temperature and pressure sensors inside an incubator, the
effect of changing temperature on the pressure output was
investigated. The output was monitored as the temperature
was increased from 22°C to 39°C and then decreased back to
22°C. Also 7.8, 10.3, 12.9 and 15.5 mmHg loads were
measured in the incubator for room temperature (23˚C) and at
an average ocular surface temperature of 36˚C [28-31], so that
the error associated with calibrating at 23°C but measuring at
36°C could be estimated.
D.
Eyelid pressure in vivo measurement apparatus
For measurement on the eye, the sensor-contact lens
combination needed to be stabilized so it could not translate or
rotate with respect to the eyelid position. The plastic support
beam (attached to the sensor-contact lens combination) was
fastened to a ball joint so its orientation could be easily altered
(Figure 3). Two video cameras provided front and side
recording of the sensor-contact lens being placed onto the eye
(Figure 3). The front (en face) recording was saved directly to
the computer that controlled the pressure measurement system
so that it was synchronised with the pressure data. The side
camera was connected to a monitor so that the image could be
viewed as the contact lens was placed onto the eye. The
sensor-contact lens and video cameras were mounted on a
platform attached to a slit-lamp biomicroscope base which
allowed their position and height (x, y and z planes) to be
adjusted simultaneously and the sensor to be accurately
positioned under the upper eyelid margin (Figure 3). This
research was approved by the university Human Research
Ethics Committee and all subjects gave informed consent
before participation.
Figure 2: Water column height calibration system. Inset shows system
without water column in place.
C.
The likely influence of drift and temperature on in vivo
eyelid pressure measurements
To examine sensor drift, ten loads (1, 2, 2.5, 3, 3.5, 4, 5, 6,
8 and 10 mmHg) were each randomly loaded two times and
the pressure recorded for at least 20 seconds. The mean raw
score between 15 and 20 seconds after loading was calculated
for each load and the time taken for the output to remain
Figure 3: Apparatus for upper eyelid pressure measurements.
Authorized licensed use limited to: Politechnika Slaska. Downloaded on May 29, 2009 at 04:19 from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
TBME-00056-2009.R1
4
III.
RESULTS
300
3 loads before
conditioning
3 loads after
conditioning
A.
Calibration
The average raw score output for a pressure cell for loads of
1, 2, 2.5, 3, 3.5, 4, 5, 6, 8 and 10 mmHg (each applied twice,
average between 10 and 30 seconds after the load is applied) is
shown in Figure 4. For this pressure range of the sensor,
linear regression provided a good fit to the data (for this
example the coefficient of determination, R² = 0.96).
250
200
150
100
50
12
y = 0.0415x + 1.2088
R² = 0.96
0
11
0
25.6
51.2
76.8
102.4
128
153.6
179.2
204.8
230.4
325
350.6
376.2
401.8
427.4
453
478.6
504.2
529.8
Time (seconds)
10
Figure 5: Six loads of 7.8 mmHg, three applied before conditioning and three
applied after conditioning.
9
8
7
300
6
5
250
4
3
200
2
1
150
0
100
0
25 50 75 100125150175200225250275300
50
Pressure raw score
0
Figure 4: Calibration data for 1 cell with applied pressure between 1 and 10
mmHg.
0 1 2 3 4 5 6 7 8 9101112131415161718192021222324
Time (secs)
B.
Conditioning
Preconditioning the sensor showed evidence of regulating
the measurement response. The response of the sensor was
variable prior to conditioning, whereas after conditioning the
application and removal of loads is obvious with a more
consistent response for each of the three loads (Figure 5).
When assessing the magnitude of the conditioning load as a
variable, it was found that after conditioning with 10.3 mmHg
the output for the 7.8 mmHg measurement took longer to
reach a stable level compared with conditioning with the
higher 25.9 or 51.7 mmHg loads (which gave similar results).
There was no significant effect of the length of the interval
between conditioning and measurement (10, 30 or 60 minutes)
on the output response.
Figure 6: Drift curves for loads of 2, 3, 4, 6 and 8 mmHg over approximately
22 seconds. Arrows indicate when the load remained within a 10% range
around the 15-20 second average.
The sensor‟s output increased only very slightly over the
17˚ temperature range. When comparing calibration curves at
23°C and 36°C, the error associated with calibrating at room
temperature (23°C) and measuring at ocular surface
temperature (36°C) was a slight overestimation of pressure
(average 2.5%).
D.
Examples of upper eyelid pressure measurements
Sample eyelid pressure measurements are shown in Figure
7 where the upper eyelid is placed on the sensor and removed
three times. This shows obvious and consistent response of
the sensor each time the pressure is applied by the upper
eyelid. A further example is the effect of tightening the upper
lid, which is achieved by a technique similar to the lid-pull
technique for removing a rigid contact lens. The sensor output
shows that the pressure applied by the upper eyelid increases
when the lid is pulled (Figure 8).
C.
Sensor properties: drift and temperature
For the first 3 seconds the sensor response was noisy
(Figure 6). After this, some small amount of drift or creep
occurred in the sensor output. The average time for all loads
to remain within 10% of the 15 to 20 second average was 10.4
seconds.
Authorized licensed use limited to: Politechnika Slaska. Downloaded on May 29, 2009 at 04:19 from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
TBME-00056-2009.R1
5
maximum load to be applied in testing and should involve
materials of similar compliance to the application. The
benefits of conditioning are reduced drift and hysteresis and
increased reliability. The importance of conditioning was
demonstrated with loads applied before and after prestressing
the sensor (Figure 5). From investigation of the conditioning
load and break time between calibration and measurement, it
was concluded that the model #4201 sensors should optimally
be conditioned with four loads of 25.9 mmHg for 1-minute
(with 30-second intervals between loads), less than 60 minutes
prior to use.
Over the relatively short time required to measure eyelid
pressure (< 1 minute), the influence of drift is insignificant
provided that the time after loading is matched for calibration
and measurement data (for example between 10 and 30
seconds) and the first 10 seconds (when the output is unstable)
is disregarded.
From the experiments concerning the influence of
temperature, it was concluded that temperature does not have
to be taken into account in the calibration and measurement of
eyelid pressure. Only small errors were recorded when a
measurement was taken at ocular surface temperature but
calibrated with data recorded at room temperature. It is also
questionable whether the sensor would heat up to ocular
surface temperature while on the eye as the piezoresistive
conductive ink inside the sensor is covered with Mylar plastic
sheets. Mylar (polyester) is known for its excellent
temperature resistance, with a coefficient of thermal
conduction of 0.0001. So it should act as an insulator for the
pressure sensitive ink. Unlike the temperature experiments
where the whole sensor was placed in the incubator for a
number of hours, for eyelid pressure measurements the sensor-
contact lens combination is only in contact with the eye and
eyelid for a few minutes. Heat from the eye would also be
absorbed by the Perspex contact lens, further limiting the
influence of temperature variations on the sensor. Therefore
the effect of temperature on the Tekscan sensors reported by
this study is most likely to be an overestimation of its
influence.
Some electrical interference with the sensor output was
observed from a fluorescent light, from the video camera LCD
panel and when using metal clips to secure the sensor to the
apparatus. The result was erratic readings from pressure cells
without the application of pressure. To reduce this
interference the fluorescent ring light was replaced with a desk
lamp which could be positioned further away from the sensor
and proximity to the video cameras was also avoided. Also
the sensor-contact lens combination was attached to the
measurement apparatus with plastic clamps.
Several techniques were developed to calibrate and measure
upper eyelid pressure using piezoresistive sensors. Previous
studies measuring eyelid pressure were disadvantaged by the
complexity of the instrumentation and the techniques
available. Using new piezoresistive sensors means that the
total thickness of the device inserted between the cornea and
eyelid is much smaller, being less than 0.7 mm (approximately
0.5 mm for the contact lens and 0.17 mm for the sensor).
Measurement of eyelid pressure
200
180
Eyelid on pressure sensor
160
140
120
100
80
60
40
20
0
200
250
300
350
400
450
500
550
600
Time (seconds)
Figure 7:
An eyelid pressure measurement with the upper
eyelid being placed on and off the sensor three times.
Figure 8:
Eyelids pulled twice to increase applied pressure.
IV.
D
I
SCUSSION AND CONCLUSIONS
A novel system was developed to use Tekscan #4201 tactile
pressure sensors to measure upper eyelid pressure. This
included designing a custom contact lens, trimming and
resealing the pressure sensor, attaching it to the contact lens
and support beam and filing a flat peripheral area on the
contact lens to which the sensor could be attached. Initially
the sensor-contact lens combination had the support beam
perpendicular to the contact lens so that the sensor lay flat on
the lens, but was bent to run along the support beam.
However pressure measurements were noisy, most likely due
to shearing effects between the back and front Mylar sheets of
the sensor. When the support beam was attached at an angle
so that the sensor remained flat from the support beam onto
the contact lens, the variability in
measurements
was
significantly reduced.
There were a number of advantages of the custom-built
hydrostatic calibration apparatus compared to commercially
available systems. The sensor could be calibrated when
attached to the contact lens, despite its thickness and shape.
Using a plastic membrane at the end of the water column to
contact and conform to the sensor surface closely resembles
the contact applied by an eyelid. Also there were no lower
pressure limits imposed by the water column calibration
apparatus and this was important since we anticipated eyelid
pressure to be relatively low. Calibration should be completed
prior to every use of the sensors with a linear fit for pressure
data between 1 and 10 mmHg.
The Tekscan I-scan manufacturer recommends that sensors
that are new or haven‟t been used for a length of time should
be exercised by loading them three to five times. For best
results it is advised that the load be 20% greater than the
Authorized licensed use limited to: Politechnika Slaska. Downloaded on May 29, 2009 at 04:19 from IEEE Xplore. Restrictions apply.
[ Pobierz całość w formacie PDF ]