sensors

Article
Pulse Oximetry Imaging System Using Spatially Uniform Dual
Wavelength Illumination
Riaz Muhammad 1,† , Kay Thwe Htun 1,† , Ezekiel Edward Nettey-Oppong 1,† , Ahmed Ali 1,2 , Dae Keun Jeon 3 ,
Hyun-Woo Jeong 4 , Kyung Min Byun 5,6, * and Seung Ho Choi 1,7, *

1 Department of Biomedical Engineering, Yonsei University, Wonju 26493, Republic of Korea;

[email protected] (R.M.); [email protected] (K.T.H.); [email protected] (E.E.N.-O.);
[email protected] (A.A.)
2 Department of Electrical Engineering, Sukkur IBA University, Sukkur 65200, Pakistan
3 Mediana, R&D Center, Wonju 26365, Republic of Korea; [email protected]
4 Department of Biomedical Engineering, Eulji University, Seongnam 13135, Republic of Korea;
[email protected]
5 Department of Biomedical Engineering, Kyung Hee University, Yongin 17104, Republic of Korea
6 Department of Electronics and Information Convergence Engineering, Kyung Hee University,
Yongin 17104, Republic of Korea
7 Department of Integrative Medicine, Major in Digital Healthcare, Yonsei University College of Medicine,
Seoul 06229, Republic of Korea
* Correspondence: [email protected] (K.M.B.); [email protected] (S.H.C.);
Tel.: +82-33-760-2463 (S.H.C.)
† These authors contributed equally to this work.

Abstract: Pulse oximetry is a non-invasive method for measuring blood oxygen saturation. However,
its detection scheme heavily relies on single-point measurements. If the oxygen saturation is measured
at a single location, the measurements are influenced by the profile of illumination, spatial variations
in blood flow, and skin pigment. To overcome these issues, imaging systems that measure the
distribution of oxygen saturation have been demonstrated. However, previous imaging systems have
relied on red and near-infrared illuminations with different profiles, resulting in inconsistent ratios
between transmitted red and near-infrared light over space. Such inconsistent ratios can introduce
Citation: Muhammad, R.; Htun, K.T.;
fundamental errors when calculating the spatial distribution of oxygen saturation. In this study, we
Nettey-Oppong, E.E.; Ali, A.; Jeon,
developed a novel illumination system specifically designed for a pulse oximetry imaging system. For
D.K.; Jeong, H.-W.; Byun, K.M.; Choi,
the illumination system, we customized the integrating sphere by coating a mixture of barium sulfate
S.H. Pulse Oximetry Imaging System
and white paint inside it and by coupling eight red and eight near-infrared LEDs. The illumination
Using Spatially Uniform Dual
Wavelength Illumination. Sensors
system created identical patterns of red and near-infrared illuminations that were spatially uniform.
2023, 23, 3723. https://doi.org/ This allowed the ratio between transmitted red and near-infrared light to be consistent over space,
10.3390/s23073723 enabling the calculation of the spatial distribution of oxygen saturation. We believe our developed
pulse oximetry imaging system can be used to obtain spatial information on blood oxygen saturation
Academic Editor: Vittorio Passaro
that provides insight into the oxygenation of the blood contained within the peripheral region of
Received: 3 February 2023 the tissue.
Revised: 25 February 2023
Accepted: 16 March 2023 Keywords: pulse oximetry imaging system; spatially uniform dual wavelength illumination;
Published: 4 April 2023 hyperspectral imaging system

Copyright: © 2023 by the authors.

1. Introduction
Licensee MDPI, Basel, Switzerland.
This article is an open access article Oxygen saturation, also referred to as blood oxygen saturation (SpO2 ), is a crucial
distributed under the terms and physiological parameter, along with heart rate, breathing rate, blood pressure, and body
conditions of the Creative Commons temperature [1,2]. The measurement of blood oxygen saturation provides the relative
Attribution (CC BY) license (https:// amount of oxygen present in the blood and serves as an indicator of whether a person
creativecommons.org/licenses/by/ has an adequate supply of oxygen [3]. A decrease in oxygen levels can cause long-term
4.0/). damage to individual cells, the brain, and the heart, as well as short-term malfunctioning

Sensors 2023, 23, 3723. https://doi.org/10.3390/s23073723 https://www.mdpi.com/journal/sensors

Sensors 2023, 23, 3723 2 of 18

of other important human organs [4,5]. Therefore, accurate and frequent measurements
of oxygen saturation are essential for monitoring the well-being of vital organs and for
the early detection of potential health issues. For measuring blood oxygen saturation, a
variety of options including invasive and non-invasive methods are available from both
the literature and commercial devices. In an invasive SpO2 estimation method, the levels of
oxyhemoglobin (HbO2 ) and deoxyhemoglobin (Hb) can be determined through a blood gas
analysis where blood samples are collected, and the number of oxygen-binding hemoglobin
is counted. Although this technique is accurate, it is costly, time-consuming, and invasive.
Pulse oximetry is commonly used as a non-invasive method for measuring the blood
oxygen saturation (SpO2 ) and the heart rate of patients in hospitals and homes [6]. Pulse
oximetry measures the percentage of oxygenated hemoglobin in the arteries [7]. The
measurements determine the supply of oxygen to peripheral tissues and the efficiency
of oxygenation in the pulmonary alveoli [8,9]. During detection, the interaction of light
with the area of measurement is achieved via transmission or reflection methods. In the
transmission method, light incidents on one side of the targeted peripheral region, and the
transmitted light is detected by a photodetector placed either on the opposite side or side-
by-side with the light source, whereas the light source and photodetector are located on
the same side of the targeted region in the reflection method [10,11]. Photoplethysmogram
(PPG) signals, generated by shining red and near-infrared light on the skin, provide
information on both oxygenated and deoxygenated hemoglobin and are commonly used
to quantify blood oxygen saturation. This is achieved by evaluating the ratio of absorption
coefficients of oxyhemoglobin to deoxyhemoglobin, which reflects the saturation level of
oxygen in the blood [12].
Conventional pulse oximetry methods are limited to single-point measurements, and
hence the measurements are influenced by the shape of illumination, spatial variations
in blood flow [13], and skin pigment [14,15]. To overcome these challenges, imaging
systems that measure the distribution of blood oxygen saturation have been demonstrated.
Humphreys and Markham [16] presented a contactless system equipped with a CMOS
camera to capture photoplethysmography (PPG) signals at 760 nm and 880 nm, respectively.
Two lights were alternately illuminated on the fingertip to detect rhythmic arterial pulsation.
Subsequently, the results were compared to those of a conventional contact-based pulse
oximeter. Although the results were comparable, the SNR of the PPG signal was highly
dependent on the location in the peripheral region of the tissue. Al-Naji et. al. [17] have
used the green and red channels of a digital camera to estimate SpO2 values. However, the
estimated SpO2 values are greatly influenced by the lighting conditions.
Moreover, previous camera-based pulse oximetry imaging systems relied on two
different patterns of red and near-infrared illuminations [18–20]. When spatial profiles of
the input red and near-infrared illumination are different, the ratio between transmitted
red and near-infrared light also differs across space [20]. The mismatch between the red
and near-infrared illumination profiles can introduce fundamental errors [21]. Therefore,
spatially uniform dual wavelength illumination is required to eliminate the occurrence of
artifacts due to spatially varying light sources.
In this study, we developed a novel illumination system specifically designed for use
in a pulse oximetry imaging system. The illumination system developed in this study
created two identical patterns of red (630 nm) and near-infrared (940 nm) illuminations that
are spatially uniform. To achieve this unique illumination, we customized an integrating
sphere by coating its interior with a mixture of barium sulfate and white paint.
An integrating sphere, also referred to as the Ulbricht sphere, is an optical device used
for the spatial integration of radiant flux, and thus has a diffusing effect or the uniform
scattering of light. It is composed of a hollow sphere with small openings which serve as
input and output ports for light. The interior of the sphere is coated with a highly reflective
material to obtain diffusely reflecting walls, which enables multiple scattering reflections of
light within the sphere. Incident light rays are scattered multiple times within the sphere,
effectively averaging out any directional variations in the light output. Consequently,
Sensors 2023, 23, 3723 3 of 18

a more homogenous illumination is achieved. The operation of the integrating sphere

is described by a mathematical expression of radiance. The general expression for the
radiance, L, corresponding to an illuminated diffuse surface with an input flux, Φi , is
given as:
Φp
L= i (1)
πA
where p, π, and A are the reflectance, the total projected solid angle from the diffuse surface,
and the illuminated area, respectively [22]. Considering monochromatic and unidirectional
irradiation, the reflectance is described by the ratio of the reflected flux, independent of
direction, to the incident flux from a differential solid angle about the θ and φ direction,
where θ is the polar angle of incidence measured from the normal of the surface, and φ is
the azimuthal angle of incidence measured in the plane of the surface from an arbitrary
reference [23]. Reflectance is expressed in terms of the reflection-distribution function f (θ,
φ0 ; θ 0 , φ0 ) integrated over half-space. The function evaluates the intensity of the radiance
corresponding to the reflected flux in the direction θ 0 , φ0 per unit radiant flux incident in
the differential solid angle about the θ and φ direction. A surface is considered diffuse for a
continuous f and perfectly diffuse for a constant f [23].

Z2π π/2
Z
p(θ, φ) = f (θ,φ0 ; θ 0 , φ0 )sin θ 0 cos θ 0 dθ 0 dφ0 (2)
0 0

For an internally illuminated integrating sphere, both the numerous surface reflections
and losses via the port apertures required for the input flux are considered in the radiance
equation [22].
Φi p
L= × (3)
π As 1 − p (1 − f )
where f is the port fraction. The radiance equation has two components; the first component
is identical to the general expression for radiance (Equation (1)), and the second component
is termed the sphere multiplier [22]. The sphere multiplier, M, is a unitless quantity that
accounts for the radiance increase resulting from multiple reflections.
p
M= (4)
1 − p (1 − f )

The value of M is determined by both the reflectance of the sphere surface and the
port fraction. The port fraction is expressed as a function of the input port area, Ai , the
output port area, Ao , and the sphere wall area, As .

Ai + A o
f = (5)
As

The developed pulse oximetry imaging system incorporates this illumination system
to capture high-precision spatial and temporal data. Using a hyperspectral imaging ap-
proach, we obtained spectral information from different locations within the peripheral
region based on the acquired spatio-temporal data. Hyperspectral imaging is a multivariate
imaging technique. An image made of I rows and J columns acquired over K variables is a
classical multivariate image. The commonly utilized variables are wavelengths, but other
variables can be used as well [24]. By interpolating between the coinciding points of the
dual images captured due to the dual wavelength light source, a spectrum is generated to
characterize different locations within the peripheral region. Utilizing the hyperspectral
imaging approach and fundamental theory of pulse oximetry, a spatial map of the blood
oxygen saturation was produced. Moreover, the developed system overcomes the limita-
tions of existing systems, where the spatial intensity distribution of red and near-infrared
lights is not uniform, yielding approximate blood oxygen saturation measurements.
Sensors 2023, 23, 3723 4 of 18

2. Materials and Methods

2.1. System Design
A transmission pulse oximetry imaging system was developed for measuring the
spatio-temporal distribution of blood oxygen saturation over time. As indicated in
Figure 1a, the developed technology has two major components: an imaging system,
and a light source. The imaging system consisted of a CMOS camera, a zoom lens, an
integrating sphere, and light emitting diodes (LEDs) as the light source. Red (630 nm)
and near-infrared (940 nm) LEDs were used as input light sources for the integrating
sphere. In addition, an Arduino microcontroller was used to generate trigger signals for
the measurements. As demonstrated by the signal patterns in Figure 1a, the LEDs alternate
in response to the trigger signals with a corresponding exposure time for the camera. The
system repeats this operation to capture spatio-temporal images for data processing. A
hyperspectral imaging approach was used to obtain spectral information from the captured
images at wavelengths of 630 nm and 940 nm. Thus, the spatio-temporal distribution data
of blood oxygen saturation is acquired. Figure 1b shows the front and top view of the
experimental set-up of the pulse oximetry imaging system. The following sections further
elaborate on the materials and components utilized in developing the imaging system.

2.1.1. Imaging Setup

The camera, responsible for capturing light from the peripheral region through reflec-
tion or transmission, plays a critical role in developing a pulse oximetry imaging system.
The performance of the camera immensely affects the quality of the acquired images and,
ultimately, the accuracy of the blood oxygen saturation measurement [25]. A monochrome
CMOS camera (model acA1300-60gmNIR, Basler AG, Ahrensburg, Germany) featuring a
1/1.8 in CMOS sensor with a maximum resolution of 1280 × 1024 square pixels of length
5.3 µm and 12-bit encoding was used. The camera was connected to a PC via IEEE 1394 and
utilized a manually adjusted C-mount zoom lens (model 87-536 0.15X-0.5X Non-Telecentric
Lens, Edmund Optics, Barrington, NJ, USA) with a focal length range of 50–179.8 mm
and an aperture of f /2.8–f /22, to focus on the tissue under investigation. Upon receiving
trigger signals, the camera captures a set of frames at 28 frames per second using a global
shutter technique.

2.1.2. Illumination System

An ideal pulse oximetry imaging system requires a light source that exhibits low
spatial intensity variations and maintains a consistent illumination pattern over time and
different wavelengths. In traditional pulse oximetry imaging, inhomogeneous illumination
arises from the inconsistent illumination profile of the input light across various peripheral
locations. This results in imprecise blood oxygen saturation measurements [26]. To over-
come these issues, we created an illumination system by customizing an integrating sphere
to produce uniform illumination at red and near-infrared wavelengths.
The illumination system design aimed to balance light distribution and brightness
while conforming to established guidelines for the proper operation of integrating spheres.
In accordance with these guidelines, commercially available integrating spheres typically
have diameters ranging from 50 mm to 250 mm, with openings that do not exceed 5% of
the sphere’s total area and an angle of light incidence not greater than 10 degrees [27]. To
maximize the light intensity at the exit port of the sphere, following the design guidelines
cited before, a 50 mm diameter sphere was designed featuring 16 input ports of 4 mm
diameter on its sides for LED lights and a 12 mm diameter output port on its top for uniform
illumination emission. The total port area was designed to occupy 4.61% of the sphere’s
surface with a light incidence angle of 10 degrees. The design of the integrating sphere
was realized using a computer-aided design (CAD) and fabricated from white Polylactic
Acid (PLA) material using a 3D printer (model Flashforge Guider II, Zhejiang Flashforge
3D Technology Co., Ltd., Jinhua, China) with a filament thickness of 1.75 mm.
 trigger signals with a corresponding exposure time for the camera. The system repeats
this operation to capture spatio-temporal images for data processing. A hyperspectral im-
aging approach was used to obtain spectral information from the captured images at
wavelengths of 630 nm and 940 nm. Thus, the spatio-temporal distribution data of blood
oxygen saturation is acquired. Figure 1b shows the front and top view of the experimental
Sensors 2023, 23, 3723
set-up of the pulse oximetry imaging system. The following sections further elaborate5 on
of 18

the materials and components utilized in developing the imaging system.

Figure 1.1.Optical/electrical configuration

Optical/electrical and materials
configuration of pulseofoximetry
and materials imaging system:
pulse oximetry imaging(a)system:
Illus-
tration of the proposed system; (b) Front and top views of the experimental setup with hardware
(a) Illustration of the proposed system; (b) Front and top views of the experimental setup with
trigger, PC connection, camera, light source, and power supply; (c) Cross-sectional view of the 3D
hardware trigger, PC connection, camera, light source, and power supply; (c) Cross-sectional view of
the 3D printed integrating sphere coated with the reflectance material (upper left), BaSO4 powder
(upper right), green laser (lower left) and red laser (lower right) projected on the integrating sphere;
(d) Reflectance spectra of white standard and the coating material (paint + BaSO4 ) used for the inte-
grating sphere; (e) Spectra of 630 nm and 940 nm wavelength LEDs, along with the deoxyhemoglobin
and oxyhemoglobin absorption spectra within the 500 nm to 1000 nm wavelength range.

2.1.3. Light Emitting Diodes (LEDs)

The transmission-type pulse oximetry imaging system requires a light source having
both visible and infrared lights with stable and uniform illumination. Red and near-infrared
lights have been used for pulse oximetry systems, since they maximize the penetration for
optical imaging techniques [28]. The illumination source used in this study consisted of
eight red and eight near-infrared light-emitting diodes (LEDs) inserted into an integrating
sphere from all four sides (refer to Figure 1a). The chosen number of LEDs per wave-
length ensured stable, uniform, and intense illumination at red (630 nm) and near-infrared
(940 nm) wavelengths.

2.1.4. Reflectance Coating Material

Barium sulfate (BaSO4 ) is a white, crystalline powder known for its high reflectance
and has been used as a reference standard in optical measurements [29]. It offers a cost-
effective alternative to more expensive white standards and provides uniform reflectance
across the visible and near-infrared wavelength range [28]. Additionally, the crystalline
structure of BaSO4 results in high scattering properties [30]. Hence, it is a suitable choice
for the in-house integrating sphere.
Sensors 2023, 23, 3723 6 of 18

A mixture of Barium sulfate (97.5–100%, SAMCHUN Chemical, Seoul, Republic of

Korea) and white paint (model PXI453901/4L, NOROO Paint and Coatings, Anyang,
Republic of Korea) at a 50:50 volume ratio with 20% water content was prepared. The
mixture was stirred and allowed to settle for 2 h to achieve a uniform mix. The interior of
the integrating sphere was then coated with three layers, each with a specified drying time
of 10 h, to ensure a consistent thickness. The cross-sectional view of the coated integrating
sphere is shown in Figure 1c—upper-left. To evaluate the uniformity of light scattering
on the sphere’s inner surface, green and red laser pointers with wavelengths of 532 nm
(22 mW output power) and 650 nm (8 mW output power), respectively, were utilized. The
lasers were focused on a sectioned piece of the sphere, and the distribution of scattered
light was recorded (see Figure 1c—bottom-left and right). The evaluation of scattered
light on the inner surface of the cut piece of the integrating sphere revealed a uniform
distribution of light, which was indicative of uniform reflection scattering across the entire
surface due to multiple scattering events (see Section 3.1 for a complete analysis of the
illumination pattern).
The reflectance spectra of the mixture of BaSO4 and white paint were measured and
compared with Spectralon white standard (model USRS-99-010, Labsphere Inc., North
Sutton, NH, USA) using an integrating sphere (model 2P4/M, Thorlabs, Newton, NJ, USA)
and a compact CCD spectrometer (model CCS200/M, Thorlabs, USA). The results showed
that a mixture of BaSO4 and white paint provides a reflection of 87% (see Figure 1d), which
is acceptable for multiple scattering purposes. The high reflectance value achieved by the
mixture of barium sulfate and white paint confirms its suitability for use as the reflective
material in the integrating sphere for efficient reflection and scattering of light inside the
integrating sphere.

2.1.5. Microcontroller for Generating Trigger Signals for Camera and Light Source
An Arduino microcontroller was used to control both the red (630 nm) and near-
infrared (940 nm) LEDs and the camera, as shown in Figure 1a. A separate power supply
was required for the 630 nm LEDs due to the higher current draw, which is shown in
Figure 1b. This power supply provided a regulated 12 volts and was switched on and
off by a trigger signal generated by the Arduino through a transistor (model MPS 2222A,
ON Semiconductor, Scottsdale, AZ, USA). In contrast, the 940 nm LED array was pow-
ered directly from the trigger signal generated by the Arduino, eliminating the need for
an additional power supply. Using the Arduino microcontroller to generate trigger sig-
nals ensures precise control of both red (630 nm) and near-infrared (940 nm) LEDs for
accurate imaging.

2.2. Fundamental Theory of Dual Wavelength Pulse Oximetry

Blood oxygen saturation (SpO2 ) is the percentage of oxygen in the arterial blood, and
it is determined using the following formula [31]:

[HbO2 ]
SpO2 = ×100% (6)
[HbO2 ]+[Hb]

where [HbO2 ] and [Hb] are the concentrations of hemoglobin with oxygen and without
oxygen, respectively.
The specific absorption coefficients of oxygenated (HbO2) and deoxygenated hemoglobin
(Hb) have distinct spectral dependence [32], as included in Figure 1e. The difference
between these coefficients allows for the quantification of each constituent, forming the
basis of pulse oximetry [33,34]. At wavelengths below 800 nm, the Hb-specific absorp-
tion coefficients are higher than the HbO2 one, while in the region above 800 nm, the
HbO2 absorptions dominate. As shown in Figure 1e, red light at 630 nm is absorbed
by deoxygenated hemoglobin (Hb) more than oxygenated hemoglobin (HbO2 ), whereas
near-infrared light at 940 nm is absorbed by HbO2 more than Hb. Upon transmission or
reflection through tissue, the red or near-infrared light is attenuated due to absorption by
Sensors 2023, 23, 3723 7 of 18

biological molecules and scattering by larger tissue structures. The modified Beer-Lambert
Law [35] is used to express the intensity of the transmitted or reflected light from the skin
tissue to quantify the light absorption by hemoglobin as follows:

It = Io e−CDα (7)

where It is the intensity of the transmitted light, Io is the intensity of incident light, C is
the concentration of the sample, α is the absorption coefficient of the tissue, and D is the
optical path length (i.e., the distance traveled through the sample). The calculation of
SpO2 based on pulse oximetry assumes that the pulsatile component of optical absorp-
tion is due to the pulsating flow of arterial blood, while the non-pulsatile component
stems from non-pulsating arterial blood, venous blood, and other tissues. The ratio of ab-
sorbances is calculated using the red (630 nm) and near-infrared (940 nm) time signals in the
following equation:
AC630
R630 =
DC630
AC940
R940 = (8)
DC940
R630
RR =
R940
where AC is the pulsatile component, DC is the non-pulsatile component, and RR is
the ratio of ratios of the absorbances at the two wavelengths. Previous studies [36,37]
have shown that the pulsatile and non-pulsatile components correspond to the standard
deviation and mean color intensities of the red and near-infrared frames, respectively. A
nearly linear relationship exists between SpO2 values and RR [38]. Therefore, the value of
SpO2 is estimated from RR according to the equation:

SpO2 = (m × RR)+c (9)

where m and c are empirically determined through linear regression for each volunteer,
m is the slope of the estimated regression line, and c is the y-intercept [39]. The empirical
approximation technique is used to correct errors in the measured values caused by the
assumption of only two substances in the light path [40].

2.3. Image Acquisition

The monochrome CMOS camera was controlled by an Arduino and took images with
a resolution of 1280 × 1024 pixels at a frame rate of 28 frames per second. Images were
captured every 36 milliseconds when either LED was on. This led to a camera trigger rate
of 28 times per second, resulting in a data acquisition rate of 14 frames per second for
each wavelength. Camera exposure time was set at 18 ms to optimize the contrast-to-noise
(CNR) ratio [41,42]. The recorded frames were sent to a personal computer (PC) via IEEE
1394 data communication and stored automatically in external memory as a sequence of
time-series images in TIFF format for future image processing. A buffer was employed
to store the images and prevent data loss during transfer to the PC. All experiments were
conducted in a completely dark environment to minimize noise.

2.4. Image Processing

In this study, image processing was used as a method to calculate the spatial distribu-
tion of blood oxygen saturation through a pulse oximetry imaging system that employs
a dual wavelength uniform illumination light source. We have used MATLAB (Version
R2022b, MathWorks, Natick, MA, USA) to process the time series images.
Once a set of recordings was successfully acquired, the raw image frames were divided
into discrete regions of interest (ROIs) to produce a new set of reduced frames, where the
value of each pixel in the new frame was set as the average of all the pixel values within
each ROI. Though compromising the spatial resolution, such a procedure significantly
 pixels. This resulted in a reduced frame size of 64 × 52 pixels, yielding time series PPG
signals at each pixel position across a sequence of frames.
The mean of the pixel values (16-bit, ranging from 0 to 65,536) within an ROI region
was computed for each frame using the following equation:
Sensors 2023, 23, 3723 X j Y 8 of 18
∑ii 1 ∑j 1 s(i,j,t)
S(t) = (10)
X·Y
where X and
improved theYsignal-to-noise
represent the width and height the
ratio. Averaging of the ROI
pixel region,
values respectively,
within each ROIandis at better
is the
frame sequence. These averaged values were used to obtain the imaging-PPG
sampling technique that preserves more spatial information as compared to capturing signal at the
corresponding wavelength (see Figure 2).
reduced images of the peripheral region. Herein, the ROI size was set at 20 × 20 pixels.
The PPGinsignals
This resulted wereframe
a reduced usedsize
as the physiological
of 64 × 52 pixels, indicators
yielding timeto analyze thesignals
series PPG pulsatileat
changes in blood volume and blood oxygen
each pixel position across a sequence of frames. saturation levels at various locations over
different wavelengths.
The mean Eachvalues
of the pixel of the(16-bit,
two PPG signals
ranging was
from divided
0 to 65,536)into a 10-s
within an subset,
ROI regionand
the AC and DC components of the PPG signals were
was computed for each frame using the following equation: obtained using average peak-to-peak
and mean values, respectively, from each of these subsets (see Figure 2). The minimum
points at each signal were interpolated∑ across −Y spatial locations to remove the baseline
i−X jall
i−1 ∑j−1 s(i, j, t)
from the PPG signals, and the DC =
S(t)component was subtracted from the AC component. (10)
X·Y
Consequently, the RR values were calculated from Equation (8) using the determined AC
where
and DC X components
and Y represent thePPG
of the width and height
signals. of the blood
Finally, ROI region,
oxygen respectively,
saturation and
(SpOt2is the
) val-
frame sequence.
ues were These
extracted averaged
from values RR
the calculated were used at
values to each
obtainspatial
the imaging-PPG
location usingsignal at the
Equation
corresponding
(9). wavelength (see Figure 2).

Figure 2. Image acquisition and data processing for SpO2 calculation: Image acquisition. The image
data is captured alternately at 630 nm and 940 nm in a time series of frames, with a resolution of
1280 × 1024 pixels for each frame. The region above the output port of the integrating sphere was
considered a valid measurement area. Image data processing. The panel illustrates the steps for
calculating the SpO2 values and displays the AC and DC components of the 630 nm and 940 nm
photoplethysmogram (PPG) signals.

The PPG signals were used as the physiological indicators to analyze the pulsatile
changes in blood volume and blood oxygen saturation levels at various locations over
different wavelengths. Each of the two PPG signals was divided into a 10-s subset, and
the AC and DC components of the PPG signals were obtained using average peak-to-peak
and mean values, respectively, from each of these subsets (see Figure 2). The minimum
points at each signal were interpolated across all spatial locations to remove the baseline
from the PPG signals, and the DC component was subtracted from the AC component.
Consequently, the RR values were calculated from Equation (8) using the determined AC
and DC components of the PPG signals. Finally, the blood oxygen saturation (SpO2 ) values
were extracted from the calculated RR values at each spatial location using Equation (9).

3. Results and Discussion

The pulse oximetry imaging system was assessed to determine its efficacy to measure
the spatio-temporal distribution of blood oxygen saturation over time. The performance
of the developed system was evaluated in three phases. First, characterization of the
illumination pattern of the designed integrating sphere to ensure uniform illumination.
Sensors 2023, 23, 3723 9 of 18

Next, analysis of the captured images to ensure photoplethysmogram (PPG) waveforms

are consistent for all the locations within the peripheral region. Finally, acquisition of the
spatio-temporal information of the distribution of blood oxygen saturation to determine
physiological parameters including heart rate, respiratory rate, and oxygen saturation
of tissues. The outcomes were compared to existing pulse oximetry imaging systems in
literature and commercial devices.

3.1. Illumination Pattern Assessment

Herein, a key focus was on developing an illumination system with a highly uniform
profile that remains constant over time to achieve precise measurements of blood oxygen
saturation. By utilizing the working principle of an integrating sphere, a virtual light source
was developed to overcome this challenge. The integrating sphere uniformly distributes
light over the entire inner surface of the sphere, creating a highly homogenous spatial
illumination pattern. The dimensions of the designed integrating sphere are illustrated
in Figure 3a. Substituting the dimensions and the determined reflectance (87%) of the
integrating sphere into Equations (4) and (5), the port fraction, f, and the sphere multiplier,
M, were calculated to be 0.04 and 5.279, respectively.
The determined sphere multiplier value indicates that the fabricated integrating sphere
ensured multiple reflections of the incident lights, resulting in an overall increase in ra-
diance. Thus, the fabricated integrating sphere used in the developed pulse oximetry
imaging system was efficient in yielding high radiance. The uniform distribution of light
within the sphere ensures a consistent uniform profile independent of time and changes in
wavelength. As shown in Figure 3b, the output light from the designed integrating sphere
was uniformly distributed over the entire outport port area for both input light sources at
630 nm and 940 nm. The realized uniform light profile is also illustrated in Figure 3c. The
3D profiles of the output light demonstrate the homogenous spatial illumination pattern
obtained from the designed integrating sphere.
In addition, the acquired homogeneity ensures that the light source is consistent with
the input trigger signals. This consistency is also critical for accurate measurements of
SpO2 in pulse oximetry, as any discrepancies between the light source and the input trigger
signals can lead to inaccuracies in the measurements. Figure 3d presents an assessment by
utilizing a photodetector and a camera to demonstrate the compatibility between the light
source and the input trigger signals. The timing diagram from the photodetector confirmed
that fluctuations in the timing or intensity of the light source relative to the input trigger
signals were trivial. Thus, the light source was highly consistent with the input trigger
signals, with minimal variations. The images captured by the camera further demonstrated
consistency by revealing that the homogenous spatial illumination pattern produced by
the integrating sphere was capable of uniformly illuminating the sample.
Additionally, we characterized the transmission pattern of our illumination system.
To achieve this, we placed a thin diffuser sheet over the output port of the integrating
sphere. Next, we alternately turned the input LEDs ON and OFF and captured images of
the transmitted light through the diffuser at two wavelengths—630 nm and 940 nm (see
Figure 4a). Subsequently, we calculated the intensity difference by subtracting the 940 nm
profile from the 630 nm profile and plotted it over space and time (see Figure 4b). We also
calculated the mean variance of the difference over time (see Figure 4c). The integrating
sphere ensured a uniform illumination profile over the dual wavelengths in both space and
time, resulting in improved image quality. Thus, this uniform and consistent illumination
system enables the light to penetrate through the skin and provide the spatial information
of the light absorbed by the blood, resulting in spatial distribution measurements of blood
oxygen saturation.
 Sensors2023,
Sensors 2023,23,
23,3723
x FOR PEER REVIEW 10
10 of 18
18

Figure 3.3. Characterization

Figure Characterization ofof illumination
illumination profile:
profile: (a)
(a) The
The physical
physical dimensions
dimensions of of the
the integrating
integrating
sphere (the
sphere (the integrating
integrating sphere
sphere has
has aa diameter
diameter of of 50
50 mm
mm with
with 44 mm
mm holes
holes for
for 16
16 LEDs
LEDs and
and aa 12
12 mm
mm
light output port) and demonstration of light-scattering patterns in the integrating
light output port) and demonstration of light-scattering patterns in the integrating sphere; (b) Thesphere; (b) The
figure displays the illumination intensity of 630 nm (left) and 940 nm (right) that comes out from
figure displays the illumination intensity of 630 nm (left) and 940 nm (right) that comes out from the
the output port with the standard error of 1.2% for 630 nm and 1.3% for 940 nm at 1.8 cm field of
output port with the standard error of 1.2% for 630 nm and 1.3% for 940 nm at 1.8 cm field of view
view (FOV); (c) The 3D illumination profile of the 630 nm and 940 nm at 1.8 cm (FOV) were gener-
(FOV); (c) The
ated using 3D illumination
MATLAB R2022b; (d) profile
Frames of the 630nm
of 630 nmand
and940
940nmnmwavelengths
at 1.8 cm (FOV)werewere generated
captured alter-
using MATLAB R2022b; (d) Frames of 630 nm and 940 nm wavelengths were captured
nately, producing homogenous spatial patterns over time. The LED ‘ON’ and ‘OFF’ timing is syn- alternately,
producing
chronized homogenous
with camera spatial patterns
acquisition, withover time.
a total The of
delay LED ‘ON’
6 ms andbefore
(3 ms ‘OFF’ timing
cameraisexposure
synchronized
starts
and 3camera
with ms after the camera
acquisition, frame
with is captured).
a total delay of 6 ms (3 ms before camera exposure starts and 3 ms after
the camera frame is captured).
In addition, the acquired homogeneity ensures that the light source is consistent with
the input trigger signals. This consistency is also critical for accurate measurements of
SpO2 in pulse oximetry, as any discrepancies between the light source and the input trig-
ger signals can lead to inaccuracies in the measurements. Figure 3d presents an assess-
ment by utilizing a photodetector and a camera to demonstrate the compatibility between
the light source and the input trigger signals. The timing diagram from the photodetector
confirmed that fluctuations in the timing or intensity of the light source relative to the
input trigger signals were trivial. Thus, the light source was highly consistent with the
 profile from the 630 nm profile and plotted it over space and time (see Figure 4b). We also
calculated the mean variance of the difference over time (see Figure 4c). The integrating
sphere ensured a uniform illumination profile over the dual wavelengths in both space
and time, resulting in improved image quality. Thus, this uniform and consistent illumi-
nation system enables the light to penetrate through the skin and provide the spatial in-
Sensors 2023, 23, 3723 11 of 18
formation of the light absorbed by the blood, resulting in spatial distribution measure-
ments of blood oxygen saturation.

Figure
Figure 4.4. Quantification
Quantificationof of
uniformity
uniformity and similarity
and of red
similarity andand
of red near-infrared illuminations:
near-infrared illuminations:(a)
Transmitted
(a) Transmittedlightlight
illumination intensity
illumination at 630atnm
intensity 630(left), 940 nm
nm (left), (center),
940 and theand
nm (center), intensity differ-
the intensity
ence between 630 nm and 940 nm (right) emitted from the output port of the integrating sphere. To
difference between 630 nm and 940 nm (right) emitted from the output port of the integrating sphere.
capture the images of the transmitted light illumination intensity, a diffuser was placed over the
To capture the images of the transmitted light illumination intensity, a diffuser was placed over the
output port of the sphere with a field of view (FOV) of 1.8 cm on the x-axis and 1.4 cm on the y-axis;
output
(b) port of thedisplays
The spectrum sphere with a field in
the change of intensity
view (FOV) of 1.8 cm
difference between x-axis
on the 630 nmand
and1.4
940cmnm onatthe y-axis;
a central
(b) The spectrum displays the change in intensity difference between 630 nm and 940
field of view of 1.2 cm over a period of time ranging from 72 ms to 720 ms; (c) The plot depicts thenm at a central
field of view ofin
mean-variance 1.2 cm overdifference
intensity a period of time ranging
between 630 nmfrom
and 72940ms
nmtoover
720 ms;
time.(c) The plot depicts the
mean-variance in intensity difference between 630 nm and 940 nm over time.
3.2. Photoplethysmography
3.2. Photoplethysmography
An experimental test was conducted by imaging a fingertip (refer to supplementary
An experimental test was conducted by imaging a fingertip (refer to supplementary
Video S1). Figure 5a illustrates the relative position of the imaged illuminated area of the
Video S1). Figure 5a illustrates the relative position of the imaged illuminated area of
the fingertip. The mean values of the pixels contained in the highlighted regions of
the fingertip are plotted against time. The waveforms in the projected PPG graph are
inverted, and the received pixel light intensity is replaced with light absorption on the
vertical axis. Light absorption is directly proportional to the peripheral arterial pressure
waveform, whereas the received light intensity is inversely proportional. In the analyzed
PPG signal, the systolic peaks and diastolic troughs are easily distinguishable. Moreover,
the dicrotic notch—an inflection in the waveform caused by the abrupt closure of the
aortic valve—is distinctly visible. The outlined observations are consistent with expected
physiological behavior and further support the validity of the PPG signal as a measure of
cardiovascular activity.
 Light absorption is directly proportional to the peripheral arterial pressure waveform,
whereas the received light intensity is inversely proportional. In the analyzed PPG signal,
the systolic peaks and diastolic troughs are easily distinguishable. Moreover, the dicrotic
notch—an inflection in the waveform caused by the abrupt closure of the aortic valve—is
distinctly visible. The outlined observations are consistent with expected physiological
Sensors 2023, 23, 3723 12 of 18
behavior and further support the validity of the PPG signal as a measure of cardiovascular
activity.

Figure 5.
Figure 5. Measurement
Measurementofofphotoplethysmogram
photoplethysmogram using pulse
using oximetry
pulse imaging
oximetry system:
imaging (a) The
system: (a)pho-
The
toplethysmogram (PPG) signals acquired at various spatial locations are displayed, demonstrating
photoplethysmogram (PPG) signals acquired at various spatial locations are displayed, demonstrating
time-varying spatial features at different wavelengths; (b) the signal-to-noise ratio (SNR) plot for
time-varying spatial features at different wavelengths; (b) the signal-to-noise ratio (SNR) plot for
both the red and near-infrared wavelengths; (c) Fast Fourier Transform (FFT) spectra of the RED
both the red
(630 nm) andand near-infrared
near-infrared wavelengths;
(NIR) (940 nm) PPG(c) Fast Fourier
signals Transform
are shown, (FFT) spectra
highlighting of therate
the heart REDat
(630 nm) and near-infrared (NIR) (940 nm) PPG signals are
1.19 Hz (71.2 bpm) and respiratory rate at 0.25 Hz (15 BR/min).shown, highlighting the heart rate at
1.19 Hz (71.2 bpm) and respiratory rate at 0.25 Hz (15 BR/min).
The signal-to-noise ratio (SNR) plot suggests that the PPG signals obtained from dif-
The signal-to-noise ratio (SNR) plot suggests that the PPG signals obtained from
ferent spatial locations are consistent and have a relatively constant SNR (see Figure 5b).
different spatial locations are consistent and have a relatively constant SNR (see
This consistency across spatial locations, distinct pulse amplitude, and phase information
Figure 5b). This consistency across spatial locations, distinct pulse amplitude, and
demonstrates the efficiency of the pulse oximetry imaging system and further confirms
phase information demonstrates the efficiency of the pulse oximetry imaging system
the uniformity of the illumination used. A high-quality PPG signal which reflects the
and further confirms the uniformity of the illumination used. A high-quality PPG
amount of transmitted or reflected light that penetrates the skin is crucial for obtaining
signal which reflects the amount of transmitted or reflected light that penetrates the
accurate physiological parameters such as heart rate, respiratory rate, and blood oxygen
skin is crucial for obtaining accurate physiological parameters such as heart rate, respi-
saturation (SpO2). Fourier spectra of the red (630 nm) and NIR (940 nm) PPG signals are
ratory rate, and blood oxygen saturation (SpO2 ). Fourier spectra of the red (630 nm)
depicted in Figure 5c. The pulsatile component, evident at 1.19 Hz (71.4 bpm), is profound
and NIR (940 nm) PPG signals are depicted in Figure 5c. The pulsatile component,
in both spectra. The respiratory component, demonstrated at 0.25 Hz (15 BR/min), is also
evident at 1.19 Hz (71.4 bpm), is profound in both spectra. The respiratory compo-
apparent
nent, in both signals.
demonstrated The acquired
at 0.25 results areis
Hz (15 BR/min), comparable to thein
also apparent results
both obtained
signals. from
The
acquired results are comparable to the results obtained from a Patient monitor (model
M40, MEDIANA Co., Ltd., Seoul, South Korea), which recorded 1.23 Hz (73 bpm) and
0.26 Hz (16 BR/min) for the pulsatile and respiratory components, respectively.
Table 1 provides a comparative analysis between the proposed system and other pulse
oximetry imaging systems. Previously demonstrated pulse oximetry imaging systems
have used LED arrays [39,43] or LED rings [16,44] that relied on two different illumination
profiles for dual wavelengths. When the spatial profiles of the two illuminations are
different, the ratio between transmitted light will also differ. Such a mismatch between
the illumination profiles can introduce fundamental errors when calculating the spatial
distribution of blood oxygen saturation. In contrast, our illumination system created two
Sensors 2023, 23, 3723 13 of 18

identical patterns of red and near-infrared illumination, of which patterns were spatially
uniform with standard error of only 1.2% and 1.3%, respectively (see Figure 3b). We utilized
this illumination system to capture spatial and temporal data, from which we obtained
spectral information at different locations within the peripheral region.

Table 1. Comparison of our illumination system with other reported illumination systems.

Imaging System Type of Illumination Signal-to-Noise Ratio

Dual wavelength (630 nm & 940 nm)
Our proposed system Integrating sphere uniform light source SNR ratio range: 6–10
(SE—1.2% & 1.3%)
Dual wavelength
Non-contact Imaging
(520 nm & 660 nm) N/A
Plethysmography [44]
LED ring light source
Non-contact monitoring of blood oxygen Dual wavelength
saturation using camera and dual (610 nm & 880 nm) N/A
wavelength imaging system [39] LED array light source
A CMOS camera-based system for Dual wavelength
clinical photoplethysmography (520 nm & 660 nm) N/A
applications [16] LED ring light source
Dual wavelength
Non-contact Imaging
(650 nm & 880 nm) N/A
Photoplethysmography [43]
LED array light source

3.3. Spatio-Temporal Information of Blood Oxygen Saturation

The spatial distribution of blood flow as a function of time (i.e., the spatio-temporal
data) and the derived PPG waveforms are shown in Figure 6a for red illumination and
in Figure 6b for NIR illumination. The spatial and temporal characteristics of blood
flow in the fingertip were analyzed by utilizing the images captured under the red
(630 nm) illumination. The processed images, captured over the period of time during
the experiment, demonstrated a clear connection between the pulsatile blood flow in the
fingertip and the cardiac cycle (see Figure 6a). To further analyze the relationship between
the cardiac cycle and pulsatile blood flow in the fingertip, the captured time-series images
of the fingertip during near-infrared (940 nm) illumination were processed after removing
the DC component. Near-infrared imaging offers increased tissue penetration depth due to
reduced photon absorbance and scattering. As a result, the use of near-infrared illumination
is advantageous in pulse oximetry imaging systems [45]. The resulting images, shown in
Figure 6b, highlight the spatial variations in the pulsatile blood flow.
Sensors 2023, 23, 3723 14 of 18
Sensors 2023, 23, x FOR PEER REVIEW 14 of 18

Figure 6. Spatio-temporal assessment of blood oxygen saturation using pulse oximetry imaging
system: (a) The image illustrates the validated area of the finger illuminated by red (630 nm). The
raw Photoplethysmogram (PPG) waveform is shown for three cardiac cycles. The figure also shows
the processed 12 images for a single cardiac cycle showing spatial variation in pulsatile blood flow;
Sensors 2023, 23, 3723 15 of 18

(b) The image illustrates the validated area of the finger illuminated by red (940 nm). The raw
Photoplethysmogram (PPG) waveform is shown for three cardiac cycles. The figure also shows the
processed 12 images for a single cardiac cycle showing spatial variation in pulsatile blood flow (for
a video demonstration of three cycles, refer to Supplementary Materials Videos S2 and S3); (c) The
SpO2 map represents the spatial distribution of blood oxygen saturation within the validated region
of measurement. The average SpO2 measured for the subject was 97.17%.

SpO2 Quantification
To evaluate the performance of our illumination system, we calculated the spatial
distribution of blood oxygen saturation (SpO2 ) in the fingertip using the fundamental theory
of pulse oximetry. In pulse oximetry, the AC component corresponds to the maximum
peak of the PPG signal. The time-varying spatial maps of pulsatile blood flow obtained
using both red (630 nm) and near-infrared (940 nm) images correspond to the amplitude
variations of a PPG signal (see Figure 6a,b). The processed images under red illumination
show the maximum peak (AC630 ) of the PPG signal, which corresponds to high blood
volume at t = 684 ms. Similarly, under near-infrared illumination, the AC940 corresponds to
high blood volume at t = 720 ms. We normalized the AC components with DC components
at these temporal points. Last, we calculated SpO2 distribution using Equation (4), and by
utilizing calibrated m and c values of −21.56 and 110.66, respectively. The average SpO2
level measured by our system was 97.15% (see Figure 6c). For comparison, the SpO2 level
of the subject was also measured using the Patient monitor (model M40, MEDIANA Co.,
Ltd., Wonju, Republic of Korea), and a value of 99% was recorded. To minimize potential
calibration errors, it is recommended that the system be tested on several volunteers. The
developed pulse oximetry imaging system successfully measured the pulsatile variation
in blood and SpO2 levels of the peripheral region. Table 2 compares the SpO2 , heart rate,
and respiratory rate data from our proposed system and a commercially available Patient
monitor device (model M40, MEDIANA Co., Ltd., Wonju, Republic of Korea).

Table 2. Comparison of physiological parameters from our system with a commercially available
patient monitoring system.

Our Proposed System Commercial Patient Monitor

SpO2 (%) 97.15 99
Heart rate (bpm) 71.4 73
Respiratory rate (BR/min) 15 16

Table 3 compares the measurement systems for blood oxygen saturation based on
the parameters analyzed. Single channel pulse oximeters only monitor temporal changes
in transmitted light at a single point to calculate the blood oxygen saturation. If blood
oxygen saturation is measured at a single location, the measurements are influenced by
the shape of illumination and spatial variations in blood flow [2]. However, our proposed
imaging system can capture the spatial and temporal changes in transmitted light at
various spatial locations over the red (I630 nm see Figure 6a) and near-infrared (I940 nm see
Figure 6b). Additionally, the proposed system can provide the spatial distribution and
temporal change of SpO2 , as demonstrated by Figure 6c. On the other hand, the single-
channel pulse oximeter provides only the temporal changes of I630 nm and I940 nm , and the
SpO2 at a single location. The table highlights the superior performance of our proposed
system over single-channel pulse oximeters in terms of spatial and spectral data acquisition.
Sensors 2023, 23, 3723 16 of 18

Table 3. Comparison of our system’s specifications with a single-channel pulse oximetry system.

System Parameter Our Proposed System Single Channel Pulse Oximetry System
Spatial resolution 106 × 106 µm N/A
Provide spatial distribution and temporal
change of I630 nm ;
Spectral data at Provide temporal change of I630 nm at a
Series of I630 nm images visualizes
630 nm, I630 nm single location
pulsatile blood flow on region of interest
(Figure 6a).
Provide spatial distribution and temporal
change of I940 nm ;
Spectral data at Provide temporal change of I940 nm at a
Series of I940 nm images visualizes
940 nm, I940 nm single location
pulsatile blood flow on region of interest
(Figure 6b).
Provide spatial distribution and temporal Provide temporal change of SpO2 at a
Blood oxygen saturation level, SpO2
change of SpO2 (Figure 6c) single location

4. Conclusions
In this study, we developed a camera-based pulse oximetry imaging system that
employs a spatially uniform dual wavelength light source to observe the spatial-temporal
changes in blood oxygen saturation over the peripheral regions, such as the fingertip or
earlobe. We used a pulse oximetry technique in combination with a hyper-spectral imaging
approach to characterize different locations within the peripheral region. Our results
showed that both 630 nm and 940 nm images contain physiological information about the
blood flow within the peripheral region of the tissue under investigation. We combined
the processed red and near-infrared images to estimate the blood oxygen saturation at
individual locations within the peripheral region. The measured physiological parameters
were comparable to those of a commercial Patient monitor.
Our pulse oximetry imaging system offers several advantages over conventional single-
channel pulse oximetry methods. Our imaging technique eliminates limitations related to
inconsistent illumination profiles between the two wavelengths in the peripheral region
and provides rich spatio-temporal information about blood oxygen saturation at different
locations. The development of an integrated device utilizing this imaging technique holds
significant potential for improving biomedical applications, such as wound monitoring and
real-time, long-term monitoring of blood oxygen saturation, where accurate tissue oxygen
content information is critical.

Supplementary Materials: The following supporting information can be downloaded at: https:
//www.mdpi.com/article/10.3390/s23073723/s1, Video S1: Real time recorded video by conducting
the experiment, Video S2: The visualization of pulsatile blood flow by using 940 nm, Video S3: The
visualization of pulsatile blood flow by using 630 nm.
Author Contributions: Conceptualization, S.H.C., R.M., A.A. and D.K.J.; Methodology, S.H.C., R.M.,
K.T.H., E.E.N.-O. and K.M.B.; Conducting experiments, R.M., K.T.H., E.E.N.-O. and A.A.; Writing-
original draft preparation, S.H.C., R.M., K.T.H., E.E.N.-O. and H.-W.J.; Writing—review & editing,
S.H.C., R.M., K.T.H., E.E.N.-O. and D.K.J.; Formal analysis, S.H.C., R.M., D.K.J., A.A., H.-W.J. and
K.M.B.; Project administration, S.H.C., A.A., D.K.J., H.-W.J. and K.M.B.; Funding acquisition, S.H.C.
and K.M.B. All authors have read and agreed to the published version of the manuscript.
Funding: This work was supported by the National Research Foundation of Korea (NRF),
grant-funded by the Korean government (MSIT) (No. 2022R1A2C1010151; 2022R1C1C1011328;
2022H1D3A2A02081592) and the Brain Korea 21 Four Program. This research was also supported
by “Regional Innovation Strategy (RIS)” through the National Research Foundation of Korea
(NRF) funded by the Ministry of Education (MOE) in 2023 (2022RIS-005).
Institutional Review Board Statement: Ethical review and approval were waived for this study, due
to non-invasive measurement.
Sensors 2023, 23, 3723 17 of 18

Informed Consent Statement: Informed consent was obtained from all subjects involved in
the study.
Data Availability Statement: The data presented in this study are available from the corresponding
authors upon request.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Manta, C.; Jain, S.S.; Coravos, A.; Mendelsohn, D.; Izmailova, E.S. An evaluation of biometric monitoring technologies for vital
signs in the era of COVID-19. Clin. Transl. Sci. 2020, 13, 1034–1044. [CrossRef] [PubMed]
2. Tamura, T. Current progress of photoplethysmography and SPO2 for health monitoring. Biomed. Eng. Lett. 2019, 9, 21–36.
[CrossRef] [PubMed]
3. Strapazzon, G.; Gatterer, H.; Falla, M.; Cappello, T.D.; Malacrida, S.; Turner, R.; Schenk, K.; Paal, P.; Falk, M.; Schweizer, J.; et al. Hypoxia
and hypercapnia effects on cerebral oxygen saturation in avalanche burial: A pilot human experimental study. Resuscitation 2021,
158, 175–182. [CrossRef] [PubMed]
4. Mas-Bargues, C.; Sanz-Ros, J.; Román-Domínguez, A.; Inglés, M.; Gimeno-Mallench, L.; El Alami, M.; Viña-Almunia, J.; Gambini,
J.; Viña, J.; Borrás, C. Relevance of oxygen concentration in stem cell culture for regenerative medicine. Int. J. Mol. Sci. 2019,
20, 1195. [CrossRef]
5. Yin, Y.; Shu, S.; Qin, L.; Shan, Y.; Gao, J.-H.; Lu, J. Effects of mild hypoxia on oxygen extraction fraction responses to brain
stimulation. J. Cereb. Blood Flow Metab. 2021, 41, 2216–2228. [CrossRef]
6. Michard, F. Hemodynamic monitoring: Would a pulse oximeter do the job? Crit. Care Med. 2021, 49, 383–386. [CrossRef]
7. Jubran, A. Pulse oximetry. Intensive Care Med. 2004, 30, 2017–2020. [CrossRef]
8. Ferrando, C.; Tusman, G.; Suarez-Sipmann, F.; León, I.; Pozo, N.; Carbonell, J.; Puig, J.; Pastor, E.; Gracia, E.;
Gutiérrez, A.; et al. Individualized lung recruitment maneuver guided by pulse-oximetry in anesthetized patients un-
dergoing laparoscopy: A feasibility study. Acta Anaesthesiol. Scand. 2018, 62, 608–619. [CrossRef]
9. Amouzeshi, A.; Salehi, F.; Ehsani, H.; Jomefourjan, S.; Jani, M.; Amouzeshi, Z. Comparison of pulse oximetry and clinical
examination for proper tissue perfusion in congenital heart patients. J. Surg. Trauma 2021, 9, 26–31.
10. Matsushita, K.; Aoki, K.; Kakuta, N.; Yamada, Y. Fundamental study of reflection pulse oximetry. Opt. Rev. 2003, 10, 482–487.
[CrossRef]
11. Grap, M.J. Pulse oximetry. Crit. Care Nurse 2002, 22, 69–74. [CrossRef]
12. Pothisarn, W.; Chewpraditkul, W.; Yupapin, P. Noninvasive hemoglobin-measurement-based pulse oximetry. In Optics in Health
Care and Biomedical Optics: Diagnostics and Treatment; SPIE: Bellingham, WA, USA, 2002.
13. Kumar, M.; Suliburk, J.W.; Veeraraghavan, A.; Sabharwal, A. PulseCam: A camera-based, motion-robust and highly sensitive
blood perfusion imaging modality. Sci. Rep. 2020, 10, 4825. [CrossRef]
14. Cabanas, A.M.; Fuentes-Guajardo, M.; Latorre, K.; León, D.; Martín-Escudero, P. Skin pigmentation influence on pulse oximetry
accuracy: A systematic review and bibliometric analysis. Sensors 2022, 22, 3402. [CrossRef]
15. Keller, M.D.; Harrison-Smith, B.; Patil, C.; Arefin, M.S. Skin Colour Affects the Accuracy of Medical Oxygen Sensors; Nature Publishing
Group: UK, London, 2022.
16. Humphreys, K.; Markham, C.; Ward, T.E. A CMOS camera-based system for clinical photoplethysmographic applications. In
Opto-Ireland 2005: Imaging and Vision; SPIE: Bellingham, WA, USA, 2005.
17. Al-Naji, A.; Khalid, G.; Mahdi, J.; Chahl, J. Non-contact SpO2 prediction system based on a digital camera. Appl. Sci. 2021,
11, 4255. [CrossRef]
18. Moco, A.V.; Stuijk, S.; De Haan, G. Ballistocardiographic artifacts in PPG imaging. IEEE Trans. Biomed. Eng. 2015, 63, 1804–1811.
[CrossRef]
19. Moço, A.V.; Stuijk, S.; de Haan, G. Motion robust PPG-imaging through color channel mapping. Biomed. Opt. Express 2016, 7,
1737–1754. [CrossRef]
20. Mannheimer, P.D. The light–tissue interaction of pulse oximetry. Anesth. Analg. 2007, 105, S10–S17. [CrossRef]
21. Torp, K.D.; Modi, P.; Simon, L.V. Pulse oximetry. In StatPearls; StatPearls Publishing: St. Petersburg, FL, USA, 2022.
22. Labsphere. Technical Guide: Integrating Sphere Theory and Applications; Labsphere North Sutton: Sutton, NH, USA, 2013.
23. Edwards, D.K.; Gier, J.T.; Nelson, K.E.; Roddick, R.D. Integrating sphere for imperfectly diffuse samples. JOSA 1961, 51, 1279–1288.
[CrossRef]
24. Geladi, P.; Burger, J.; Lestander, T. Hyperspectral imaging: Calibration problems and solutions. Chemom. Intell. Lab. Syst. 2004, 72,
209–217. [CrossRef]
25. Sun, Y.; Thakor, N. Photoplethysmography revisited: From contact to noncontact, from point to imaging. IEEE Trans. Biomed. Eng.
2015, 63, 463–477. [CrossRef]
26. Taylor-Williams, M.; Spicer, G.; Bale, G.; Bohndiek, S.E. Noninvasive hemoglobin sensing and imaging: Optical tools for disease
diagnosis. J. Biomed. Opt. 2022, 27, 080901. [CrossRef] [PubMed]
27. Torrent, J.; Barrón, V. Diffuse reflectance spectroscopy. In Methods of Soil Analysis Part 5—Mineralogical Methods; SSSA: Madison,
WI, USA, 2008; Volume 5, pp. 367–385.
Sensors 2023, 23, 3723 18 of 18

28. Vaqar, A.; Haque IR, I.; Zaidi, T. Spectroscopic Properties of Blood for Pulse Oximeter Design. In Proceedings of the 2019 9th
International Conference on Biomedical Engineering and Technology, Tokyo, Japan, 28–30 March 2019.
29. Caramizoiu, S.; Mihalache, I. Highly reflective surface coating using BaSO4 . In Proceedings of the 2022 International Semiconduc-
tor Conference (CAS), Poiana Brasov, Romania, 12–14 October 2022; IEEE: Piscataway, NJ, USA, 2022.
30. Dias, L.D.S.; Junior, J.C.D.S.; Felicio, A.L.D.S.M.; de Franca, J.A. A NIR photometer prototype with integrating sphere for the
detection of added water in raw milk. IEEE Trans. Instrum. Meas. 2018, 67, 2812–2819. [CrossRef]
31. Nitzan, M.; Romem, A.; Koppel, R. Pulse oximetry: Fundamentals and technology update. Med. Devices Evid. Res. 2014, 7,
231–239. [CrossRef] [PubMed]
32. Smuda, K.; Gienger, J.; Hönicke, P.; Neukammer, J. Function of hemoglobin-based oxygen carriers: Determination of methe-
moglobin content by spectral extinction measurements. Int. J. Mol. Sci. 2021, 22, 1753. [CrossRef] [PubMed]
33. Casalino, G.; Castellano, G.; Zaza, G. A mHealth solution for contact-less self-monitoring of blood oxygen saturation. In
Proceedings of the 2020 IEEE Symposium on Computers and Communications (ISCC), Rennes, France, 7–10 July 2020; IEEE:
Piscataway, NJ, USA, 2020.
34. Rosa, A.d.F.G.; Betini, R.C. Noncontact SpO2 measurement using Eulerian video magnification. IEEE Trans. Instrum. Meas. 2019,
69, 2120–2130. [CrossRef]
35. Kocsis, L.; Herman, P.; Eke, A. The modified Beer–Lambert law revisited. Phys. Med. Biol. 2006, 51, N91. [CrossRef]
36. Selvaraju, V.; Spicher, N.; Wang, J.; Ganapathy, N.; Warnecke, J.M.; Leonhardt, S.; Swaminathan, R.; Deserno, T.M. Continuous
monitoring of vital signs using cameras: A systematic review. Sensors 2022, 22, 4097. [CrossRef]
37. Labati, R.D.; Piuri, V.; Rundo, F.; Scotti, F.; Spampinato, C. Biometric recognition of PPG cardiac signals using transformed
spectrogram images. In Pattern Recognition, Proceedings of the ICPR International Workshops and Challenges, Virtual Event, 10–15
January 2021; Springer: Berlin/Heidelberg, Germany, 2021.
38. Tian, X.; Wong, C.-W.; Ranadive, S.M.; Wu, M. A Multi-Channel Ratio-of-Ratios Method for Noncontact Hand Video Based SpO $
_2 $ Monitoring Using Smartphone Cameras. IEEE J. Sel. Top. Signal Process. 2022, 16, 197–207. [CrossRef]
39. Shao, D.; Liu, C.; Tsow, F.; Yang, Y.; Du, Z.; Iriya, R.; Yu, H.; Tao, N. Noncontact monitoring of blood oxygen saturation using
camera and dual-wavelength imaging system. IEEE Trans. Biomed. Eng. 2015, 63, 1091–1098. [CrossRef]
40. Wu, J.; Gunther, J.; Jayet, B.; Ray, N.; Andersson-Engels, S.; Kainerstorfer, J.M. Self-calibrated pulse oximetry based on absorption
changes in photon pathlength. In Proceedings of the European Conference on Biomedical Optics, Munich, Germany, 20–24 June
2021; Optica Publishing Group: Washington, DC, USA, 2021.
41. Yuan, S.; Devor, A.; Boas, D.A.; Dunn, A.K. Determination of optimal exposure time for imaging of blood flow changes with laser
speckle contrast imaging. Appl. Opt. 2005, 44, 1823–1830. [CrossRef]
42. Fercher, A.; Briers, J. Flow visualization by means of single-exposure speckle photography. Opt. Commun. 1981, 37, 326–330.
[CrossRef]
43. Sun, Y.; Hu, S.; Azorin-Peris, V.; Kalawsky, R.; Greenwald, S. Noncontact imaging photoplethysmography to effectively access
pulse rate variability. J. Biomed. Opt. 2013, 18, 061205. [CrossRef]
44. Fan, Q.; Li, K. Noncontact imaging plethysmography for accurate estimation of physiological parameters. J. Med. Biol. Eng. 2017,
37, 675–685. [CrossRef]
45. Okubo, K. NIR hyperspectral imaging. In Transparency in Biology: Making the Invisible Visible; Springer: Berlin/Heidelberg,
Germany, 2021; pp. 203–222.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.