Simulating the Effects of Melanin and Air Gap Depth on the

Accuracy of Reflectance Pulse Oximeters

Miodrag Bolic

School of Electrical Engineering and Computer Science, University of Ottawa, 800 King Edward Avenue, Ottawa, Canada

Keywords: PPG, Skin Pigmentation, Skin Color, Melanin, Racial Bias, Motion Artifacts, Monte Carlo Simulation,

Oxygen Saturation, Reflectance Pulse Oximeter.

Abstract: The work aims to understand the effects of skin color and air gap depth on the accuracy of oxygen saturation

estimation. In this paper, this is done by simulating light propagation through the tissue. It is very important

to understand light propagation through the tissue when designing a reflectance pulse oximeter to know what

tissue layers are illuminated by the LEDs, how to position the emitter and the detector depending on the

measurement site, and what kind of signal is expected at the photodetector. This knowledge could also

contribute to developing robust pulse oximeters whose accuracy does not depend on a subject’s skin color.

Our simulation results confirm a larger variation of SpO

for lower saturation levels for dark-skinned subjects

if the SpO

calibration curve is mainly obtained based on measurements from light-skinned subjects. Also, if

the device is calibrated with a small air gap, increasing the air gap will result in overestimated SpO

1 INTRODUCTION

The major difficulty when using pulse oximeter

devices is motion artifacts. Besides motion artifacts,

pulse oximeters are inaccurate in poorly oxygenated

patients (Webster, 1997). This is because there is a

lack of data for fitting the parameters of the models

used in pulse oximeters for patients whose oxygen

saturation levels are below 70%. A similar lack of

data for calibrating the SpO

curve has been noticed

for darker-skinned individuals, which might suggest

the existence of racial bias. There has been significant

interest in racial bias in pulse oximetry. A recent

review analyzed the influence of skin pigmentation

on the accuracy of pulse oximeters (Fuentes-

Guajardo, 2022). It was pointed out that there is

growing evidence that pulse oximeters are less

accurate in dark-skinned individuals at oxygen

saturation levels lower than 80%, resulting in

overestimations of the oxygen saturation.

Development of tools for simulating light

propagation through non-homogeneous tissues based

on Monte-Carlo methods allowed for computational

approaches in analyzing the effects of different

parameters and skin properties on the distribution of

light in the tissue. One of the first works analyzing

https://orcid.org/0000-0002-8013-8645

light propagation through homogeneous and non-

homogeneous tissue is (Tuchin, 1997). More recent

works include (Reuss, 2004), (Reuss, 2005) and

(Chatterjee, 2017).

In this paper, we address issues of estimating

oxygen saturation levels in reflectance pulse

oximeters for people of different skin colors and, in

the case of the different distances between the probe

and the tissue, using a model-based design. We

simulate light propagation through the tissue using

Monte Carlo simulation-based library called

MCMatlab (Marti, 2018).

The pulse oximeter probe contains LEDs and a

photodetector placed on the same side of the tissue so

that the pulse oximeter operates in the reflectance

mode. LEDs with 660 nm and 940 nm wavelengths

were used. Six layers of skin with different optical

properties were simulated. In addition, we added an

air gap between the skin and the simulated probe. The

light from the LED gets scattered through the tissues,

and the photodetector collects only a small percentage

of it. We simulated different blood volumes during

the systolic and diastolic instants of the cardiac cycle,

as well as different oxygen saturation levels of blood.

All these parameters affect the percentage of light

collected by the photodetector. After obtaining the

Bolic, M.

Simulating the Effects of Melanin and Air Gap Depth on the Accuracy of Reﬂectance Pulse Oximeters.

DOI: 10.5220/0011749300003414

In Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 1: BIODEVICES, pages 64-71

ISBN: 978-989-758-631-6; ISSN: 2184-4305

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

fraction of illuminated light from the photodetector

for the light at 660 nm and 940 nm, we computed the

ratio of the fractions of illuminated light (called the

ratio of ratios) and used it to estimate the oxygen

saturation SpO

(Bolic, 2023).

We were especially interested in:

• Analyzing the effects of different levels of

melanin on estimated SpO

• Analyzing the effect of changing the air gap

depth on estimated SpO

The analysis showed the following results:

• For a given setup, the slope of estimated SpO

vs.

the ratio of ratios curve differs for different

melanin levels. This might explain why there is a

bias in estimating oxygen saturation in darker

skin individuals.

• Changing the air gap causes large changes in the

ratio of ratios which then results in inaccurate

SpO

estimates.

Besides the Introduction, the paper contains four

additional sections. Section 2 provides an overview

of the state-of-the-art. Section 3 describes the

simulation setup. Section 4 presents the results.

Finally, the discussion of the results, conclusion, and

future directions are presented in Section 5.

2 BACKGROUND

In this section, we will review the state-of-the-art

from several points of view, including racial bias in

pulse oximetry, simulation of light propagation

through the tissue, and simulation of the effects of

skin pigmentation and motion artifacts on pulse

oximetry.

The existence of melanin in the skin significantly

increases the absorption of light and reduces the

fraction of illuminated light received at the detector.

The absorbance of 660 nm light by melanin is

hundreds of times greater than the absorbance of light

at the same wavelength by blood- and melanin-free

tissues.

Evaluating the accuracy of existing pulse

oximeters for darker skin patients was mainly

performed by comparing the pulse oxygen saturation

against the reference SaO

measurements. The

experiments were done on subjects whose oxygen

saturation levels were controlled. The results are

mixed, but several studies showed an overestimated

SpO

at lower saturation levels (Bickler, 2005) for

darker skin patients.

Light propagation simulators have been

developed, including MCMatlab (Marti, 2018), and

used to simulate the effect of melanin on SpO

accuracy. Monte Carlo simulations were performed to

simulate a transmittance-based pulse oximeter

(Arefin, 2022) for different skin color subjects.

Virtual patients’ finger tissues were generated by

modifying the parameters of the tissue. Oxygen

saturation was in the range of 70% to 100% for each

simulated patient (Arefin, 2022). Monte Carlo light-

tissue interaction model that investigated changes in

melanin level was proposed for reflectance pulse

oximetry (Al-Halawani, 2022). The results show only

the photon penetration depth and absorbance and not

the effect of melanin on the ratio of ratios and SpO

estimation.

It is important to control the pressure when

attaching the reflectance probe to the skin. It was

shown that over the range of pressure applied to the

pulse oximetry probe, the DC amplitude and the

normalized pulse area changed significantly (Teng,

2004).

Motion artifacts can be caused even by small

movements, such as typing if the probe is placed on

the wrist, or breathing if the probe is placed on the

shoulder. Motion artifacts can corrupt the signal in

different ways. They can cause baseline shift and/or

induce changes in the morphology of the signal.

Motion artifacts result mainly from the changes in the

depth of the air gap between the probe and the tissue

with motion. The changes in depth cause changes in

the amount of light absorbed in the tissue as well as

the amount of light scattered and reflected from the

tissue. Motion artifacts are mainly low-frequency

interference. The work that included air gap changes

in the Monte Carlo model assumed three types of

motion: fast arm swing, slow arm swing and typing

(Zhou, 2020). The sensor was used in reflectance

mode and was placed on the wrist. Recently, several

works have been done on characterizing the type of

noise or motion artifacts of photoplethysmogram

(PPG) signals obtained when people perform

different activities such as sitting, walking and

running (Cajas, 2020).

3 THE MODEL AND

SIMULATION SETUP

3.1 Simulating Light Propagation

Through the Tissue

The first step in simulating light propagation through

the tissue is to describe the experiment, which

includes the modeled tissue and its properties, the

Simulating the Effects of Melanin and Air Gap Depth on the Accuracy of Reﬂectance Pulse Oximeters

light source and the detector. In the next step, the light

is simulated as photons that are absorbed or randomly

scattered as they travel through the medium. The rate

of absorption and scattering depends on the

absorption and scattering coefficients of each layer of

the skin. This light distribution is described by the

solution to the radiative transfer equation, which is

solved using Monte Carlo (MC) methods (Wilson,

1983).

Using tools such as MCMatlab, the

implementation of the radiative transfer equation

solver is provided so that the user only needs to

describe the experiment and interpret the simulation

results. MCMatlab is an open-source Matlab-based

software for modeling light interaction with

biological tissue. MCMatlab includes the radiative

transfer equation solver for finding the light

distribution in complex media and a thermal solver,

which was not used here. It is user-friendly and

contains many examples that can be modified and

adjusted to a particular problem. However, even

though the authors claim significant improvements in

computational time compared with other tools (Marti,

2018), MCMatlab is based on Monte Carlo

simulations and takes significant time to complete a

simulation.

The simulation is mainly based on the work

described in (Reuss, 2004) and (Reuss, 2005). The

simplified model of a reflectance PPG probe consists

of an emitter and a detector. The wavelengths of 660

nm and 940 nm were used in the simulation. Emitter

and detector separation was set to 3 mm. The emitter

(LED) is centered at the coordinates (-0.15 cm, 0 cm,

0 cm) while the detector is placed at (0.15 cm, 0 cm,

0 cm) and normal to the tissue surface – see Figure

1a). On the emitter side, a Gaussian beam of a radial

width of 4 mm was simulated. The Gaussian beam

was selected because it approximates noncollimated

LED (the light does not have parallel rays). The

detector was simulated with a diameter of 2 mm.

Tissue is modeled using several layers and

presented as a cuboid in a 3D Cartesian coordinate

system. The thicknesses and blood volume

distributions in the tissue layers are adapted from

(Reuss, 2005). Human skin is divided into epidermis,

dermis and hypodermis (Figure 1a)). The optical

barrier between the LED and the photodetector is also

modeled and shown in red in Figure 1a). The blue

color layer around the barrier represents the air gap.

Figure 1b) shows the normalized fluence rate of

the collected light at the photodetector. This example

is simulated for 660 nm wavelength, with 10 min

simulation duration, which allows for simulating

about 10



photons. We can see a non-uniform

distribution of light.

Figure 1: a) Modeled skin layers and b) normalized fluence

rate of the collected light. The figure is obtained by running

a modified example provided by MCMatlab (Marti, 2018).

3.2 Modeling Changes in Blood

Volume

Table 1: Sublayers of the skin together with their thickness

and diastolic blood volume.

Layer

# (i)

Layer

names

Sublayer

Thickness

(mm)

Blood volume (

1 Epidermis 0.2 0

2 Dermis 0.2 0

3 Papillary

lexus

0.2 0.0556

4 Dermis

(perfused)

0.8 0.0417

5 Cutaneous

plexus

0.6 0.2037 in

diastole, 0.2454

in systole

6 Hypoderm. 8 0.0417

Since each layer is uniformly thick and has the same

area, their relative volumes are expressed only by

their thicknesses. Modeled thickness 𝑑



of skin layer

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

𝑖 and the relative volume of blood 𝐵



in each layer are

presented in Table 1.

The total blood fraction 𝐵



is defined as the mean

concentration of blood in the total tissue volume

during the diastolic stage, and it is assumed to be 5%.

Therefore, the values of the diastolic blood volume

and the depth of the tissue are selected in a way that

the total blood fraction ends up being 5% by

computing the total blood fraction as:

𝐵



=(𝑑



𝐵







)/𝑑



(1)

where 𝑑



∑

𝑑







Software packages for simulating light

propagation through the tissue do not automatically

include information about the parameters that change

over time. Therefore, modifications must be done to

address variables that change over time, such as

arterial blood volume. The systole was simulated by

increasing blood volume only in the skin layer that

contains larger blood vessels. The increase in blood

volume per different sublayers was based on the work

of Reuss (2005). The arterial pulsation was simulated

by blood volume increase in the cutaneous plexus

layer only by adding the arterial blood. The pulse

fraction 𝐵



is defined as the fraction of the total

volume displaced by the arterial pulse, which is

assumed to be 0.25%. The relative volume of systole

in the cutaneous plexus layer is then computed as

𝐵





=𝐵



+𝐵



𝑑



/𝑑



= 0.2037 + 0.0025 ⋅ 10mm/

0.6mm = 0.2454%. In this way, the relative volume

of blood in the cutaneous plexus changed from 𝐵



0.2037% to 𝐵





= 0.2454%. In all other layers, the

relative volume of blood is the same during systole

and diastole.

In this simulation, about 84% of incident light hits

the cuboid boundaries, and 16% is absorbed within

the cuboid. Out of the total incident light hitting the

boundaries of the cuboid, about less than 1% is

detected by the detector placed on the skin’s surface.

Selected simulation parameters for dermis layers

include the water content W = 0.65, Mie scattering

coefficient 1.0, and anisotropy g=0.9.

3.3 Modeling the Level of Melanin

The level of melanin is initially set to 0.3%

(Chatterjee, 2017), which is ten-fold below what is

commonly found in lightly-pigmented human skin.

The reason was to collect a larger illumination

fraction than the one collected for the skin with larger

levels of melanin. The other melanin levels that are

simulated are 3%, 8%, 12% and 16%. The barrier and

the airgap depths were kept close to zero, which

simulated the situation without the air gap.

The experiment is set to take 20 minutes to obtain

one data point on the graph in Figures 2-3. In Figure

2, there are 44 data points resulting in an 8-hour long

simulation. These experiments were done for

mentioned five levels of melanin. Ideally, the

experiment should run for more than 20 minutes per

point and be repeated multiple times to reduce the

results’ variability. The reason for variability is that

the photodetector actually collects a very small

percentage of light, and if the experiment is short, the

illumination fraction will vary a lot.

3.4 Modeling Changes in Air Gap

Depth

The effect of different amounts of pressure applied to

the probe is simulated by changing the depth of the

air gap. Simulation is performed starting with the air

gap of 0.05 mm, which corresponds to the situation

without the air gap, to 0.2 mm, which corresponds to

a small air gap for a smartwatch. The barrier between

the LED and the photodetector is simulated as well.

Without the barrier, the majority of light collected by

the photodetector would be reflected from the skin

surface. Motion artifacts are commonly simulated as

changes in the air gap depth; therefore, this simulation

could be further extended to the simulation of motion

artifacts. The melanin level was set to 0.3%.

4 RESULTS

4.1 Oxygen Saturation vs. the Level of

Melanin

Figure 2 was obtained by simulating the interaction

of light with the tissue for the experimental setup

described previously when oxygen saturation was set

to 60%, 70%, 80%, 90%, 95% and 100%. The blood

volume was changed in the cutaneous plexus layer to

simulate the increase of blood volume during the

systole. Therefore, there are two different curves for

each wavelength, one obtained during systole and

another obtained during the diastole period of the

cardiac cycle. Figure 2a) shows the relative

values of the illumination fraction

𝐼



(



)

,𝐼



(



)

,𝐼



(



)

, and 𝐼



(



)

vs. SpO

for a melanin level of 3% that corresponds to lighter

skin color. I



(



)

and I



(



)

are the relative intensity of

light for a particular wavelength (red 660 nm and

infrared 940 nm) detected during systole or diastole.

Simulating the Effects of Melanin and Air Gap Depth on the Accuracy of Reﬂectance Pulse Oximeters

Figure 2b) shows the relative values of

𝐼



(



)

,𝐼



(



)

,𝐼



(



)

, and 𝐼



(



)

vs. SpO

for a melanin level of 12% that corresponds to darker

skin color. The illumination fractions are reduced

when the melanin level is increased because more

light gets absorbed with increased melanin levels.

Figure 2: Illumination fractions of light versus oxygen

saturation levels obtained at the detector for 660 nm and

940 nm wavelengths for the melanin levels of a) 3% and b)

12%.

Figure 3 shows the illumination franctions for the

case when melanin levels take values of 0.3%, 3%,

8%, 12% and 16% for oxygen saturation of 95%. The

illumination fractions for both wavelengths are

reduced when the melanin level is increased.

The ratio of ratios 𝑅 is computed as

𝑅=

𝐼



(



)

−𝐼



(



)

𝐼



(



)

𝐼



(



)

−𝐼



(



)

𝐼



(



)

We use the illumination fraction instead of the

light intensity to compute 𝑅. After 𝑅 is computed,

𝑆𝑝𝑂



is estimated using regression. An example of

linear regression is shown below:

𝑆𝑝𝑂



=𝐶



+𝐶



⋅𝑅

where 𝐶



and 𝐶



are regression coefficients.

Figure 3: Illumination fraction of light versus the level of

melanin for 95% oxygen saturation level obtained at the

detector for 660 nm and 940 nm wavelengths.

For each ratio of ratios obtained using different

levels of melanin, we fit different curves. Fitted

curves of SpO

versus the ratio of ratios for each level

of melanin, are shown in Figure 4. The fitted curve

for a melanin level of 8% (yellow line) is an outlier.

Figure 4: SpO

versus the ratio of ratios computed for

different melanin levels.

Next, let us consider the following scenario. The

simulated data was collected for the case of a low

level of melanin (3%), and the regression curve SpO

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

vs. the ratio of ratios was fitted for this melanin level.

Then, new simulated data (illumination fraction at

660 nm and 940 nm at systolic and diastolic

instances) was collected for subjects with other

melanin levels.

Figure 5: Error in estimated SpO

versus simulated (ground

truth) SpO

computed for different melanin levels.

To compute the SpO

for these subjects, the

regression coefficients computed for the subject with

3% melanin were used. The error between the

computed oxygen saturation levels versus true,

simulated oxygen saturation when melanin levels

take values of 0.3%, 3%, 8%, 12% and 16% are

shown in Figure 5. The abscissa is the simulated SpO

at the saturation levels of 60%, 70%, 80%, 90%, 95%

and 100%. The ordinate shows the SpO

errors for

these simulated levels of oxygen saturation. There are

several outliers in the figure, including a large outlier

at SpO

=80% for a melanin level of 0.3%. The

number of outliers could have been reduced if the

simulation time was longer.

4.2 Oxygen Saturation Changes with

Air Gap Depth Changes

Figure 6 shows the change of the illumination fraction

for both red and infrared light at the points of minimum

and maximum blood volume (diastolic and systolic

points) for an oxygen saturation level of 95% and a

very low level of melanin. The ratio of AC and DC

components decreases with the increase of the air gap.

The ratio of ratios is computed from Figure 6 and

used to estimate SpO

which is shown in Figure 7.

Since the SpO

is fitted using a curve based on almost

no air gap, we see that the SpO

gets overestimated as

the depth of the air gap increases. The simulation was

performed with SpO

=95%. As can be seen in Figure

Figure 6: Simulated changes in the illumination fraction

with an increased air gap between the probe and the tissue

for SpO

=95%.

7, for larger values of the air gap, we are getting

oxygen saturation levels of more than 100%, which is

not possible. In a realistic device, an algorithm would

limit the SpO

level to a maximum of 100%. Devices

are normally calibrated for a fixed air gap depth.

However, when the air gap depth changes due to

motion artifacts, it causes significant changes in the

ratio of ratios and errors in estimating SpO

Figure 7: Estimated SpO

for the air gap depth of 0.05 mm,

0.1 mm, 0.15 mm and 0.2 mm with a simulated barrier

placed between the emitter and the detector for SpO

=95%.

The estimated values of SpO

for larger air gaps are outside

the bounds (larger than 100%).

5 DISCUSSION AND

CONCLUSIONS

This work builds on the modeling works of (Reuss,

2005) and the simulator design of (Marti, 2018). It

Simulating the Effects of Melanin and Air Gap Depth on the Accuracy of Reﬂectance Pulse Oximeters

analyzes the effects of melanin and the depth of the

air gap on the ratio of ratios and the accuracy of SpO

estimates. The main observations from this work and

future directions are discussed next.

It can be observed from Figure 4 that the slopes of

SpO

vs. the ratio of ratios curves are different for

different skin colors. Therefore, if we could estimate

the tone of the subject’s skin, we would be able to

adjust the curves or select the appropriate regression

curve for calibration. In addition, if the SpO

calibration curve computed for subjects with a lower

level of melanin is used for subjects with a larger level

of melanin, the results from Figure 5 do not confirm

that the overestimation would occur. This might be

because of outliers of SpO

estimates at 0.3% and 8%

melanin levels. The results, however, show variability

with an increased level of melanin.

The illumination is significantly reduced with the

increased level of melanin, as shown in Figure 3. It

drops faster for red than for infrared light. This

information can be used to increase the radiated

power of the LEDs in case a darker skin color is

detected.

In reflectance pulse oximetry, it is difficult and

sometimes impossible to control how well the probe

is attached to the skin. This can result in a varying air

gap. Increasing the air gap results in an increased

level of reflection of light from the surface of the skin,

resulting in less light propagating in the tissue and a

weaker AC component of the signal. Figure 6 shows

changes in illumination fraction at the photodetector

with increasing the air gap and barrier depth. Figure

7 shows changes in the estimated SpO

for the same

case. Therefore, the pressure applied to the device

should be controlled and constant to allow for a fixed

air gap. Also, motion artifacts cause varying air gaps;

therefore, they should be detected, and the signal

obtained during motion artifacts should be removed

from the analysis.

The study can be extended in several ways. One is

to repeat measurements for different points and to run

Monte Carlo simulations longer to get less noisy data.

We can also modify the tissue model to more

precisely reflect the body site where the probe is

attached. For example, the optical properties and

thickness of the skin layers on the finger, ear, wrist,

neck, or forehead (which are some of the typical sites)

are different.

The simulated depth of the blood vessels can be

modified based on the selected site. In the simulation,

we assume that the blood volume changes occur only

in the cutaneous plexus. It was also observed

(Mannheimer, 2004) that placing the reflectance

pulse oximetry probe directly over a larger blood

vessel can degrade SpO

estimation accuracy. This

can also be modeled.

Future work will also include building an end-to-

end pulse oximeter simulator that will include the

model of the LEDs and driving circuits, the model of

the tissue and light propagation and the model of the

photodiode and the front-end electronics. This will

allow us to understand further how noisy the signal is

in case of small levels of illumination of the

photodetector and to perform sensitivity analysis to

understand what components/parameters of the

system affect the output the most.

REFERENCES

Al-Halawani, R., Chatterjee, S. & Kyriacou P.A. (2022).

Monte Carlo Simulation of the Effect of Human Skin

Melanin in Light-Tissue Interactions. 2022 44th Annual

International Conference of the IEEE Engineering in

Medicine & Biology Society (EMBC).

Arefin, M.S., Dumont, A.P. & Patil, C.A.(2022). Monte

Carlo based simulations of racial bias in pulse oximetry.

Proc. SPIE 11951,Design and Quality for Biomedical

Technologies XV, 1195103.

Bickler, P.E., Feiner, J.R & Severinghaus, J.W., (2005).

Effects of Skin Pigmentation on Pulse Oximeter

Accuracy at Low Saturation, Anesthesiology, 102:

715–9.

Bolic M. (2023). Pervasive Cardiovascular and Respiratory

Monitoring Devices: Model-Based Design. Elsevier.

Cajas, S. A., Landínez, M. A. & López, D. M. (2020)

Modeling of motion artifacts on PPG signals for heart-

monitoring using wearable devices. 15th International

Symposium on Medical Information Processing and

Analysis, 11330(3).

Chatterjee, S. & Phillips, J. P. (2017). Investigating optical

path in reflectance pulse oximetry using a multilayer

Monte Carlo model. Clinical and Preclinical Optical

Diagnostics, edited by J. Quincy Brown, Ton G. van

Leeuwen, Proc. of SPIE-OSA, 10411.

Fuentes-Guajardo, M., Latorre, K., León, D. & Martín-

Escudero, P. (2022). Skin Pigmentation Influence on

Pulse Oximetry Accuracy: A Systematic Review and

Bibliometric Analysis. Sensors, 22(3402).

Hartmann, V., Liu, H., Chen, F., Qiu, Q., Hughes, S. &

Zheng D. (2019). Quantitative Comparison of

Photoplethysmographic Waveform Characteristics:

Effect of Measurement Site. Front Physiol., 10(198).

Mannheimer, P.D., et al. (2004). The Influence of Larger

Subcutaneous Blood Vessels on Pulse Oximetry.

Journal of Clinical Monitoring. 18, 179-88.

Marti, D., Aasbjerg, R. N., Andersen P. E. & Hansen A. K.

(2018). MCmatlab: an open-source, user-friendly,

MATLAB-integrated three-dimensional Monte Carlo

light transport solver with heat diffusion and tissue

damage. Journal of Biomedical Optics, SPIE, 23, 1-6..

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

Reuss, J. L., Siker, D. (2004). The pulse in reflectance pulse

oximetry: Modeling and experimental studies. Journal

of Clinical Monitoring and Computing, 18, 289–299.

Reuss, J.L. (2005). Multilayer modeling of reflectance

pulse oximetry. IEEE Trans Biomed Eng., 52(2), 153-

159.

Teng, X.F. & Zhang, Y.T. (2004). The effect of contacting

force on photoplethysmographic signals. Physiological

Measurement, 25(5), 1323-1335.

Tuchin, V.V. (1997). Light scattering study in tissues.

Physics—Uspekhhi, 40(5), 495–515.

Webster J. G. (Ed.) (1997). Design of Pulse Oximeters. IOP

Publishing Ltd.

Wilson, B.C. & Adam, G. (1983). A Monte Carlo model for

the absorption and flux distributions of light in tissue.

Med. Phys., 10(6), 824 – 830.

Zhou, C.C., Wang, J.Y., Qin, L.P. & Ye, X.S. (2020)

Model Design and System Implementation for the

Study of Anti-motion Artifacts Detection in Pulse

Wave Monitoring. Proceedings of the 13th

International Joint Conference on Biomedical

Engineering Systems and Technologies, 1, 102-109.

Simulating the Effects of Melanin and Air Gap Depth on the Accuracy of Reﬂectance Pulse Oximeters