FULL PAPER

www.advhealthmat.de

3D Printing Silicone Elastomer for Patient-Specific Wearable

Pulse Oximeter
Sara Abdollahi, Eric J. Markvicka, Carmel Majidi, and Adam W. Feinberg*

saturation (SpO2 ) to determine physiologic

Commercial pulse oximeters are used clinically to measure heart rate and status for diagnosis and decision mak-
blood oxygen saturation and traditionally made from rigid materials. However, ing. In critical care, this includes contin-
these devices are unsuitable for continuous monitoring due to poor fit and uous monitoring of neonates, postopera-
mechanical mismatch. Soft materials that match the elastic properties of tive patients, and those with cardiopul-
monary disease.[1] However, current devices
biological tissue provide improved comfort and signal-to-noise but typically
are rigid, bulky and burdensome to wear for
require molding to manufacture, limiting the speed and ease of customizing long-term monitoring, which has limited
for patient-specific anatomy. Here, freeform reversible embedding (FRE) 3D their use outside the hospital for preven-
printing is used to create polydimethylsiloxane (PDMS) elastomer cuffs for tive care, health tracking, and the manage-
use on the hand and foot. FRE enables printing liquid PDMS prepolymer in ment of chronic disease.[2] This is especially
a concern for diseases such as diabetes with
3D geometries within a sacrificial hydrogel bath that provides support during
a risk of neuropathy and peripheral arterial
cure. This serves as proof-of-concept for fabricating patient-specific pulse disease, which can lead to loss of blood flow
oximeters with pressure sensing, termed P3 -wearable. A sizing analysis in the limbs and ultimately amputation.[3]
establishes dimensional accuracy of FRE-printed PDMS compared to The emergence of consumer-grade wrist-
anatomical computer-aided design models. The P3 -wearable successfully band activity trackers (e.g., Apple Watch,
outputs photoplethysmography (PPG) and pressure amplitude signals FitBit) has also demonstrated the value
of continuous monitoring of heart rate
wirelessly to a tablet in real time and the PPG is used to calculate heart rate,
and SpO2 , which has driven the advance
blood oxygen content, and activity state. The results establish that FRE of wearable technology including reduced
printing of PDMS can be used to fabricate patient-specific wearable devices battery size,[4] wireless operation,[5] con-
and measure heart rate and blood oxygenation on par with commercial tinuous measurement,[6] and elimination
devices. of motion-induced artifacts.[7] This one-
size-fits-all approach, however, has proved
challenging to implement for long-term
1. Introduction medical monitoring, where there is need to fit patient-specific
anatomy and record high-quality and repeatable measurements.
Traditional pulse oximeters are used in the clinic to measure Further, current pulse oximeter devices on the market are lim-
vital signs such as heart rate and peripheral capillary oxygen ited to a few anatomic locations, such as the clip-on pulse oxime-
ter for the finger, which cannot provide information on local-
Dr. S. Abdollahi, Prof. A. W. Feinberg
ized perfusion in other areas of the body. The need to fabricate
Department of Biomedical Engineering custom-fit medical wearables has spurred interest in advanced
Carnegie Mellon University manufacturing methods that enable iterative design and low-cost
5000 Forbes Avenue, Pittsburgh, PA 15213, USA customization.[8]
E-mail: [email protected] 3D printing is uniquely positioned to fabricate a wearable de-
Dr. E. J. Markvicka, Prof. C. Majidi vice with patient-specific geometry in a single step based on
Robotics Institute
Carnegie Mellon University a computer-aided design (CAD) model from anatomical scans.
5000 Forbes Avenue, Pittsburgh, PA 15213, USA However, standard polymeric materials such as polylactic acid
Prof. C. Majidi (PLA), acrylonitrile butadiene styrene (ABS), and thermoplastic
Department of Mechanical Engineering polyurethane (TPU) elastomer are up to four orders of magnitude
Carnegie Mellon University stiffer than soft tissue.[9] Using these leads to a rigid and bulky
5000 Forbes Avenue, Pittsburgh, PA 15213, USA
interface that is uncomfortable for long term wear and restricts
Prof. A. W. Feinberg
Department of Materials Science and Engineering
movement because of the mechanical mismatch with the skin.[10]
Carnegie Mellon University Pulse oximeters made of rigid materials can also cause pressure
5000 Forbes Avenue, Pittsburgh, PA 15213, USA ischemia and skin breakdown.[11] Softer materials with lower
elastic modulus and bending stiffness, and increased stretchabil-
The ORCID identification number(s) for the author(s) of this article ity that are comparable to human skin (E ≈ 1 MPa) are needed.[12]
can be found under https://doi.org/10.1002/adhm.201901735
Silicone elastomers, in particular polydimethylsiloxane (PDMS),
DOI: 10.1002/adhm.201901735

Adv. Healthcare Mater. 2020, 1901735 1901735 (1 of 9) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advhealthmat.de

can be tuned to have these properties and have been used to make using a silicone adhesive (Figure 1b). This was done purposefully
wearable sensors that have been tested on users’ skin.[12b,13] Keep- because we wanted to 1) have a CAD design that minimized the
ing the sensor in place is also a challenge to ensure consistent total number of fabrication steps and 2) avoid additional postpro-
measurements longitudinally through time and adhesives have cessing that would be required to cut out the window manually.
been used for this purpose.[14] Yet, chemical adhesives can irritate The CAD model was then scaled linearly to produce slightly
or breakdown the skin over continuous measurements.[12a,13a] different sizes, FRE printed in PDMS, and the dimensional accu-
This suggests that a well-designed PDMS wearable could apply racy of each cuff size was compared with the CAD model counter-
light pressure to maintain placement without needing an adhe- part (Figure 1c). The error between the CAD model and the actual
sive. However, to date there has not been a way to directly 3D print print dimensions was found to increase linearly with cuff size.
PDMS for this type of design where pulse oximetry is needed at The print height (z-axis) was larger than the CAD model while the
peripheral sites of the body such as the foot, which are suscepti- inner major and minor axes (xy plane) were smaller. This makes
ble to edema and skin breakdown.[15] sense because the g-code generated from the CAD model out-
In this study, we combined 3D printing of PDMS with flexi- lines a path for the center of the PDMS filaments to be printed,
ble electronics to develop a Patient-specific Pulse oximeter with and the 17-gauge needle has an inner diameter of 1.219 mm,
Pressure sensing that we call a P3 -wearable. The goal was to lever- which should result in increased height and decreased inner di-
age freeform reversible embedding (FRE) 3D printing to directly ameter. A similar analysis was done for the window cutout de-
fabricate a PDMS cuff to fit the finger or toe, and thus demon- signed to fit the fPCB, which had a smaller height and larger
strate proof-of-concept production of an anatomically accurate width than the CAD model (Figure 1d). There was no change
soft wearable. FRE is an embedded 3D printing technique de- as a function of overall cuff size because the window cutout di-
rived from freeform reversible embedding of suspended hydro- mensions were kept constant. The decrease in height was ex-
gels (FRESH) 3D bioprinting and enables printing of soft materi- pected due to the filament diameter and the g-code pathing, how-
als such as liquid PDMS prepolymers.[16] In FRE, liquid materials ever the larger width was likely due to the challenge of starting
are printed layer by layer within a sacrificial support bath that acts and stopping the flow of a relatively viscous print material like
as a Bingham plastic with a yield stress. As the print nozzle moves PDMS. Having quantified the difference in dimensions between
through the bath and exerts a stress above the bath’s yield stress, the CAD design and the actual printed PDMS cuff, linear correc-
the support briefly liquifies, enabling the deposition of the liquid tion factors were used to produce a final print of a desired size.
materials being printed. By optimizing the rheological properties The complete P3 -wearable consists of the soft PDMS cuff
of the bath and the ink being printed, FRE can be used to print integrated with the fPCB and battery and was first tested on the
low viscosity liquids in 3D that would otherwise flow and form index finger in various sizes to determine the impact of fit on
a puddle on the build platform. Using FRE printing of PDMS, performance (Figure 2a). We previously reported the elastic mod-
we developed a sizing analysis framework to assess the dimen- ulus of FRE 3D printed PDMS as 1.2 ± 0.1 MPa and showed that
sional accuracy of P3 -wearable cuffs for the hand and foot based the layers have excellent fusion and can withstand tensile strain
on patient measurements, and analyzed device performance. As up to 130% elongation to failure, similar to cast PDMS.[16b,17]
proof-of-concept, the P3 -wearable device was tested at rest, while The waveforms recorded by the pulse oximeter optoelectronics
sitting and during walking. The results demonstrate the ability to and pressure sensor were output wirelessly in real time to a
track activity, pulse, and oxygen saturation, showing the potential tablet computer (Figure 2b,c; Figure S2a,b, Supporting Infor-
to continuously monitor peripheral blood perfusion. mation). The P3 -wearable photodetector successfully recorded
optical signals from the LEDs (Figure 2b; Figure S2a, Supporting
Information) as well as pressure changes (Figure 2c; Figure S2b,
2. Results Supporting Information). The size of the PDMS finger cuff was
then linearly scaled by small percentage increments to determine
The P3 -wearable was first developed for the hand and the soft the impact of fit on the optical and pressure signals (Figure 2d,e).
PDMS cuffs were subject to a sizing analysis to determine di- These results for pressure due to the heartbeat and photoplethys-
mensional accuracy of the printing process. The purpose of the mography (PPG) (Figure 2d,e) show that the 100% cuff has
finger cuff was to validate that we could fabricate a complete and reduced signal amplitudes, which suggests this smaller size is
functional P3 -wearable and assess print fidelity and reproducibil- restricting blood flow (i.e., volume) into the finger. For the larger
ity. While this is a simple cylindrical shape that could be easily 105% and 110% cuffs, the reduced contact force with the skin
molded, it serves as a basic design for validation before proceed- (Figure 2f) also results in decreased signal amplitudes for pres-
ing to more complex patient-specific prints. Figure 1a depicts the sure due to the heartbeat and PPG (Figure 2d,e). The pressure
design of the P3 -wearable, starting from i) measurement of the amplitudes for the heartbeat for the 102% finger cuff (≈1.1 kPa)
finger, ii) FRE printing of the PDMS cuff, iii) integration of the were nearly fourfold higher than the 105% finger cuff (≈0.3 kPa)
flexible printed circuit board (fPCB) (Figure S1, Supporting Infor- size (Figure 2d) and were generally in sync with the PPG ampli-
mation) with the PDMS cuff, and iv) testing of the P3 -wearable tude with respect to trends across cuff sizes (Figure 2e). Next, the
on the finger. First, finger measurements made with a caliper contact pressure of the finger on the pressure sensor was evalu-
were used to design a patient-specific CAD model for 3D print- ated and found to have a trend whereby it decreased as the finger
ing, into which a window cutout was designed as a place to insert cuff size increased (Figure 2f). Importantly, a statistically sig-
the fPCB with the pulse oximeter and pressure sensor chip. FRE nificant decrease in contact pressure resulted when going from
printing enabled the fabrication of the PDMS cuff with a well- the 102% to 105% finger cuff sizes, which also corresponded
defined window cutout into which the fPCB was easily integrated to the drop in pressure amplitude for the heartbeat (Figure 2d).

Adv. Healthcare Mater. 2020, 1901735 1901735 (2 of 9) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advhealthmat.de

Figure 1. 3D printing of PDMS finger cuff for the P3 -wearable device. a) Schematic of the entire process that shows i) the use of a digital caliper to take
measurements on the patient’s finger, ii) the design of a CAD model for the finger cuff based on the measurements, iii) the 3D printing of the finger cuff
with PDMS using the FRE technique, iv) the integration of the fPCB into the PDMS finger cuff, and v) the P3 -wearable worn on the hand to record pulse
rate, SpO2 , and pressure measurements. The fPCB for the P3 -wearable includes the pulse oximeter chip (MAX30102, Maxim Integrated) with IR, and
red LEDs, a barometric chip for pressure sensing, a Bluetooth low energy wireless module, and a coin cell battery. b) Examples of the PDMS finger cuff
(top view and front window view) and the PDMS finger cuff integrated with the fPCB without the battery attachment. Scale bar is 1 cm. c) Comparison
of print dimensions of the 3D printed PDMS finger cuff as a function of the CAD model dimensions for different sizes (85%, 90%, 95%, 97.5%, 98%,
100%, 101%, 102%, 102.5%, 105%, 110%, and 115%) (n = 3 per size). The linear fits are y = 1.01x + 1.06 (R2 = 0.97) for the height, y = 0.94x − 0.48
(R2 = 0.99) for ID major axis, and y = 0.97x − 0.24 (R2 = 0.98) for ID minor axis. d) Dimensional analysis of the internal window cutout in the PDMS
finger cuff compared to the CAD dimensions (5.61 mm × 12.06 mm) (n = 3 per size).

This suggests that the finger cuff size of 102% gave the highest calipers (Figure 3a). At first, the CAD model was designed with
PPG and pressure amplitude for the heartbeat because there is a the window cutout for the pulse oximeter fPCB on the toenail for
minimum contact pressure for reading the heartbeat through the comfort and to avoid excessive pressures on the pressure sensor
pressure sensor, but too much contact pressure actually reduces (Figure 3b). The 3D model generated for the big toe was more
blood flow and thus degrades the signals. Still, the difference in complex than the simple cylinder used for the finger and con-
PPG amplitude between the 100% and 102% cuff are noteworthy tained anatomical detail such as the toenail (Figure 3c). More
(Figure 2e) even though the contact pressures are within the complex FRE prints with overhangs and large internal voids
measurement error (Figure 2f). This is because the contact that are challenging or impossible to 3D print in air or mold
pressure is the pressure applied to the sensor, and thus only an were also created in PDMS to show the versatility of the ap-
approximate indicator of the pressure that the cuff exerts around proach, including scanned models of the human face and ad-
the entire finger. Indeed, comparing PPG amplitude to contact ditional models of the toe (Figure S3, Supporting Information).
pressure for the 100% to 110% finger cuff sizes suggested that FRE printing produced a PDMS toe cuff with a cutout for mount-
contact pressures in the 2.5–15.5 kPa range were in the optimal ing the pulse oximeter and pressure sensor chip and was able
range (Figure 2g). Overall, these results enabled us to understand to recreate the shape of the CAD model (Figure 3d). To assess
the relationship between size and contact pressure on heartbeat dimensional accuracy of the printed toe cuff, we scanned it us-
amplitude and PPG amplitude, enabling us to identify a 102% ing the same 3D laser scanner used to originally scan the toe
finger cuff size as the best choice in this specific example. and generated a 3D model (Figure 3e). We performed quanti-
The approach used to develop the P3 -wearable for the foot tative gauging by comparing the two CAD models and creat-
was similar to the finger, except the anatomy of the toe was ing a heat map showing deviation of the toe cuff superimposed
captured using a 3D laser scanner instead of measured using on the toe (Figure 3f). There were clear differences on the nail

Adv. Healthcare Mater. 2020, 1901735 1901735 (3 of 9) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advhealthmat.de

Figure 2. Performance of the 3D printed wearable pulse oximeter for the finger. a) Representative image of the patient-specific wearable pulse oximeter
tested on the index finger. b) Raw signal output of the red PPG waveform from the pulse oximeter shows the ability to read from the finger and overall
system performance with data transmitted wirelessly to the tablet computer for collection. c) Raw signal output of the pressure sensor, which is the sum
of the noncontact pressure, contact pressure, and pressure from the pulsatile blood flow. d) Average amplitude of the pressure signals from the pulsatile
blood flow due to the heartbeat (i.e., oscillating waveform in panel (c)) as a function of cuff size. Statistically significant differences indicated by * (P
< 0.05) by one-way ANOVA followed by a Tukey–Kramer post hoc test (n = 3). e) Average amplitude of the PPG from the red and IR LEDs. Statistically
significant differences indicated by * (P < 0.05) by one-way ANOVA followed by a Tukey–Kramer post hoc test (n = 3). f) The contact pressure between
the finger and the pressure sensor as a function of cuff size. Statistically significant differences indicated by * (P < 0.05) by one-way ANOVA followed
by a Tukey–Kramer post hoc test (n = 3). g) Pulse oximeter PPG signal amplitude as a function of the different contact pressures applied by cuff sizes
100%, 102%, 105%, and 110% (n = 3).

region, surrounding the pulse oximeter window, and more evenly signal, and thus none of these sizes were further analyzed. Ad-
spread throughout the cuff. However, throughout the entire toe ditionally, when the sensor was placed beneath the toe (bottom),
cuff these differences were less than 2 mm, and the average devi- the PPG amplitude for the red LED was overall higher than on
ation was less than 1 mm, with multiple prints of the same size the nail (top) (Figure 3g). Yet the pressure amplitude for the 90%
showing similar dimensions Table S1 (Supporting Information). cuff size was larger with the sensor placed on the nail (top) ver-
The dimensional accuracy across different cuff sizes from 90% sus beneath the toe (bottom) (Figure 3i). Taken together, these
to 110% were overall quite similar, providing confidence that the results highlight the fact that cuff size and location impact the
toe cuff can be FRE printed using PDMS with predictable dimen- P3 -wearable’s functional performance and that all locations (e.g.,
sions (Figure S4, Supporting Information). The cuffs were then bottom, top, and side) have the potential to work. Based on the
integrated with the pulse oximeter fPCB to make the complete P3 - results though, we identified the 90% toe cuff size with the sen-
wearable for the foot (Figure 3b), which was tested in the same sor on the nail (top) as the best option for recording both the PPG
three step sequence (detached–worn–detached) as for the hand. and pressure signal from the heartbeat. There are also other vari-
Given the shape of the large toe, we decided to evaluate the P3 - ables such as the wavelength of light, the anatomical location,
wearable with the fPCB located in three different positions; on and the proximity of the arteries to the surface of the skin that
the top of the toenail, on the bottom of the toe and on the side of are a determinant of the output signal. Thus, customization may
the toe (Figure 3a). be necessary depending on the sensitivity needed for a given ap-
Results showed that there were clear differences based on the plication.
cuff size and the location of the sensor on the toe (Figure 3g– Finally, we evaluated the performance of the P3 -wearable in
i). However, a change in the size and location did not necessar- various usage scenarios with the 90% toe cuff size and the fPCB
ily result in a similar change to both the PPG and pressure sig- integrated on the nail in order to minimize direct pressure being
nals. For example, the 95% and 100% cuff sizes on the nail (top) placed on the fPCB electronics. The P3 -wearable was tested at rest
and beneath the toe (bottom) had overall higher PPG amplitudes. with the foot extended flat and without putting any weight on the
In contrast, a much smaller cuff size of 83% with the sensor on device (Figure 4a), and then compared to sitting (Figure 4b) and
the side of the toe also had both high PPG and pressure signals, walking (Figure 4c) (note that the battery and Bluetooth module
but other sizes with the sensor on the side (78%, 80%, 85%, and were on top of the foot). The signal recordings during the at rest
100%) either did not fit (i.e., too small to wear) or had very low and sitting conditions had the typical pulsatile waveforms for the

Adv. Healthcare Mater. 2020, 1901735 1901735 (4 of 9) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advhealthmat.de

Figure 3. Fabrication, assessment and performance of the 3D printed wearable pulse oximeter for the toe. a) Schematic of the fabrication process
showing i) 3D scanning of the patient’s toe, ii) generation of a CAD model of the toe cuff based on the scan, iii) 3D printing the toe cuff with PDMS using
the FRE technique, iv) integrating the fPCB with pulse oximeter chip into the PDMS toe cuff, and v) wearing the completed sensor on the toe to record
pulse and SpO2 , with the chip positioned on the top, bottom or side of toe. b) Complete wearable toe pulse oximeter with the pulse oximeter window
placement on the bottom of the toe. Scale bar = 1 cm. The fPCB includes the pulse oximeter chip (MAX30102, Maxim Integrated) with IR and red LEDs, a
barometric chip for pressure sensing, a Bluetooth low energy wireless module, and a coin cell battery. c) CAD model for 3D printing the toe cuff obtained
from the 3D anatomical scan. d) Toe cuff 3D printed from PDMS using the FRE technique. e) CAD model of the printed PDMS toe cuff obtained by 3D
scanning of the printed toe cuff in (d). f) Quantitative gauging analysis showing the deviation of the CAD model for the toe cuff compared to the CAD
model from the scan of the printed PDMS toe cuff with the color code indicating the differences between the two. g) Signal amplitude for the red LED
of the pulse oximeter as a function of cuff size and location. One-way ANOVA was used to determine statistical significance followed by a Tukey–Kramer
post hoc test indicated by * (P < 0.05, n = 3). h) Signal amplitude for the IR LED of the pulse oximeter as a function of cuff size and location. One-way
ANOVA was used to determine statistical significance followed by a Tukey–Kramer post hoc test indicated by * (P < 0.05, n = 3). i) Signal amplitude
for the pressure from the pulsatile blood flow as a function of cuff size and location. One-way ANOVA was used to determine statistical significance
followed by a Tukey–Kramer post hoc test indicated by * (P < 0.05, n = 3).

foot red and IR PPG, and pressure (Figure 4a,b). However, there While sitting, the P3 -wearable recorded SpO2 of 98% and heart
was a marked decrease in the signal amplitude sitting compared rate of 71 beats min−1 , while the commercial pulse oximeter
to at rest. As seen for the finger, the peaks of the cyclical pres- recorded SpO2 of 97% and 71 beats min−1 . While these values
sure recordings for both conditions (Figure 4a,b) were inversely could be quantified for the at rest and sitting conditions, walking
correlated to the PPG signal. The heart rate and SpO2 calculated resulted in large increases in the pressure and large decreases in
from the PPG signals recorded by the P3 -wearable matched val- the PPG signals (Figure 4c). During the period of contact of the
ues obtained with a commercial pulse oximeter worn at the same foot with the ground, the pressure signal increased by 5–10 kPa
time on the index finger. At rest, the P3 -wearable recorded SpO2 over baseline, obscuring any pulsatile signal from blood the flow,
of 100% and heart rate of 63 beats min−1 , while the commer- and corresponded to a decrease in the PPG signals. While the
cial pulse oximeter recorded SpO2 of 98% and 64 beats min−1 . PPG signals increased and the pressure decreased during the

Adv. Healthcare Mater. 2020, 1901735 1901735 (5 of 9) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advhealthmat.de

Figure 4. Activities testing with the 90% cuff wearable pulse oximetry on top of the toenail. The three states of pulse oximeter signal recording of the
Red PPG, IR PPG, and contact pressure during wear are a) at rest (no weight), b) sitting, and c) walking.

noncontact phase of the gait cycle, oscillations remained and nei- rection factors for the finger cuff, but these likely cannot be uni-
ther produced a usable signal. It is thought that this was due to versally applied to other sizes and shapes, in part because of dif-
movement of the fPCB relative to the toe, producing motion ar- ferences in the way initial anatomical measurements are made.
tifacts. While not usable during walking, the signals from the Still, the correction factors provide a starting point, and the print
P3 -wearable could easily be used to determine when the user is dimensions may be within a clinically acceptable range depend-
moving and restrict reading of usable PPG and pressure signals ing on the application. For example, Ashtiani et al. reported that
to periods of rest or sitting. Further studies are needed to assess 3D printed dental onlays, compared to onlays made with conven-
the comfort of the device for long term wear and will ultimately tional methods, were within the clinically acceptable marginal
be part of product development if commercialized. discrepancy (20–150 µm) for this application.[19] In another study
on an ear prosthesis, the clinically acceptable mean difference
between the printed prosthesis and the subject was reported as
3. Discussion 2 mm (1.5%).[20] Thus, our average difference between the prints
and digital cuff dimensions may be within the clinically accept-
The P3 -wearable we have developed has its basis in the abil- able range for a wearable pulse oximetry application.
ity to 3D print PDMS elastomer using the FRE process to cre- Overall, differences across the signal amplitudes of the PPG
ate a patient-specific soft cuff. However, FRE is a recently devel- and pressure sensors revealed that both the location of the sen-
oped approach and there has been limited data on dimensional sors and the size of the cuff affect the device’s performance (Fig-
accuracy.[16a-e] Like most 3D printing,[18] there are many print ures 2d,e and 3g–i). The PPG signal oscillations resulted from
parameters such as speed, flowrate, and layer height that have changes in blood volume during the systolic and diastolic phases
an effect on the dimensions of the PDMS cuffs relative to the of the cardiac cycle, and the pressure sensor was able to detect
CAD model. Here, we selected these parameters based on our the same changes. Having tested different cuff sizes provided a
previously published Expert Guided Optimization approach.[17] means to study the effect of contact pressure on signal ampli-
Our results found that the absolute mean deviation of the PDMS tudes (Figure 2f,g). Contact pressures in the 2.5–15.5 kPa range
prints compared to the CAD model did not exceed 1 mm for that corresponded to 102% and 105% cuff sizes were found to
the finger (Figure 1c) or the toe cuffs (Figure S4 and Table S1, have the largest amplitudes (Figure 2g). Dresher and Mendelson
Supporting Information). The sizing analysis found linear cor- found a similar pressure range of 8–12 kPa was optimal to obtain

Adv. Healthcare Mater. 2020, 1901735 1901735 (6 of 9) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advhealthmat.de

higher PPG amplitudes.[7a] In the latter study, the contact pres- troduced a workflow to match the design to different anatomical
sure was changed crudely by pressing on the sensor, however, the locations, which can serve as a model for potential clinical imple-
results corroborated our findings on the implications of cuff fit mentation. This approach can also be used to determine the im-
on device performance.[7a] Additional studies will be required to plications of patient matching on the device performance. Look-
assess the effect of anatomic shape and sensor location on signal ing forward, the P3 -wearable can be improved with additional
quality. sensors, for example an accelerometer, to better track patient ac-
The PPG and pressure signals were different through each ac- tivity. Though designed for the finger and toe, this approach may
tivity (at rest, sitting, and walking) when testing the P3 -wearable be adapted for sensors in other anatomic locations. In addition,
on the toe with the fPCB on the nail in the 90% cuff size (Fig- future testing of the P3 -wearable should be done with a sample
ure 4). The signal when sitting was smaller than at rest likely population to evaluate the ability to conform to variable anatomy
because the pressure exerted by the ground on the bottom of the between individuals. Ultimately, commercial translation will re-
toe deformed the PDMS cuff and slightly decreased fPCB con- quire meeting the U.S. Food and Drug Administration (FDA)
tact with the nail. Still, both at rest and sitting, the P3 -wearable guidelines on the requirements for receiving a 510(k) approval
produced usable signals to extract the heart rate and blood oxy- to market oximeters as a medical device.
gen content. The calculated values were comparable to the output
from a commercial oximeter worn simultaneously on the index
finger with a ±2% accuracy (Figure 4a,b).[21] The signal record- 5. Experimental Section
ings when walking could not be used due to motion artifacts as
Freeform Reversible Embedding (FRE) 3D Printing of Silicone Process
the foot was lifted and then the foot hit the ground. This caused and Preparation: The FRE 3D printing process has been previously
a change in pressure in the toe and thus the amount of blood vol- described.[16b,17] Briefly, a fused deposition modeling (FDM) 3D printer,
ume as well as contact with the sensor. However, for our purpose used to 3D print standard filament-based thermoplastics, was modified to
we do not need to be continually reading SpO2 , and taking mea- 3D print liquid silicone. The modification involved using the 3D printer
surements even as infrequently as once per hour during short to build a syringe-based extruder that then served to replace the origi-
nal extruder on the Makerbot Replicator dual 3D Printer (Makerbot In-
periods of rest is adequate. In this case the distinctive shape of
dustries, LLC, NY). Aside from hardware modification, FRE involved 3D
the pressure and PPG signals were indicative of walking and to- printing within a sacrificial support bath to sustain the build of liquid lay-
gether inform on patient activity status. Future iterations of this ers from the bottom up. The support bath was a yield stress fluid, the
device could automatically determine the activity state from the polyacrylic acid microgel Carbopol 940 (Lubrizol Corporation, OH) at 2%
PPG and/or pressure signal and then only calculate heart rate w/v. The bath was prepared by first mixing Carbopol 940 in deionized wa-
and blood oxygen content during sitting or rest states. ter and thereafter neutralizing the solution to pH ≈ 7 by adding sodium
We will need to perform longer term studies to evaluate perfor- hydroxide. The silicone elastomer used for 3D printing was Sylgard 184
(Dow Corning Corporation, MI), the preparation of which required mix-
mance of the P3 -wearable and any impact on the skin over time. ing a two-component base-to-catalyst at a 10:1 ratio by weight. Both bath
The benefit of a customized cuff over a bandage lies in avoid- (≈75 g) and the silicone were mixed prior to use for 2 min in a Thinky-
ing direct contact of the skin with chemical adhesives, which can Conditioning planetary centrifugal mixer (Phoenix Equipment Inc, NY) and
cause irritation and increase risk of infection in patients with ul- then defoamed for another 2 min at 2000 RPM. The silicone was loaded
cers, sensitive skin, and other complications. Similarly, the ben- into the syringe-based extruder of the 3D printer and extruded through a
efit of a customized cuff over a molded rubber band-type cuff is 17-gauge needle with a 1.219 mm inner diameter (Jensen Global Inc, CA)
and then inserted into the cup containing the bath that was placed onto
the ability to tailor to the anatomical shape of the body in order
the print platform. The 3D printing software Simplify3D (Simplify3D, OH)
to minimize pressure points while maintaining adequate signal. was used and the code was customized to complement the syringe-based
We recognize that traditional molding techniques remain feasi- 3D printing process. Each cuff printed in less than 2 min, after which the
ble to produce soft wearable cuffs of different sizes in limited print container was placed in an oven at 65 °C to heat cure the PDMS for
geometries. Future work should thus look toward assessing this 2 h. After this the printed cuff was manually removed from the Carbopol
(molding versus 3D printing) and other competing technologies support bath and then briefly washed off with water to remove any loosely
adhered bath residue.
through cost-benefit analysis to determine the most suitable for a
Finger Cuff Design, Sizing, and 3D Printing: The dimensions of the dis-
given application. Additionally, for the P3 -wearable we will need tal phalanx segment of a person’s index finger (major axis = 17.9 mm; mi-
to perform long term testing of durability and on a wider number nor axis = 13.7 mm; height = 23 mm) was measured with a digital caliper
of users as the device moves toward commercialization. at the widest point for the width (major axis) and thickness (minor axis).
The measurements were then used to design a cylindrical shaped cuff CAD
model. The CAD was also designed with a window to the size of the pulse
oximeter chip (6.35 mm width × 10.55 mm height) that was measured us-
4. Conclusion ing a caliper. The CAD model as is and with the pulse oximeter window
were 3D printed in silicone (Sylgard 184). A sizing analysis was performed
In this study, the P3 -wearable was developed by combining FRE to determine and compare the dimensional accuracy of the prints relative
3D printing of PDMS with a pulse oximeter and pressure sensor to the CAD models. The relative mean difference between the silicone cuff
chip on an fPCB. Pulse oximetry noninvasively measured blood and the CAD model from the sizing analysis was then used as correction
oxygenation and pulse rate by relying on the optical detection of factors (5.77% height, −9.49% ID minor axis, and −11.28% ID major axis)
volume changes in blood vessels. Traditionally, in the clinic a one- to design the final finger cuff CAD model (height = 21.67 mm; ID major
axis = 19.92 mm; ID minor axis = 15 mm). Similarly, correction factors
size-fits-all rigid device such as a finger clip is used for short-term for the pulse oximeter window (width = 18.67%; height = −14.3%) based
monitoring. In contrast, the P3 -wearable is wireless, small sized on the relative mean difference between the silicone print and the CAD
(≈8 cm), provides continuous real-time feedback on a tablet, and were used to design the final CAD window (width = 5.61 mm; height =
made from soft and flexible materials throughout. Further, we in- 12.06 mm; note: the corrected width is slightly smaller than the 18.67%

Adv. Healthcare Mater. 2020, 1901735 1901735 (7 of 9) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advhealthmat.de

correction factor). The final CAD model dimensions was thenceforth used the patient-matched silicone cuff and held in place by friction fit, using the
as the reference size model, 100% to the dimensions of the finger, which silicone-based adhesive Sil-Poxy (Smooth-On, Inc.) on the cuff window
served as basis to create other CAD sizes both smaller and larger (85%, brims. Each wearable was tested in three stages (Figure S2, Supporting
90%, 95%, 97.5%, 98%, 101%, 102%, 102.5%, 105%, 110%, and 115%). Information) that recorded while 1) detached and on a table (≈30 s), 2)
The window dimensions across CAD sizes were kept constant because the worn on the finger or toe (≈60–90 s), and again 3) detached and on a
size of the fPCB does not change. The wall thickness of all CAD models table (≈30 s). The data collected at a sampling frequency of 100 Hz was
was also maintained constant at 1.2 mm, the diameter of the print nozzle transmitted wirelessly at 15 Hz to a tablet via a Bluetooth low energy (BLE)
throughout the study. Each CAD size was printed trice and each dimen- module (Nordic nRF5122) powered by a lithium coin cell battery (CR1220).
sion (height, ID major and minor axes, wall thickness major and minor A graphical user interface (GUI) was developed in C# to display the PPG
axes, and window width and height) was measured trice by a caliper (n and pressure waveforms in real time, which were thereafter processed
= 108). All CAD models were designed using the Meshmixer (Autodesk, in MATLAB. Informed signed consent was obtained from coauthors S.A.
Inc.) software and 3D printed with the Simplify3D (S3D) software (extru- and E.J.M. for application of the body-mounted pulse-oximetry cuff (P3 -
sion multiplier = 1.4). wearable), which includes the recording of PPG, SpO2 , pressure, and/or
Toe Cuff Design, Sizing, and 3D Printing: A patient-specific CAD model any other data reported throughout all figures. For the period signals that
of the toe was developed by scanning a cast of a toe using the NextEngine the device was worn, the average peak amplitudes (IR PPG, red PPG, and
Desktop 3D scanner (NextEngine, Inc.). The cast was made from plaster pressure) that reflect arterial blood flow were extracted from a subset (30
using the Lifecasting Alja-Safe Starter Kit (Smooth-On, Inc.). Briefly, the s interval) of the recordings. The full detached–worn–detached recording
first step involved creating the Alja-Safe molding gel that required mixing set was used to obtain the contact pressure on the barometer. Specifically,
the Alja-Safe powder with water at a 1:1 mixing ratio by volume. The toe the average pressure offset while the cuff was detached was subtracted
was inserted into the gel and left in for the gel to set for 8–10 min. Sub- from the average pressure while the cuff was worn (Pcontact, toe/finger =
sequent removal of the toe produced a toe-shaped mold cavity for casting Pcontact, worn − Pcontact, detached ). To calculate heart rate and arterial oxygen
with plaster. The plaster was prepared by mixing with water at a 2:1 ratio saturation of hemoglobin (SpO2 ), data were processed by applying stan-
by volume. After pouring into the mold, the plaster was left to cure for ≈3 dard feature extraction techniques. The analysis was based on the modified
h. The life-size toe cast was scanned, repaired and then processed with a Beer–Lambert law.[23] Heart rate was calculated from the maximum peak
combination of software that include ScanStudio (NextEngine, Inc.), Net- between 0.8 and 1.8 Hz of a fast Fourier transform (FFT). The molar extinc-
fabb (Autodesk, Inc.), and Meshmixer (Autodesk, Inc.). The CAD model tion coefficients for red and near-infrared light for oxyhemoglobin and de-
was then offset to have a 1.2 mm wall thickness. The top of the toe was oxyhemoglobin (ɛ660 nm, HbO = 0.77 mm−1 cm−1 , ɛ660 nm, Hb = 7.921 mm−1
cut from the CAD model mainly to avoid nonadhering silicone layers dur- cm−1 , ɛ880 nm, HbO = 2.957 mm−1 cm−1 , ɛ880 nm, Hb = 1.936 mm−1 cm−1 )
ing 3D printing that resulted from the reduction in layer dimeters and for were obtained from.[24] The distance between the LEDs and photodetec-
potential comfort. To adjust for the dimensional accuracy of the prints, the tor was 3.8 mm. The ratio of differential pathlength factor (DPF), which
correction factors found from the finger cuff sizing analysis (5.77% height; accounts for wavelength dependent scattering differences, was set to 1.4
−9.49% ID minor axis; −11.28% ID major axis) were used in the sizing of based on Kohl et al.[25] The SpO2 is an average based on values calculated
the toe cuff CAD model. This CAD was considered the model to the di- using the fast Fourier transform (FFT) and the Hilbert transform. This out-
mension of the toe that is the 100% size. This and other CAD sizes were put was compared with a commercial oximeter, the Masimo MightySat
made with the window placed on the nail (85%, 90%, 95%, 100%, 102.5%, Fingertip Pulse Oximeter (Masimo Corporation), which was worn simul-
and 105%), beneath (85%, 90%, 95%, 97%, and 100%), and to the side taneously on the index finger.
(78%, 80%, 83%, 85%, and 100%) of the toe. The pulse oximeter window Statistical Analysis: For the finger cuff design, a one-sample t-test (n
was the same size as the fPCB previously measured and designed on the = 18, 𝛼 = 0.05) was performed to compare the average print dimensions
finger cuff (width = 5.61 mm; height = 12.06 mm). The toe was printed in to the CAD model. Additionally, statistical tests were performed on the
silicone (Sylgard 184) using the S3D software with an extrusion multiplier wearable hand and foot pulse oximetry signal output. Overall, the data
of 2 to achieve layer adhesion. passed the W/S normality test (n = 3, 𝛼 = 0.05) and were assumed normal.
Assessing the 3D Printed Toe Cuff Dimensional Accuracy Relative to the For the amplitudes of IR PPG, red PPG, and pressure from heartbeat, a
CAD Model: The dimensional accuracy of the printed cuffs was assessed one-way ANOVA (n = 3, 𝛼 = 0.05) and Tukey–Kramer post hoc test were
by a surface area and a deviation analysis with the original and a CAD applied separately to assess for statistical significance across the cuffs.
model obtained by scanning the prints. This was done for nine cuff sizes For the finger, the cuffs were of different sizes (100%, 102%, 105%, and
(90%, 95%, 97.5%, 100%, 102.5%, 105%, and 110%) with the pulse oxime- 110%). For the toe, the cuffs were of different sizes and placed at different
ter window designed on the nail. The reverse engineering of the 3D prints locations (nail 85%, nail 90%, nail 95%, nail 100%, beneath 85%, beneath
(n = 2) was performed using the FARO ScanArm laser scanner (FARO 90%, beneath 95%, beneath 100%, and side 83%). The contact pressure
Technologies, Inc.) for all sizes. The wall thickness of the printed toe was output that could be quantified for the hand wearable was subject to the
captured by scanning both the inside and outer surface. The deviation, in same statistical analysis as the finger cuff sizes. The statistical analysis
millimeters, between the CADs was determined by first aligning the two was performed using Microsoft Excel’s Analysis ToolPak feature. The data
models using the Geomagic Wrap (3D Systems, Inc.) software. A best-fit in the figures were presented as the mean ± the standard deviation (SD).
alignment was used with 1500 sample size and high precision fitting. Sub-
sequently, the built-in deviation analysis of Geomagic was used with the
original CAD set as the reference model. The surface area of each model Supporting Information
was determined through the Geomagic surface area analysis feature.
Pulse Oximetry and Pressure Sensing Wearable Fabrication, Testing, and Supporting Information is available from the Wiley Online Library or from
Data Processing: A flexible printed circuit board (fPCB) harboring the the author.
pulse oximetry and pressure sensor capability was developed (Figure
S1, Supporting Information) using a fabrication technique described
recently.[22] Briefly, a UV laser micromachining system was used to pat-
tern and assemble soft layers and rigid parts such as polyacrylic acid elas-
Acknowledgements
tomer and the pressure sensor, respectively. The MEMS piezoresistive The authors thank Prof. Jana Kainerstorfer for writing the MATLAB code to
pressure sensor was commercially available, the barometer cover was re- extract SpO2 and heart rate values from raw PPG signals, Niki Abdollahi
moved and embedded in the elastomer. The reflectance pulse oximeter for her contributions to the activity schematics, and Andrew Holmes for
sensor was also commercially available (MAX30102, Maxim Integrated) providing access to the FARO ScanArm laser scanner. This work was finan-
and the chip contained a red LED (660 nm), infrared LED (880 nm), a cially supported by the Disruptive Health Technology Institute at Carnegie
photodetector, and analog signal processing. The fPCB was inserted into Mellon University (A017261-HIGHMARK-Feinberg).

Adv. Healthcare Mater. 2020, 1901735 1901735 (8 of 9) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advhealthmat.de

Conflict of Interest [9] M. A. Meyers, P.-Y. Chen, A. Y.-M. Lin, Y. Seki, Prog. Mater. Sci. 2008,
53, 1.
A.W.F. has an equity stake in FluidForm Inc., which is a startup company [10] a) L. E. Dunne, H. Profita, C. Zeagler, J. Clawson, S. Gilliland, E. Y. Do,
commercializing FRE 3D printing. FRE 3D printing is the subject of patent J. Budd, presented at 2014 36th Annu. Int. Conf. on IEEE Engineering
protection including U.S. Patent 10150258 and others. in Medicine and Biology Society, August 2014; b) Y. Liu, M. Pharr, G.
A. Salvatore, ACS Nano 2017, 11, 9614.
[11] a) D. Jackson, A. M. Sarki, R. Betteridge, J. Brooke, Int. J. Nurs. Stud.
Keywords 2019, 92, 109; b) M. Schallom, D. Prentice, C. Sona, C. Arroyo, J.
Mazuski, Heart Lung 2018, 47, 93; c) C. Ceran, O. F. Taner, F. Tekin,
3D printing, health monitors, pulse oximeters, soft materials, wearable S. Tezcan, O. Tekin, B. Civelek, J. Pediatr. Surg. 2012, 47, e27.
devices [12] a) L. Gao, Y. Zhang, V. Malyarchuk, L. Jia, K.-I. Jang, R. Chad Webb, H.
Fu, Y. Shi, G. Zhou, L. Shi, D. Shah, X. Huang, B. Xu, C. Yu, Y. Huang,
Received: December 3, 2019 J. A. Rogers, Nat. Commun. 2014, 5, 4938; b) S. Harada, K. Kanao, Y.
Revised: May 3, 2020 Yamamoto, T. Arie, S. Akita, K. Takei, ACS Nano 2014, 8, 12851.
Published online: [13] a) M. K. Choi, O. K. Park, C. Choi, S. Qiao, R. Ghaffari, J. Kim, D. J.
Lee, M. Kim, W. Hyun, S. J. Kim, H. J. Hwang, S.-H. Kwon, T. Hyeon,
N. Lu, D.-H. Kim, Adv. Healthcare Mater. 2016, 5, 80; b) K.-I. Jang, S.
Y. Han, S. Xu, K. E. Mathewson, Y. Zhang, J.-W. Jeong, G.-T. Kim, R.
[1] J. W. Salyer, Respir. Care 2003, 48, 386. C. Webb, J. W. Lee, T. J. Dawidczyk, R. H. Kim, Y. M. Song, W.-H. Yeo,
[2] M. Kozioł, P. Piech, M. Maciejewski, W. Surtel, Acta Angiol. 2019, 25, S. Kim, H. Cheng, S. I. Rhee, J. Chung, B. Kim, H. U. Chung, D. Lee,
28. Y. Yang, M. Cho, J. G. Gaspar, R. Carbonari, M. Fabiani, G. Gratton, Y.
[3] a) N. Narula, A. J. Dannenberg, J. W. Olin, D. L. Bhatt, K. W. John- Huang, J. A. Rogers, Nat. Commun. 2014, 5, 4779; c) Y. S. Rim, S.-H.
son, G. Nadkarni, J. Min, S. Torii, P. Poojary, S. S. Anand, J. J. Bax, S. Bae, H. Chen, N. De Marco, Y. Yang, Adv. Mater. 2016, 28, 4415.
Yusuf, R. Virmani, J. Narula, J. Am. Coll. Cardiol. 2018, 72, 2152; b) [14] G. S. Jeong, D.-H. Baek, H. C. Jung, J. H. Song, J. H. Moon, S. W.
H. Reinecke, M. Unrath, E. Freisinger, H. Bunzemeier, M. Meyborg, Hong, I. Y. Kim, S.-H. Lee, Nat. Commun. 2012, 3, 977.
F. Lüders, K. Gebauer, N. Roeder, K. Berger, N. M. Malyar, Eur. Heart [15] C. Y. L. Chao, Y.-P. Zheng, G. L. Y. Cheing, Diabetes Technol. Ther. 2012,
J. 2015, 36, 932. 14, 602.
[4] a) Y. Mendelson, R. J. Duckworth, G. Comtois, presented at 2006 Int. [16] a) T. Bhattacharjee, S. M. Zehnder, K. G. Rowe, S. Jain, R. M. Nixon,
Conf. on IEEE Engineering in Medicine and Biology Society, August– W. G. Sawyer, T. E. Angelini, Sci. Adv. 2015, 1, e1500655; b) T. J. Hin-
September 2006; b) M. Savage, C. Pujary, Y. Mendelson, presented ton, A. Hudson, K. Pusch, A. Lee, A. W. Feinberg, ACS Biomater. Sci.
at 2003 IEEE 29th Annu. Proc. Bioengineering Conference, March 2003; Eng. 2016, 2, 1781; c) H. Kathleen, S. Bjorn, G. Christophe, L. Jared,
c) M. Tavakoli, L. Turicchia, R. Sarpeshkar, IEEE Trans. Biomed. Circuits T. Skylar, 3D Print. Addit. Manuf. 2017, 4, 123; d) C. S. O’Bryan, T.
Syst. 2010, 4, 27. Bhattacharjee, S. Hart, C. P. Kabb, K. D. Schulze, I. Chilakala, B. S.
[5] a) F. Adochiei, C. Rotariu, R. Ciobotariu, H. Costin, presented at Sumerlin, W. G. Sawyer, T. E. Angelini, Sci. Adv. 2017, 3, e1602800; e)
2011 7th Int. Symp. on Advanced Topics in Electrical Engineering R. L. Truby, J. A. Lewis, Nature 2016, 540, 371; f) A. Lee, A. R. Hudson,
(ATEE), May 2011; b) N. Constant, O. Douglas-Prawl, S. Johnson, D. J. Shiwarski, J. W. Tashman, T. J. Hinton, S. Yerneni, J. M. Bliley, P.
K. Mankodiya, presented at 2015 IEEE 12th Int. Conf. on Wear- G. Campbell, A. W. Feinberg, Science 2019, 365, 482.
able and Implantable Body Sensor Networks (BSN), June 2015; c) [17] S. Abdollahi, A. Davis, J. H. Miller, A. W. Feinberg, PLoS One 2018,
P. B. Crilly, E. T. Arakawa, D. L. Hedden, T. L. Ferrell, presented 13, e0194890.
at IEEE Instrumentation and Measurement Technology Conference [18] a) I. El-Katatny, S. H. Masood, Y. S. Morsi, Rapid Prototyping J. 2010,
Sensing, Processing, Networking (IMTC Proceedings), May 1997; 16, 36; b) D. Ibrahim, T. L. Broilo, C. Heitz, M. G. de Oliveira, H. W.
d) J. W. Lebak, J. Yao, S. Warren, in Proc. 25th Annu. Int. Conf. on de Oliveira, S. M. W. Nobre, J. H. G. dos Santos Filho, D. N. Silva, J.
the IEEE Engineering in Medicine and Biology Society, Vol. 4, IEEE, Cranio-Maxillofac. Surg. 2009, 37, 167; c) M. Salmi, K.-S. Paloheimo,
Cancun, Mexico 2003, p. 3196; e) M. Poh, N. C. Swenson, R. W. J. Tuomi, J. Wolff, A. Mäkitie, J. Cranio-Maxillofac. Surg. 2013, 41, 603;
Picard, IEEE Trans. Inf. Technol. Biomed. 2010, 14, 786; f) J. Yao, d) D. N. Silva, M. Gerhardt de Oliveira, E. Meurer, M. I. Meurer, J. V.
R. Schmitz, S. Warren, IEEE Trans. Inf. Technol. Biomed. 2005, 9, Lopes da Silva, A. Santa-Bárbara, J. Cranio-Maxillofac. Surg. 2008, 36,
363. 443; e) A. Unkovskiy, S. Spintzyk, D. Axmann, E.-M. Engel, H. Weber,
[6] a) K. Reinhart, H.-J. Kuhn, C. Hartog, D. L. Bredle, Intensive Care Med. F. Huettig, J. Prosthodontics 2019, 28, e460.
2004, 30, 1572; b) T. Tamura, Y. Maeda, M. Sekine, M. Yoshida, Elec- [19] R. Eftekhar Ashtiani, L. Nasiri Khanlar, M. Mahshid, A. Moshaverinia,
tronics 2014, 3, 282; c) S. Warren, Y. Jianchu, G. E. Barnes, presented at J. Prosthet. Dent. 2018, 119, 233.
Proc. Second Joint 24th Annual Conference and the Annual Fall Meet- [20] A. Unkovskiy, S. Spintzyk, D. Axmann, E. M. Engel, H. Weber, F. Huet-
ing of the Biomedical Engineering Society (Engineering in Medicine tig, J. Prosthodontics 2019, 28, e460.
and Biology), October 2002; d) M. W. Wong, H. F. Tsui, S. H. Yung, [21] B. Fahy, S. Lareau, M. Sockrider, in Am J Respir Crit Care Med, Ameri-
K. M. Chan, J. C. Cheng, J. Trauma: Inj., Infect., Crit. Care 2004, 56, can Thoracic Society, 2011, 1.
356. [22] a) M. D. Bartlett, E. J. Markvicka, C. Majidi, Adv. Funct. Mater. 2016,
[7] a) R. P. Dresher, Y. Mendelson, presented at 2006 Int. Conf. on IEEE 26, 8496; b) T. Lu, E. J. Markvicka, Y. Jin, C. Majidi, ACS Appl. Mater.
Engineering in Medicine and Biology Society, August–September 2006; Interfaces 2017, 9, 22055.
b) Y. Lee, S. Jung, Y. Seo, W. Chung, presented at 2008 Third Int. Conf. [23] a) E. M. C. Hillman, J. Biomed. Opt. 2007, 12, 051402; b) J. Kainer-
on Convergence and Hybrid Information Technology, November 2008; storfer, A. Sassaroli, S. Fantini, J. Biomed. Opt. 2016, 21, 101408; c)
c) R. Sokwoo, Y. Boo-Ho, H. H. Asada, IEEE Trans. Biomed. Eng. 2001, P. D. Mannheimer, Anesth. Analg. 2007, 105, S10.
48, 795. [24] S. Prahl, Optical Absorption of Hemoglobin, omlc.org/spectra/
[8] a) K. Xu, Y. Lu, K. Takei, Adv. Mater. Technol. 2019, 4, 1800628; b) hemoglobin (accessed: June 2018).
Y. Huang, X. Fan, S. C. Chen, N. Zhao, Adv. Funct. Mater. 2019, 29, [25] M. Kohl, C. Nolte, H. R. Heekeren, S. Horst, U. Scholz, O. Hellmuth,
1808509. V. Arno, Phys. Med. Biol. 1998, 46, 1771.

Adv. Healthcare Mater. 2020, 1901735 1901735 (9 of 9) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim