Design and validation of a prototype to calibrate the pressure

variable in tonometry •
Uriel Pineda-Zapata a & Juan Fernando Daza-Fernández b
a
Grupo GIEIAM, Facultad de Ingeniería, Universidad Santiago de Cali, Cali, Colombia. [email protected]
b
Facultad de Ingeniería, Universidad Santiago de Cali, Kaika, Cali, Colombia. [email protected]

Received: November 20th, 2020. Received in revised form: July 6th, 2021. Accepted: July 30th, 2021.

Abstract
This paper corresponds to the results of an experimental research that to led to the design of a useful prototype for quality control of
measurements in the field of tonometry to diagnose and check glaucoma. All tonometry techniques are based on Goldmann´s tonometers
(GT) and the Imbert-Fick principle. Taking into account the principles and techniques of industrial metrology, the prototype could be used
as a standard to calibrate the pressure readings of this class of tonometers. The developed protoype eliminates the indirect measurement
that is required with current standards and facilitates a greater control of the sources of error in the measurement. The model is based on
the integration of a sensor to a hydraulic model to measure pressures from 0 to 80 mmHg range, whose precision was validated through
R&R studies. The precision of the prototype was validated according to the Rigel UNI-SIM standard and was complemented by calibrating
15 tonometers.

Keywords: tonometry; metrology; uncertainty in measurement; calibration.

Diseño y validación de un prototipo para calibrar la variable presión

en tonometría
Resumen
Este artículo corresponde a los resultados de una investigación experimental que derivó en el diseño de un prototipo útil para el control de
calidad de las mediciones en el campo de la tonometría para diagnosticar y controlar el glaucoma. Todas las técnicas de tonometría se
basan en los tonómetros de Goldmann (TG) y en el principio de Imbert-Fick. Teniendo en cuenta los principios y técnicas de la metrología
industrial, el prototipo podría usarse como un estándar para calibrar las lecturas de presión de esta clase de tonómetros. El prototipo
desarrollado elimina la medición indirecta que se requiere con los patrones actuales y facilita un control mayor de las fuentes de error en
la medición. El modelo se basa en la integración de un sensor a un modelo hidráulico para medir presiones de 0 a 80 mmHg de alcance,
cuya precisión se validó a través de experimentos R&R. La precisión del prototipo fue validada según el estándar Rigel UNI-SIM y se
complementó mediante la calibración de 15 tonómetros.

Palabras clave: tonometría; metrología; incertidumbre en la medición; calibración.

1. Introduction membrane is measured by the counter force which just

flattens the membrane to a plane” [2]. Based on his studies on
Intraocular Pressure (IOP) is the fluid pressure inside the corpses’ eyes, Goldmann added tear and cornea parameters
eye, commonly measured through applanation tonometry. to this principle so that it could be applied to human eyes,
First introduced by Goldmann [1], applanation tonometry is thus building an instrument for measuring intraocular
based on the Imbert–Fick principle which states, according pressure: The "Goldmann tonometer" (GT). These
with Duke-Elder, that “the pressure in a sphere filled with tonometers are widely accepted in usual practice [3,4].
fluid and surrounded by an infinitely thin and flexible

How to cite: Pineda-Zapata, U. and Daza-Fernández, J.F., Design and validation of a prototype to calibrate the pressure variable in tonometry,. DYNA, 88(218), pp. 185-193,
July - September, 2021.
© The author; licensee Universidad Nacional de Colombia.
Revista DYNA, 88(218), pp. 185-193, July - September, 2021, ISSN 0012-7353
DOI: https://doi.org/10.15446/dyna.v88n218.91715
 Pineda-Zapata & Daza-Fernández / Revista DYNA, 88(218), pp. 185-193, July - September, 2021.

These tonometers are calibrated using standard weights 2. Subject and materials
provided by the manufacturers, where 1 g = 10 mmHg. For
example, manufacturer companies like Haag-Streit A., provides a Based on the experimental method, the design stages for
weight bar engraved with rings at 0, 2, and 6 g. Therefore, based this prototype were as follows: Initial construction of four
on the abovementioned conversion, this standard calibrates the hydraulic models; selection of an optimum hydraulic model
tonometer at 0, 20, and 60 mmHg, respectively. Also, other from the previous stage and selection of the appropriate
brands on the market B, provide four weights, of 2g each, to pressure sensor for the selected model. Then, design of the
calibrate the tonometers at the points of 20, 40, 60 and 80mmHg. circuit associated with the pressure sensor, digital processing
Therefore, manufacturers disagree on the calibration patterns. In of the output signal from the circuit and, finally, assembly,
fact, some manufacturers provide different tonometer validation and assessment of the prototype.
calibrations according to the model, as observed in their The first step involved selecting materials with structures
corresponding manuals A. similar to the anatomical eye for the preparation of the initial
GTs must comply with a standardized calibration hydraulic models.
procedure, as specified in requirement 7.2 of standard [5]
regarding the selection, verification and validation of the 2.1. Construction of the hydraulic models
methods in metrology field. In this context, a new
standardization procedure was implemented [6]. This Cornea is the tissue that comes in contact with the GT
procedure verified method traceability by directly comparing applanation prism. Therefore, the technical specifications of
the weights provided by the manufacturers against the soft contact lenses were reviewed to select the appropriate
calibrated standards provided by accredited laboratories. contact surface for the prototype. Considering their
Although GTs are devices designed to measure pressure, they approximation to the corneal parameters involved in
are calibrated indirectly using weight standards. tonometry, the characteristics of these soft contact lenses
Furthermore, GTs are used in combination with slit were compared with that of cornea, as shown in Table 1. The
lamps. In these setups, both devices must be correctly aligned specifications were taken from [13,14]
as any deviation may generate source measurement errors. Four hydraulic models were assembled using four different
As such, other sources of error not identified in the lenses readily available in the market. These lenses were attached
calibration procedures must also be considered. [7-10] For to spherical ECG suction cups, which were hollow inside. In
example, misalignment between the GT and the slit lamp, accordance with a previous study [15], the dimensions of these
adjustment issues in the GT feeler arm, and the condition of suction cups were similar to the average axial length of the eye.
the GT measuring prism. These factors are usually not
assessed during weight-based calibration, which focuses on Table 1.
the mechanics of the tonometer and not on its optics. As Contact Lenses vs. Cornea.
"even the slightest calibration error can adversely affect Central Young's Water
Anterior
treatment decisions for patient with advanced glaucomatous LENS Thickness Modulus Content
Spherical
damage"[7]. These possible sources of error must be assessed Radius
(µm) (MPa) (%)
(mm)
to achieve greater accuracy in the calibration procedure and, Cornea 544 0.17 to 0.40 78 7.5 to 8.2
hence, in tonometry. EV3E 180 0.50 74 8.6
Regarding the development of new tonometry techniques, EV3D 180 0.40 55 8.6
other authors have built calibration assemblies in an attempt EV3F 84 0.30 48 8.3
to verify the pressure measured by their equipment; for EV3C 70 0.40 47 8.3
Instance, [11] where contact lenses were used to monitor IOP EV3G 80 1.20 33 8.6
EV3A and
on enucleated porcine eyes connected to a pressure sensor EV3B
70 0.60 38 8.4
through a cannula. These tests needed to be performed Source: The authors.
quickly, before the eyes lost their biomechanical properties.
In another study, [12] the applanation surface of the GT’s
measuring prism was modified to eliminate errors caused by
the biomechanical variability of the cornea. Indeed, the
central corneal thickness varies from person to person and
decisively influences the measurement of intraocular
pressure (IOP) with conventional techniques [13].
For this study, the authors used an assembly similar to the
one used by [11], but used corpses’ eyes. Both studies proved
that models can be constructed where pressure is regulated
and the results can be compared with the measurements
provided by tonometers. Using experimental methods, this
paper proposes the design, construction, and validation of a
calibration prototype for GTs. However, the structure of this
prototype is different from that of the common standard,
which uses weights provided by the corresponding
Figure 1. Calibration Test with Hydraulic Models.
manufacturers.
Source: The Authors.

186
 Pineda-Zapata & Daza-Fernández / Revista DYNA, 88(218), pp. 185-193, July - September, 2021.

2.2. Selection of hydraulic mode

The models were tested as shown in Fig. 1. Each model

was connected to an insufflation bulb and an analog
pressure gauge through a hose, duly calibrated and
certified to measure pressures from 0 to 300 mmHg.
As part of this test, pressure was applied to each model
through the bulb, monitored on the referential pressure
gauge and compared against a duly certified GT. For each
model, three different measurements were recorded at 20,
40, 60, and 80 mmHg, as per the procedures described in Vcc and VDD: Supply voltages.
the corresponding manuals[A] [B]. Figure 2. Basic Circuit for Improving Prototype Resolution.
Source: Higuchi, Hotta, and Iwabuchi, 1992.
According to the results from the previous stage, a
repeatability and reproducibility study, used for to
identify variability in measurement due to instrument [16]
2.4. Circuit calculation
was conducted at seven points (20, 30, 40, 50, 60, 70, and
80 mmHg) for the two models that exhibited the greatest
To leverage the maximum possible number of ADC
accuracy. In addition, two operators were used, taking
converter combinations for the DAQ system, given that the
three measurements per operator for a total of 42 readings
resolution of the measurement system must be optimized, a
per model. In this way, the variability due to the device
circuit for introducing gain and offset values to the sensor output
or model was obtained. The expression of repeatability
signal was designed. The offset converts the minimum signal
and reproducibility estimates in terms of standard
value to 0 V and the gain converts the maximum signal value to
deviations.
5 V. For these purposes, a circuit based on two operational
amplifiers, was used, as shown in Fig. 2, according to [18,C].
2.3. Selection of pressure sensor
The amplifier on the right (seen Fig 2.) carries out the
conversion process. The sensor input is connected to the non-
The criteria for sensor selection were accuracy and
inverting pin and the input (R1) and feedback (R2) resistors
availability, considering that pressures to be measured ranged
are connected to the inverter. An Offset voltage is connected
between 0 and 80 mmHg (with a resolution of 2 mmHg), as
to the input resistance, coming from a potentiometer.
per the corresponding GT scale. In addition, the selected
sensor utilizes a silicon, micromachined, stress concentration
The circuit’s characteristic equation is as follows:
enhanced structure to provide a very linear output to
measured pressure, according to the manufacturer's manual. 𝑅𝑅2 𝑅𝑅2
Consequently, the technical specifications of seven 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 = 𝑉𝑉𝑉𝑉𝑉𝑉 ∗ � + 1� − 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 ∗ (3)
𝑅𝑅1 𝑅𝑅1
different sensors were assessed.
In addition, to process the output signal from each sensor, However, it can be simplified as follows:
Arduino UNO R3® data acquisition (DAQ) card was used;
its input resolution was estimated as follows: 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 = 𝑉𝑉𝑉𝑉𝑉𝑉 ∗ (𝐺𝐺 + 1) − 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 ∗ 𝐺𝐺 (4)

𝑅𝑅𝑅𝑅𝑅𝑅 =
𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉−𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉
(1) Where:
2ᶰ

𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉: 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉

Where: 𝑉𝑉𝑉𝑉𝑉𝑉: 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉
𝑁𝑁: 𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁 𝑜𝑜𝑜𝑜 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵. 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉: 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 𝐺𝐺: 𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺
𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉: 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉- 𝑉𝑉𝑉𝑉𝑉𝑉𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓: 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉
As we used a 10-bit ADC (Analog to Digital Converter)
converter whose input voltage ranged between 0 and 5 V, the 2.5. Digital signal processing
resolution of the DAQ was 4.88 mV, according to Equation
(1). Using the test block shown in Fig. 3, a correlation analysis
After expressing the sensitivity of the selected sensor in between the circuit’s output signal and the pressure applied to the
mV/mmHg its resolution can be determined using a simple model was performed. This signal was coupled to an analog input
rule of three. This calculation verified that the sensor was from the DAQ card, followed by reading the corresponding input
indeed suitable for tonometer calibration as per the criteria after applying 1000 readings (approximately) to dampen the
described by [17], which establishes that a calibration process noise. This information was stored in a variable and was
is successful when resolution ratio (R), as given by Equation subsequently transferred to the computer.
(2), is ≧4, and preferably 10. The signal was displayed as a digital variable that was
proportional to the pressure, with values ranging between 0
𝑅𝑅 =
𝐼𝐼𝐼𝐼( 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟)
(2) and 1023. To convert these values to mmHg, the response
𝑆𝑆𝑆𝑆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟)
from the sensor was linearized using the function that defines
the behavior of a straight line:

187
 Pineda-Zapata & Daza-Fernández / Revista DYNA, 88(218), pp. 185-193, July - September, 2021.

2.6. Prototype assembly

The prototype was constructed following the schematics

depicted in Fig. 4. To induce pressure in the model in a
controlled manner, a millimetric screw was coupled to a 3-
mL syringe.

2.6.1. Prototype validation and assessment

The prototype was validated by Metrology Laboratory at

A: Manometer, range 0-300mmHg. B: One-way valve. C: Inflated the University, where the values—20, 30, 40, 50, 60, 70, and
pear. D: Sensor + Circuit. E: DAQ Card. F: Voltmeter. 80 mmHg—were simulated using a duly certified Rigel UNI-
Figure 3. Test Block for Pressure Sensor. SIM vital signs simulator. Then, these results were
Source: The Authors. statistically assessed in a normal distribution, calculating the
corresponding errors at each measurement to obtain the
uncertainty in measurement.
𝑌𝑌 = 𝐴𝐴𝐴𝐴 + 𝐵𝐵 (5) Next, Equation (8) determined the combined uncertainty
(µc) for the prototype using the known uncertainties from
At this stage, to obtain ample number of samples for an both electronic block (µe) and GT (µt) used for the hydraulic
accurate linearization, the digital values were read in model measurements during the R & R study. In this manner,
increments of 2 mmHg, from 0 to 80 mmHg. In Equation (5), this uncertainty value was compared against the uncertainty
variable Y is the digital number read by the DAQ card values reported by the laboratory.
corresponding to a pressure value of X. Next, these readings
were tabulated and graphed to determine the resulting 𝜇𝜇𝜇𝜇 = �(𝜇𝜇𝜇𝜇)2 + (𝜇𝜇𝜇𝜇)2 (8)
characteristic function. Then, the equation obtained after
solving for variable X, can be implemented in the To ascertain prototype compliance, its uncertainty was
programming code to convert these data to mmHg. compared against the ISO 8612 standard [19], which
Subsequently, Equation (6) was used to calculate the establishes a ±2.5-mmHg tolerance in tonometer readings.
mean absolute deviation (MAD) of the readings as well as the Subsequently, the prototype was evaluated by calibrating 15
standard deviation (Sx) of the mean absolute errors. GTs at a renowned clinic in the Valle del Cauca region. The
Afterwards, type A uncertainty (µa) for the electronic block procedures issued by the GTs manufacturers were followed
was determined using Equation (7). Finally, to achieve to record the tonometry readings, according to the
greater accuracy, we performed two additional rounds of recommended measurement points. Finally, an assessment of
linearization following the steps from the previous method. the readings from the tonometer calibrations was conducted
∑ |𝑋𝑋𝑋𝑋−𝑋𝑋𝑋𝑋|
using the prototype and the results were compared against the
𝐷𝐷𝐷𝐷𝐷𝐷 = (6) calibration data from the current standards.
𝑛𝑛

Where:
𝑋𝑋𝑋𝑋: 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 Table 2.
𝑋𝑋𝑋𝑋: 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 Results of the Goldmann tonometers calibration test with the Hydraulic
𝑛𝑛: 𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁 𝑜𝑜𝑜𝑜 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 Models Contact Lenses vs. Cornea.
EV3A, EV3B,
𝑆𝑆𝑆𝑆 PP M1, M2, M3, �
𝑿𝑿 Sx M1, M2, M3, �
𝑿𝑿 Sx
𝜇𝜇𝜇𝜇 = (7) mmHg mmHg
√𝑛𝑛
20 22 24 22 22.67 1.15 22 22 24 22.67 1.15
40 42 42 44 42.67 1.15 42 44 44 43.33 1.15
60 62 64 64 63.33 1.15 64 64 64 64.00 0
80 82 82 84 82.67 1.15 84 84 84 84.00 0
PP: Pressure point (mmHg).
Source: The Authors.

Table 3.
Results of the Goldmann tonometers calibration test with the Hydraulic
Models Contact Lenses vs. Cornea.
EV3C EV3D
PP M1, M2, M3, �
𝑿𝑿 Sx M1, M2, M3, �
𝑿𝑿 Sx
mmHg mmHg
A: Hydraulic model. B: Applanation prism of the Goldmann 20 24 24 26 24.67 1.15 20 20 22 20.67 1.15
tonometer. C: Contact lens. D: Sensor + Circuit. E: DAQ Card. F: 40 44 46 46 45.33 1.15 40 40 40 40.00 0
PC/Display. G: Pressure exerted by the liquid. H: Aqueous humor. I: 60 64 64 66 64.67 1.15 60 62 60 60.67 1.15
Pressure inducing device. J: Electronic block. 80 86 86 84 85.33 1.15 82 80 80 80.67 1.15
Figure 4. Prototype Diagram. PP: Pressure point (mmHg).
Source: The Authors. Source: The Authors.

188
 Pineda-Zapata & Daza-Fernández / Revista DYNA, 88(218), pp. 185-193, July - September, 2021.

3. Results and discussion Table 5.

Intermediate Results from the R&R study.
EV3A EV3D EV3A EV3D
Price and commercial availability were the selection criteria POINTS �1 �2 �1 �2
R1 R2 R1 R2 𝑿𝑿 𝑿𝑿 𝑿𝑿 𝑿𝑿
for the four lenses used in the hydraulic models. As shown in (mmHg)
mmHg mmHg mmHg mmHg
Table 1, the EV3E lens is the closest to the corneal parameters, 20 0 0 0 0 18.0 18.0 20.0 20.0
but it is an expensive lens and its delivery time depends on its 30 0 2 0 0 32.0 30.7 30.0 30.0
manufacturing specifications, whereas the EV3D lens is more 40 2 0 0 0 41.0 39.0 40.0 40.0
affordable and available for immediate delivery. The EV3C and 50 1 0 1 0 51.3 50.0 49.3 50.0
EV3A lenses were selected on the basis of the same criteria. From 60 0 1 1 0 62.0 60.3 59.3 60.0
this last type, two lenses with different positive optical powers 70 2 2 1 1 73.0 70.7 69.3 70.3
80 1 2 1 0 83.3 81.3 78.7 79.0
(EV3A and EV3B) were selected to assess the effect of the shape
Source: The Authors.
of lenses (from the same manufacturer and material) in tonometry
readings. Then, the appropriate solution was selected to
accurately simulate aqueous humor in the models. Table 6.
Once the models were built, the test as described in Technical Specifications of Pressure Sensors.
Section 2.2 was conducted. Tables 2 and 3 list the Range Accuracy Output Sensitivity
Sensor
measurement readings at each point. Three readings are (mmHg) (±%) Range (mV) (mV/mmHg)
required normally.A Longer readings can cause eye fatigue. A 0 to 375.03 0.50 10 to 40 0.110
The first three models used positive lenses as, according to B −7.35 to 88.26 0.25 280 to 4500 47.070
−120.01 to
[20,21], corneas can be considered as converging lenses. C
120.01
0.25 500 to 4500 12.500
Therefore, as depicted in Fig. 5 [20], the selected lenses had an −187.52 to
D 5.00 250 to 4750 12.000
optical area with a convergent aspherical surface and a smoothing 187.52
zone with the following optical powers, in diopters (D): +4.25D E 0 to 300.02 0.70 0 to 100 0.330
(EV3A), +6.50D (EV3B), +7.00D (EV3C). −50.03 to
F 2.00 0 to 1.75 0.005
300.02
During the test, we found that aspherical lens with higher G 0 to 367.53 0.30 0 to 100 0.272
optical power exhibited greater central thickness with regard to Source: Technical Datasheets Supplied by the Manufacturers
their periphery. This interferes with tonometry readings as
greater force must be applied to flatten the surface, a situation
that can lead to measurement errors, according to [22]. This On the contrary, according to Table 3, the best
finding is reflected in the values reported in Table 2 for the performance was reported for the EV3D model, which uses a
EV3A and EV3B models, which use lenses from the same 0D lens. As there is no difference in the thickness between
manufacturer but with different optical powers, and validated this lens’ optical zone and its periphery, the applanation
by the values reported for the EV3C model, Table 3, a more process is easier and more uniform. Input data for the R&R
powerful model. The difference in pressure between the EV3A study obtained from Goldmann tonometer Calibration using
and EV3C models can easily be observed. the EV3A and EV3D Models Therefore, the EV3A and
EV3D models were selected for the R&R study. Their input
data is reported in Table 4.
The corresponding R&R study parameters were
discussed in Section 2.2, and the intermediate results from
variation ranges (R1 and R2) and averages (X) for each
operator are described in Table 5.
The expression of repeatability and reproducibility
estimates in terms of standard deviations (S). The device
variation errors (by repeatability and reproducibility) were
Figure 5. Aspherical Lens with Smoothing Zone. estimated based on the results reported in Table 4. The results
Source: Alm and Kaufman, 2003. from this variation-error estimation, using the calculation
method of [16,23], were: SEV3A = 1,458 mmHg and SEV3D =
0.39819 mmHg. Hence, the EV3D model was selected as it
Table 4.
produced the least margin of error. The following formula
Input Data for the R&R study.
was used:
EV3A EV3D
Point Operator1 Operator2 Operator1 Operator2
(mmHg) M1 M2 M3 M1 M2 M3 M1 M2 M3 M1 M2 M3 𝑆𝑆𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 = �(Srepet)2 + (Sreprod)2 (9)
mmHg mmHg mmHg mmHg
20 18 18 18 18 18 18 20 20 20 20 20 20
30 32 32 32 30 32 30 30 30 30 30 30 30
where:
40 41 40 42 39 39 39 40 40 40 40 40 40
(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)2 : 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑜𝑜𝑛𝑛 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
50 51 52 51 50 50 50 49 50 49 50 50 50 (𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)2 : 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
60 62 62 62 60 61 60 60 59 59 60 60 60
70 72 74 73 70 72 70 70 69 69 71 70 70 After selecting the hydraulic model, the best pressure sensor
80 83 83 84 82 80 82 78 79 79 79 79 79
was selected after reviewing seven technical datasheets supplied
Source: The Authors.
by the manufacturers, as shown in Table 6.

189
 Pineda-Zapata & Daza-Fernández / Revista DYNA, 88(218), pp. 185-193, July - September, 2021.

Figure 6. Comparison between Sensors’ and Prototype’s Ranges. (Range of

prototype: 0 to 80 mmHg).
Source: The Authors.

Figure 7. First Linearization for the Variable Transformation.

Sensor B displays greater accuracy than sensors A, E, and Source: The Authors.
G, as shown in Fig. 6. The three discarded sensors (A, E, and
G) feature a wider measurement scale, so the signal had to be
amplified to reach the input range of the DAQ card; this Then, using the test block from Fig. 3, after displaying the
affected the instrument resolution. However, the measuring circuit output on the computer, the first linearization was
range for Sensor B adjusts better to the requirements; therefore, performed. Its corresponding equation and graphics are
its output range is approximately the same as its technical provided below.
specification (0.28 to 4.5 V), which facilitates circuit design as
complex amplification stages are not required. Equation Fig. 7:
Then, based on its sensitivity (47.07 mV/mmHg), the
resolution of Sensor B was estimated at 0.104 mmHg using 𝑌𝑌 = 11.446𝑋𝑋 − 8.0523. (12).
the rule of three described in Section 2.3.
In according to Equation (2), and considering that the Fig. 7 shows that X-axis corresponds to the pressure
tonometer’s resolution (IR) is 2 mmHg and the sensor’s applied to the sensor, whereas the Y-axis represents the
resolution, in this case pattern resolution (PR), is 0.104 digital response, which ranged between 0 and 1023, from
mmHg, the ratio R calculated is: the electronic block to said pressure. Solving for variable
X from Equation (12) our digital values were converted to
R=(IR/PR) = 2.000/0.104, this is, mmHg using the DAQ card programming code. Next, the
R=19.230, which satisfies the criteria (R ≧ 10) described mean absolute error, MAD, standard deviation, and type A
by [17]. uncertainty measurements for the electronic block, were
calculated, in according Equations (6) and (7). Table 7 lists
Subsequently, a 2 × 2 equation system (10, 11) was the results from the two subsequent linearization stages;
formulated using Equation (4) by assigning the desired stage two exhibits less uncertainty, which is why the DAQ
equivalent output voltages (Vout)—both minimum and card programming ended there.
maximum—for the circuit to the output voltages from the After the prototype was completed, it was sent to
sensor (Vin): Metrology Laboratory at the University for validation. As
observed in Fig. 8, the uncertainty lies within the range
0𝑉𝑉 = 0.28𝑉𝑉 ∗ (𝐺𝐺 + 1) − 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 ∗ 𝐺𝐺 (10) ±0.11–±0.14 mmHg, thus complying with ISO 8612, which
establishes a ±2.5-mmHg tolerance for tonometer readings.
5𝑉𝑉 = 4.5𝑉𝑉 ∗ (𝐺𝐺 + 1) − 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 ∗ 𝐺𝐺 (11) To verify this, the combined uncertainty for the
prototype was calculated using Equation (8), considering
The solution to this system is G = 0.1838 and Voffset = that the uncertainty of the electronic block (µe) is ±0.090
1.78V. According to Equation (3), G = R2/R1. As a value of 1 mmHg (as per Table 7) and that the uncertainty (µt) of the
KΩ was assigned to R2; accordingly, R1 would be 5,382 KΩ. GT, used in the R&R study, is± 0.085 mmHg at 95%
Next, the circuit was optimized for operation through trial confidence, according to Table 8. Table 8 indicates the
and error, and the calculated components were assembled as pressure measurements in mmHg. The test conditions were
per Fig. 4. Furthermore, as the operational instrumentation performed at 20,4 ° C, with a relative humidity of 55.3%,
amplifiers denoted high accuracy and reliability, two in the metrology laboratory of the University.
AD620Ns were used and two potentiometers were placed to In addition, the lens itself is a source of uncertainty; this
vary resistors R1 and R3, thus obtaining our offset and gain value corresponds to the repeatability component of the R
values. In addition, to reduce noise, 100-nF capacitors were & R study conducted for the EV3D lens, which is ±0.211
placed between the terminals of each voltage source. Given mmHg. Therefore, the total uncertainty of the prototype is
the simplicity of the circuit, there was no need to verify its ±0.244 mmHg, which is well within the stipulated criteria.
operation through a specialized simulation software.

190
 Pineda-Zapata & Daza-Fernández / Revista DYNA, 88(218), pp. 185-193, July - September, 2021.

Table 7.
Results from Each Linearization Stage.
MAD Standard Deviation Type A Uncertainty
|STAGE
(mmHg) (mmHg) (mmHg)
Stage 2 0.92 0.491 0.090
Stage 3 1.16 0.714 0.128
Source: The Authors

Figure 9. Example of Readings Outside the Goldmann tonometer’s

Resolution.
Source: Instructions for use Applanation tonometer AT 900® / AT 870. A

Table 9 describes the results from the assessment conducted

for the prototype. Each measurement point of the variable
pressure (mmHg), was assessed within the corresponding
tolerance range, according to the GT manuals provided by the
manufacturers. For example, the 20-mmHg point was assessed
at 19.5 and 20.5 mmHg; the 40-mmHg point, at 39.5 and 40.5
Figure 8. Prototype Validation.
Source: Laboratory of Metrology, University Santiago de Cali. mmHg; the 60-mmHg point, at 59 and 61 mmHg; and the 80-
mmHg point, at 79 and 81 mmHg.
For each measurement point was determined,
Table 8. respectively: Nominal value, Average, St. deviation,
Data generated from the Goldmann Tonometer calibration certificate. Error, K. factor (95%), Uncertainty and Concept (LES+U;
Nominal Average SD Error K U LES-U). The measurement points 40 and 50 mmHg,
0,0 0,0 0,0 0,0 1,96 0,085 present minus variation, according to intervals in column
0,5 0,5 0,0 0,0 1,96 0,085 “Concept”. Results are satisfactory, because permitted
19,5 20,0 0,5 0,0 1,96 0,085
20,0 20,5 0,5 0,0 1,96 0,085 error maximus (PEM) is 2.
20,5 21,0 0,5 0,0 1,96 0,085 According to Table 8, at the 20-mmHg point, only 53%
59,5 59,5 0,0 0,0 1,96 0,085 of the GTs comply with the specifications from the
60,0 60,0 0,0 0,0 1,96 0,085 manufacturers. Therefore, we infer that the tonometries
60,5 60,5 0,0 0,0 1,96 0,085 conducted using other devices on people with normal IOP
SD: Standard deviation. U: Uncertainty. K: Factor for 95% confidence.
Source: The Authors.
(11–21 mmHg) and in accordance with [24] would be
erroneous, with some even presenting errors exceeding 1
mmHg. This observation is relevant for the purposes of this
Table 9. study as many people with glaucoma may report normal
Results of the Calibrations of 15 GTs Performed with the Prototype. IOP values, especially considering that glaucoma is the
20mmHg, 40mmHg, second leading cause of blindness for worldwide,
Tonometers
Sx �
𝑿𝑿 E
Out
Sx 𝑿𝑿 E
Out according to [25]. As glaucoma is usually diagnosed when
come come eye pressure exceeds 21 mmHg, only 40% of GTs comply
EB0360, 0.29 19.83 −0.17 C(+) 0.58 39.67 −0.33 C(+)
at 40 mmHg. This situation is unsettling as “even the
EB0138, 0.58 19.33 −0.67 C(+) 0.29 40.17 0.17 C
EB0150, 0.29 19.67 -0.33 C(+) 0.00 39.00 −1.00 NC
slightest calibration error can adversely affect treatment
EB0313, 0.29 19.17 -0.83 C(+) 0.29 39.67 −0.33 C decisions for patients with advanced glaucomatous
OVEB-144, 0.00 20.00 0.00 C(+) 0.50 40.5 0.50 C damage” [7].
OVEB-112, 0.29 19.17 −0.83 C(+) 0.00 39.00 −1.00 NC However, after reviewing the calibration certificates issued in
EB0249, 0.00 20.00 0.00 C 0.50 40.00 0.00 C August 2018 for the assessed GTs, all complied with the
EB0378, 0.00 16.00 −4.00 NC 0.29 31.67 −8.33 NC specifications from the corresponding manufacturers, with the
EB0181, 0.00 19.50 −0.50 C 0.00 39.00 −1.00 NC exception of model EB0378. These calibrations were performed
OVEB-185, 0.00 20.50 0.50 C 0.00 39.00 −1.00 NC using the weight standards and specifications provided by the
OVEB-173, 0.29 19.33 −0.67 C 0.50 39.50 −0.50 C(+) manufacturers. After comparing these values against the
OVEB-161, 0.00 20.00 0.00 C 0.29 40.17 0.17 C
developed prototype, it can be concluded that accurate error
EB0280, 0.58 20.33 0.33 C 0.29 39.17 −0.83 C(+)
OVEB-207, 0.29 20.33 0.33 C 0.29 39.83 −0.17 C
measurement can only be achieved using a validated pressure
OVEB-256 0.00 20.00 0.00 C 0.50 37.50 −2.50 NC standard as weight-based calibration does not yield reliable data
C (+): Compliance above Measuring Point. on the offset of GTs.
C (-): Compliance below Measuring Point. Nevertheless, in the procedures included in their manuals,
NC: Non-Compliant within Measuring Range. manufacturers recommend performing calibrations at
C: Compliant within Measuring Range. pressure points that are not within the resolution of the GTs.
Source: The Authors.
For example, in the 20-mmHg pressure point, manufacturers

191
 Pineda-Zapata & Daza-Fernández / Revista DYNA, 88(218), pp. 185-193, July - September, 2021.

recommended verifying between 19.5 and 20.5 mmHg, when References

the GTs have a specified resolution of 2 mmHg, and 0,5
mmHg would be a best resolution. This adds the element of [1] Goldmann, H. and Schmidt, T.H., Über applanationstonometrie.
subjectivity to a tonometry. For example, if a patient presents Ophthalmologica, 134(4), pp. 221-242, 1957. DOI:
10.1159/000303213
a tonometry as the one depicted in Fig. 9, it is not clear [2] Stuckey, G.C., Tonometry and the Imbert‐Fick. Journal of
whether the doctor must use the reading below or the reading Opthalmology, 2(2), pp. 68-70. 1974. DOI: 10.1111/j.1442-
above. Therefore, a more accurate resolution should be one 9071.1974.tb00199.x
of the criteria when purchasing a new tonometer. [3] Cook, J.A., Botello, A.P., Elders, A., Ali, A.F., Azuara-Blanco, A.,
Fraser, C. and McCormack, K. and Burr, J.M., Systematic review of
the agreement of tonometers with Goldmann applanation tonometry.
4. Conclusions Ophthalmology, 119(8), pp. 1552-1557, 2012. DOI:
10.1016/j.ophtha.2012.02.030
During the development of the hydraulic model, the [4] Stamper, R.L., A history of intraocular pressure and its measurement.
behavior of the eye structures involved in tonometry was Optom Vis Sci, 88(1), pp. E16-E28, 2011. DOI:
simulated, which showed that when materials similar to 10.1097/opx.0b013e318205a4e7
[5] ISO/IEC 17025:2017. General requirements for the competence of
these structures are used, a model can be designed whose testing and calibration laboratories; International Organization for
pressure can be measured using a GT. In addition, by Standardization, Geneva, Switzerland, 2017.
integrating the model into an electronic block specifically [6] Tobón, O. y Rodríguez, V., Desarrollo y estandarización de métodos
designed for the given range (0–80 mmHg), suitable de calibración para equipos utilizados en salud visual (Queratómetros,
uncertainty and resolution values could be determined for Lensómetros y Tonómetros), implementados en el Hospital
Universitario de San Vicente Fundación. Revista Ingeniería
the calibration of these tonometers as per the corresponding Biomédica, 11(22), pp. 13–20, 2017. DOI:
ISO standard. 10.24050/19099762.n22.2017.1179
The traceability of the prototype was verified by [7] Choudhari, N.S., George, R. and Vijaya, L., Have you tested your
validation at Metrology Laboratory at the University; this goldmann tonometer today? Journal of Current Glaucoma Practice,
validation can be performed at any accredited laboratory in 2(2), pp.13-16, 2008. DOI: 10.5005/jp-journals-10008-1027.
the country. The assessment of the tonometers revealed a [8] Kirstein, E.M., Elsheikh, A. and Gunvant, P., Tonometry–past,
present and future. Glaucoma–Current Clinical and Research Aspects.
higher degree of accuracy than the commonly used standards InTech, pp.85-108, 2011. DOI: 10.5772/37393
in the industry, thus evidencing shortcomings in weight- [9] Cho, P. and Lui, T., Comparison of the performance of the Nidek NT-
based calibration procedures and tonometries performed 2000 noncontact tonometer with the Keeler Pulsair 2000 and the
using these GTs. Goldmann applanation tonometer. Optometry and vision science:
The contribution from this prototype to GT calibration, in official publication of the American Academy of Optometry, 74(1),
pp. 51-58, 1997. DOI: 10.1097/00006324-199701000-00020
addition to its structure, operation, and the weights currently [10] Moseley, M.J., Thompson, J.R., Deutsch, J., Misson, G.P., Naylor,
used, is evidenced when the prototype is placed on the optical G., Tan-Yee, A. and Fielder, A.R., Comparison of the Keeler Pulsair
mechanism of the tonometer. In addition, to making direct 2000 non-contact tonometer with Goldmann applanation. Eye, 7(1),
pressure comparisons, this study also included the sources of pp.127-130, 1993. DOI: 10.1038/eye.1993.26
error associated with its use with the slit lamp, thus enabling [11] Leonardi, M., Leuenberger, P., Bertrand, D., Bertsch, A. and Renaud,
P., First steps toward noninvasive intraocular pressure monitoring
a comprehensive assessment of the entire instrument, both with a sensing contact lens. Investigative Opthalmology & Visual
mechanically and optically. Therefore, when calibrating 15 Science, 45(9), pp. 3113-3117, 2004. DOI: 10.1167/iovs.04-0015
tonometers, two faulty applanation prisms were replaced, [12] McCafferty, S., Lim, G., Duncan, W., Enikov, E. and Schwiegerling,
three GTs were aligned as they were found to be out of focus, J., Goldmann tonometer prism with an optimized error correcting
and two loose prism arms were adjusted—situations that applanation surface. Translational Vision Science & Technology,
5(5), art.4, 2016. Available at: DOI: 10.1167/tvst.5.5.4
would not have been easy to identify should the tonometers [13] Bowling, B., Cornea. In: Bowling, B., Ed. Kanski's Clinical
have only been calibrated using the weights. Ophthalmology. Elsevier, 2015, pp. 167-237. DOI: 10.1016/b978-0-
According to [7], the error sources for GT calibration are 7020-4093-1.00006-9
rarely investigated, which is why other patterns that may [14] Stapleton, F., Stretton, S., Papas, E., Skotnitsky, C. and Sweeney, D.F.
provide more information on these sources should be Silicone hydrogel contact lenses and the ocular surface. The ocular
surface, 4(1), pp. 24-43, 2006. DOI: 10.1016/s1542-0124(12)70262-
explored. In this manner, tonometry accuracy can be 8
improved and more stringent calibration criteria can be [15] Villate, M.A.M., Méndez, J.D.B. y Echeverry, J.E.P., Anatomía
realized against lax procedures that lead to measurement quirúrgica del ojo: revisión anatómica del ojo humano y comparación
errors. con el ojo porcino. Morfolia, 8(3), pp. 21-44. 2016.
For future studies, we recommend assessing different [16] Montgomery, D.C., Introduction to statistical quality control. 6th Ed.
John Wiley & Sons, USA, 2007.
materials for the hydraulic model, specifically for the surface [17] Pennella, R., Managing the metrology system. Milwaukee, Wis.:
in contact with the cornea, as its parameters significantly ASQ Quality Press, 3rd ed, 2004.
influence tonometry, as suggested by [26]. As evidenced in [18] Higuchi, M., Hotta, M. and Iwabuchi, M., U.S. Patent No. 5,121,119.
Table 3, owing to its optical power, the shape and thickness U.S. Patent and Trademark Office. Washington, DC, USA, 1992.
of the contact lens directly influenced the measurement; in [19] International Organization for Standardization ISO. International
Standard ISO 8612 Ophthalmic instruments — Tonometers.
this case, it contributed to the error. Hence, the selection or International Organization for Standardization, 2009. DOI:
development of a membrane that approaches corneal 10.1109/IEEESTD.2010.5733835
parameters will be a key factor for the development of a [20] Alm, A. and Kaufman, P.L., (Eds.). Adler's physiology of the eye:
pressure standard suitable for measuring GTs. clinical application, 10th Ed., Mosby, St. Louis, USA, 2003. DOI:
10.1001/archopht.121.11.1667

192
 Pineda-Zapata & Daza-Fernández / Revista DYNA, 88(218), pp. 185-193, July - September, 2021.

[21] Fannin, T.E. and Grosvenor, T., Optics of contact lenses. Clinical Pineda-Zapata, U., is a BSc. Eng. in Industrial Engineer, who graduated
optics. Butterworths, Stoneham, MA, 1987. 415-453. DOI: from the University Nacional, Colombia in 1992. He obtained his MSc. in
10.1016/b978-0-409-90060-6.50017-8 Engineering in 2011, University Nacional de Colombia. He is working as an
[22] Andreanos, K., Koutsandrea, C., Papaconstantinou, D., Diagourtas, associate professor of industrial engineering at the University Santiago de
A., Kotoulas, A., Dimitrakas, P. and Moschos, M.M., Comparison of Cali, Colombia. His areas of interest and research included: Statistical
Goldmann applanation tonometry and pascal dynamic contour Process Control, Process Strategic Management, Service Quality,
tonometry in relation to central corneal thickness and corneal Operations Management, where.
curvature. Clinical Ophthalmology, 10, pp. 2477-2484, 2016. DOI: ORCID:0000-0002-4702-6537.
10.2147/OPTH.S115203
[23] McAlinden, C., Khadka, J. and Pesudovs, K., Statistical methods for Daza-Fernández, J.F., is a BSc. Eng. in Bioengineer, from the University
conducting agreement, (comparison of clinical tests) and precision Santiago de Cali in 2019. He is working as technical support professional at
(repeatability or reproducibility) studies in optometry and Kaika S.A.S. His areas of interest included: biomedical, metrology.
ophthalmology. Ophthalmic and Physiological Optics, 31(4), pp. 330- ORCID: 0000-0003-2019-1230.
338, 2011. DOI: 10.1111/j.1475-1313.2011.00851.x
[24] Mansouri, K. et al., Review of the measurement and management of
24-hour intraocular pressure in patients with glaucoma. Survey of
ophthalmology 65(2), pp. 171-186, 2020. DOI:
10.1016/j.survophthal.2019.09.004
[25] Pascolini, D. and Mariotti, S.P., Global estimates of visual
impairment: 2010. British Journal of Ophthalmology, 96(5), pp. 614-
618. 2012. DOI: 10.1136/bjophthalmol-2011-300539
[26] Sródka, W., Evaluating the material parameters of the human cornea
in a numerical model. Acta of Bioengineering and Biomechanics /
Wrocław University of Technology, 13(3), pp.77-85, 2011.

Other cited material

A. Haag-Streit, Instructions for use applanation tonometer AT 900® /

AT 870. Koeniz. [online]. 2017. [Accesed January 18, 2019].
Available at https://www.haag-streit.com/fileadmin/Haag-
Streit_USA/Diagnostics/tonometry/download/01-IFU_AT900-
7006000-04230_eng_01.pdf.
B. Appasamy Associates. Empowerming Vision. [online]. Available
at:https://www.appasamy.com/equipmenthome.php?product=App
a_applanation_tonometer_AATM_5001.
C. Gomez, E., Ajustar señal para ADC y aprovecha el ADC completo.
[online]. 2017. Available at. https://www.rinconingenieril.es/
ajustar-senal-adc/. Accesed Febrero 05, 2019.

193