Articles
https://doi.org/10.1038/s41551-022-00859-5
Low-cost gastrointestinal manometry via
silicone–liquid-metal pressure transducers
resembling a quipu
Kewang Nan1,5, Sahab Babaee1,2,5, Walter W. Chan 2, Johannes L. P. Kuosmanen1, Vivian R. Feig2,
Yiyue Luo3,4, Shriya S. Srinivasan 1,2, Christina M. Patterson1, Ahmad Mujtaba Jebran1,2 and
Giovanni Traverso 1,2 ✉
The evaluation of the tone and contractile patterns of the gastrointestinal (GI) tract via manometry is essential for the diagnosis of GI motility disorders. However, manometry is expensive and relies on complex and bulky instrumentation. Here we report
the development and performance of an inexpensive and easy-to-manufacture catheter-like device for capturing manometric
data across the dynamic range observed in the human GI tract. The device, which we designed to resemble the quipu—knotted
strings used by Andean civilizations for the capture and transmission of information—consists of knotted piezoresistive pressure sensors made by infusing a liquid metal (eutectic gallium-indium) through thin silicone tubing. By exploring a range
of knotting configurations, we identified optimal design schemes that led to sensing performances comparable to those of
commercial devices for GI manometry, as we show for the sensing of GI motility in multiple anatomic sites of the GI tract of
anaesthetized pigs. Disposable and customizable piezoresistive catheters may broaden the use of GI manometry in low-resource
settings.
G
astrointestinal (GI) dysmotility can affect any part of the
alimentary tract and may manifest in or contribute to digestive conditions, including gastroesophageal reflux disease,
gastroparesis, intestinal pseudo-obstruction, irritable bowel syndrome, chronic constipation and faecal incontinence1. Not only do
these conditions rank among the most common patient presentations to outpatient clinics, they are also associated with substantial
morbidity, including malnutrition, feeding tube dependency, need
for invasive procedures, frequent hospitalizations and death2,3.
The current evaluation of patients with these symptoms involves
multiple diagnostic elements, with manometry playing one of the
most important roles1. GI manometers are catheter-like devices
containing a series of pressure transducers that can measure
real-time pressure changes along the length of the device when
placed endoluminally in the GI tract4. Several forms of this technology have been developed to evaluate the various segments of the
alimentary tract, including oesophageal, antroduodenal, colonic and
anorectal manometry. High-resolution manometry (HRM), which
consists of a higher number of pressure transducers spaced closer
together, emerged in the last decade and considerably enhanced
the identification of abnormal findings. Advances in hardware and
software technology further allowed the standardization of clinical
interpretation of manometry results, such as the development of
the Chicago Classification4, which facilitate and expand the clinical
utility of manometry.
Nonetheless, these systems suffer from high cost, complexity
and bulkiness, which limit their use in less developed regions
or non-hospital settings. Since physical examination findings,
conventional imaging studies and endoscopic evaluations are most
often unrevealing in patients with motility disorders, physicians in
resource-constrained settings may be limited in the assessment and
diagnosis of these conditions1. In addition, although most mano­
metry catheters can be re-used up to ~500 times, the complexity and
cost associated with disassembling and disinfecting place burdens
to even the most resource-rich regions or hospitals, leading not
only to increased risk of cross-contamination, but also reduced case
throughput. An easy-to-assemble, inexpensive, disposable and portable alternative could allow expanded use in regions with limited
resources, as well as potential reduction of infection risk associated
with the reuse of medical equipment.
Furthermore, a portable GI manometry with low-cost and disposable catheters affordable by average households can potentially
promote the decentralization of GI-related health care, which has
been the trend for modern medicine5, evident in the development of
radiography6 and electrocardiogram (ECG)7. For example, although
the single-lead ECG results obtained from Apple watch show less
resolved signals and higher false-positive rates than those from the
conventional 12-lead ECG performed during hospitalization, they are
US Food and Drug Administration (FDA)-approved8 for detection
of atrial fibrillation in non-hospital settings and are useful for early
detection of acute coronary syndrome in patients who are far away
from centralized medical resources. We therefore sought to explore
the potential for a low-cost, disposable and portable device that can
offer similar effectiveness and utility as commercial GI manometry.
A quipu is an ensemble of knotted cords, a device used by
Andean civilizations for storing and conveying information.
A simple and elegant system, the quipu utilizes the colours, spatial
distributions, as well as configurations of the knots to establish a
Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA. 2Division of Gastroenterology, Hepatology
and Endoscopy, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA. 3Computer Science and Artificial Intelligence Laboratory,
Massachusetts Institute of Technology, Cambridge, MA, USA. 4Electrical Engineering and Computer Science Department, Massachusetts Institute of
Technology, Cambridge, MA, USA. 5These authors contributed equally: Kewang Nan, Sahab Babaee. ✉e-mail: cgt20@mit.edu
1
Nature Biomedical Engineering | www.nature.com/natbiomedeng
Articles
form of logical-numerical recording that attracts academic interest9,10 Liquid metal represents a class of materials that is liquid at
or near room temperature and is highly conductive, enabling a
range of electronic applications in flexible and wearable sensors11–15.
Inspired by the quipu design and recognizing the potential of liquid
metal, here we describe the development of quipu-inspired, liquid
metal-enabled pressure transducers (QUILT) for achieving desired
pressure sensitivity across the dynamic range of the human GI tract.
We present simple fabrication schemes that can be completed using
basic bench tools, resulting in a device cost of less than US$0.26
per centimetre. We also exploit machine-aided fabrication and
finite-element (FE) simulations for enhanced sensor performances
and strategies for multiplexed measurements. Through in vitro
tests, we validate the system for pressure sensing in a wide range of
force scenarios. We further demonstrate preliminary clinical utility
of the system by investigating simulated oesophageal motility,
distention-induced oesophageal peristalsis (DIEP) and rectoanal
inhibitory reflex (RAIR) in a porcine model, and by benchmarking against the commercially available endoluminal functional
luminal-imaging probe (EndoFLIP) as well as HRM.
Results
QUILT. To date, the clinically applied GI manometry techniques
have relied on water-perfused catheters or solid-state (SS) trans­
ducers as the pressure-sensing elements (see Supplementary Table 1
for detailed comparison). It was therefore intuitive to seek new
manometry technologies that exploited: (1) liquid sensing elements
for low manufacturing cost and high sensor reconfigurability, similar to a water-perfused system16,17, (2) mechanoelectrical transductions for high temporal resolutions and interfacing with portable
recording hardware, similar to as SS system17, and (3) thin, tubular
geometries for facilitating placement endoluminally in the GI tract.
We identified eutectic gallium-indium (EGaIn) as the pressuresensing component due to its liquid nature under body conditions (melting point, 15.5 °C), low viscosity (1.99 × 10−3 Pa s), excellent electrical conductivity (3.4 × 106 S m−1), great moldability11,
low cytotoxicity15,18,19 and current use in several body-interfacing
pressure-sensing applications12–14,20. An extremely simple yet functional pressure sensor was built by infusing elastic medical cathe­
ters (for example, silicone tubing, outer diameters from 0.64 to
1.96 mm) with EGaIn and sealing both ends. The EGaIn-infused
catheter underwent cross-sectional narrowing if sufficient pressure was applied, resulting in an increase in the electrical resistance
across the EGaIn due to the piezoresistive effect. The resulting pressure sensor was made of medically approved encapsulation materials and integrated into a catheter configuration, which facilitated
clinical implementations. However, two major issues excluded this
simple system as a viable solution to GI manometry. First, the pressure generated from typical human GI contractions (that is, 0 to
250 mm Hg, or 0 to 33.3 kPa) was insufficient to cause considerable
narrowing of the catheter, leading to negligible resistive changes
(ΔR/R0) and pressure sensitivity (slopes of the red curves, Fig. 1a,
NATure BIomeDIcAl EnGIneerInG
characterized using a customized mechanical compression system
detailed in Supplementary Fig. 1). Second, there was no way to
identify the spatial location of the pressure source as the catheter
had a uniform cross-sectional area throughout.
Inspired by the quipu, we found that both issues can be simultaneously eliminated by simply tying knots on the EGaIn-infused
catheter. As illustrated in Fig. 1b, a large change in electrical resistance through the EGaIn-infused catheter was recorded by the
source metre when a small pressure within 50 kPa was applied onto
the knots, whereas no signal was detected when the pressure was
applied onto the neighbouring unknotted region. Not only did the
knots exhibit a highly linear (R2 > 0.985), low-hysteresis (hysteresis
error <0.5%) response in the human GI pressure range (black
curve, Fig. 1a), they also acted as spatial coordinates of the pressure
source since they were more pressure-sensitive than the neighbouring unknotted portions of the catheter. We speculated two reasons
for the enhanced pressure sensitivity due to knotting. First, the
process increased the aspect ratio of the catheter cross-section due
to shear stress, which made it more sensitive to deformations than
the unknotted case under the same loading conditions. Second,
similar to serially connected springs, each catheter layer that was
folded and stacked onto itself due to knotting experienced the same
external pressure component perpendicular to the stacking plane,
resulting in an amplification of the effective total pressure. To some
extent, the increases in the aspect ratio of the catheter cross-section
and in the number of the stacked catheter layers are analogous to
microbumps as stress concentrators20 and the meandering channel
designs13,14,20, respectively, for boosting the sensitivity of previously
reported EGaIn-based pressure sensors. Using simplified FE simulations (see Methods for more details), we verified that an increase
in the aspect ratio of the catheter cross-section as well as in the
number of the stacked catheter layers contributed to a larger change
in the total cross-sectional area under the same loading conditions
(Supplementary Fig. 2).
We experimented on a range of catheter diameters and found
that the resulting knots became increasingly distorted as the outer
diameter increased from 0.64 mm to 1.96 mm (Supplementary
Fig. 3), which manifested in both lower sensitivity (that is, slope
of the curve) and shorter linear regimes (Fig. 1a). Silicone tubing
with outer diameter and wall thickness of 0.64 and 0.17 mm, respectively, exhibited the highest linear sensitivity of 0.0084 mm Hg−1
in the human GI pressure range among the commercially available options (Corning Silastic laboratory tubing) and was therefore chosen for the rest of the study unless stated otherwise. These
QUILT showed a temporal resolution of ~10 Hz, as well as a stable
baseline for at least 400 s of continuous operation without using a
Wheatstone bridge circuit (Supplementary Fig. 4). To rigorously
evaluate the frequency response of QUILT, we performed cyclic
compression tests at several different frequencies between 0.1 and
20 Hz using Instron. As shown in Supplementary Fig. 5, the signal
fidelity was well-preserved at frequencies below 5 Hz. Interestingly,
at 10 Hz, a wave packet of frequency ~1.5 Hz started to emerge,
Fig. 1 | QUILT. a, Representative (n = 1) resistive changes (ΔR/R0) of liquid metal-infused silicone tubing with different outer diameters (0.64–1.96 mm),
with or without an overhand knot, as a function of applied pressure. Samples with no knots showed close-to-zero pressure sensitivity defined by the
slope of the curve. Samples with knots showed less linear sensitivity as the outer diameters of the tubing increased. Samples with 0.64 mm outer
diameter exhibited the highest linear sensitivity of 0.0084 mm Hg−1 and the largest linear range from 0 to 375 mm Hg. b, Schematic of sensor operation.
A large change in electrical resistance was detected only when the pressure was applied onto the knots. c, Schematic of the fabrication procedures.
Inset: minimum requirement of bench tools and materials for the assembly of QUILT. d, Photograph of quipus. Credit: National Museum of the American
Indian, Smithsonian Institution (14/3866). Photo by NMAI Photo Services. e, ΔR/R0 of 12 different knot types at high (150 mm Hg) and low (15 mm Hg)
pressures, demonstrating a wide distribution. Data reported as mean ± s.d. for n > 5 measurements for each group. f, Optical images of the 12 different
knot types in e and the corresponding large-scale models to illustrate the knotting process. Scale bars, 2 mm. g, ΔR/R0 at 150 mm Hg pressure as a
function of temperature. The coloured curve shows the average of 5 samples. h, ΔR/R0 at 150 mm Hg pressure as a function of soaking time in 37 °C
phosphate-buffered saline. The coloured curve shows the average of 5 samples. i, ΔR/R0 at 150 mm Hg pressure as a function of number of autoclave
cycles. The coloured curve shows the average of 5 samples.
Nature Biomedical Engineering | www.nature.com/natbiomedeng
Articles
NATure BIomeDIcAl EnGIneerInG
and the spike magnitude became non-uniform. We speculated
that this was caused by either the viscoelasticity of silicone or the
drifting errors of Instron at higher frequencies. The spikes became
distorted and no longer distinguishable at 20 Hz and beyond.
Pressure (kPa)
0
10
20
30
No knot 0.64 mm
No knot 0.94 mm
No knot 1.19 mm
No knot 1.96 mm
Knot 0.64 mm
Knot 0.94 mm
Knot 1.19 mm
Knot 1.96 mm
3
∆R/R0
2
40
b
50
c
Large signal
∆R/R0
a
Overall, the frequency range of QUILT (~0.1–5 Hz) is notably lower
than commercial pressure sensors based on piezoelectric materials,
but should be adequate for evaluating GI motility that has a typical
frequency on the order of 1 Hz and lower.
No signal
Silicone tube
No signal
Liquid metal
0.0084 mmHg–1
Less linear
Knot
Silicone sealant
1
Close-to-zero
sensitivity
0
Copper wire
0
75
150
225
Pressure (mmHg)
300
d
375
e
3
150 mmHg
15 mmHg
2
∆R/R0
1
0
0.15
0.10
0.05
0
r’s
ilo
op
lo
e
bl
D
ou
Sa
9
pu
8
ui
Q
pu
7
ui
Q
ui
pu
6
Q
pu
5
Q
ui
pu
4
ui
Q
pu
3
ui
Q
pu
Q
ui
N
on
e
(o Qu
ve ipu
rh 1
an
(fi Qu d)
gu ip
re u
ei 1
gh
t)
Q
ui
pu
2
–0.05
f
Quipu 1
(overhand)
Quipu 1
(figure eight)
Quipu 2
Quipu 3
g
Quipu 4
Quipu 6
Quipu 7
h
Quipu 8
2
Quipu 9
Double
loop
Sailor’s
i
150 mmHg
150 mmHg
2
150 mmHg
50
60
70
Temperature (°C)
Nature Biomedical Engineering | www.nature.com/natbiomedeng
0
1
2
3
4
Elapsed time (d)
5
6
Au
cy toc
cl la
e ve
10
40
Au
cy toc
cl lav
e
5 e
30
cy toc
cl lav
e
1 e
0
0
20
1
Au
0
1
au Bef
to or
cl e
av
e
1
∆R/R0
2
∆R/R0
∆R/R0
Quipu 5
Articles
Key advantages of QUILT are the simple fabrication procedures
and low cost compared with existing GI manometry devices. The
fabrication involves only three steps, with basic bench tools such
as syringes, scissors and fast-drying silicone sealants (Fig. 1c, see
Methods for details), and can be completed by a moderately trained
person in less than an hour, amounting to an overall device cost of
less than US$0.26 per centimetre. In addition, a rudimentary digital
multimetre suffices as the minimal requirement for data recording
and display, as opposed to the bulky perfusion pumps required by
water-perfused systems, or the dataloggers specific to each SS system that are often non-interchangeable among different models.
The simple and robust fabrication procedures also allow for quick
customization of the sensor configurations and reconfigurations
without additional cost, and interfacing with various other tools
and substrates to accommodate different portions of the GI tract
with distinct anatomy and motility profiles, as were demonstrated
in the in vivo implementations below. These and other features
make QUILT promising for deployment in less-developed regions
and non-hospital settings where state-of-the-art manufacturing
equipment and external hardware may be limited.
Inspired by the quipu (Fig. 1d), we explored the effects of different knot types on pressure sensitivity. Resistive change (ΔR/R0)
of 12 different knot types at low pressure (15 mm Hg) and high
pressure (150 mm Hg) exhibited a wide distribution (Fig. 1e,f). It
is worth noting that the knots generated from the silicone tubing
deviated geometrically from the large-scale models using polypropylene ropes as the quipu numbers increased (Fig. 1f), probably due
to the viscoelastic nature of silicone materials. This simple study
demonstrated the potential to tailor sensitivity for specific application needs using different knot geometries, a concept similar to
differentiating numbers and letters in the quipu. We further tested
alternatives to knots as localized stress concentrators on the silicone
tubing, such as O-rings and/or ultraviolet (UV) curing adhesive,
and three-dimensional (3D)-printed micro-fixtures, none of which
resulted in as good linear sensitivity as the overhand knots in the
GI-relevant pressure range (Supplementary Fig. 6). Methods such
as laser texturing21 could be used to modify the tubing surface to
induce changes in mechanical behaviours in response to pressure,
but this was not pursued here due to increased complexity and
cost. The overhand knots were used for the rest of the study unless
stated otherwise.
Next, we characterized the robustness of QUILT through
a heating test up to 70 °C (Fig. 1g) and a soaking test in 37 °C
phosphate-buffered saline up to 1 week (Fig. 1h), during which the
changes in sensitivity were within 4.3% and 6.1%, respectively. The
thermal influence on sensitivity (~4.3%) was comparable and on the
same order of magnitude as those reported in the literature20,22,23,
which can be further improved by the Wheatstone bridge circuit14
NATure BIomeDIcAl EnGIneerInG
that was not included in this work due to cost and manufacturing
considerations. In addition, we performed storage tests of QUILT
under normal laboratory conditions (22 °C, ~20% relative humidity)
for 1 week. The averaged changes in ΔR/R0 between day 0 and
day 6, −2.4% (Supplementary Fig. 7), were smaller than those from
the heating and soaking tests, indicating that QUILT has a good
shelf stability of at least 1 week.
Another important feature of reusable medical devices is the
compatibility with an autoclave for quick and effective sterilization,
one that SS systems lack due to the delicacy of the electronic components17. We found that the change in sensitivity of QUILT was
within 5.2% after undergoing at least 10 standard autoclave cycles
(121 °C, 1 atm for 30 min; Fig. 1i and Supplementary Fig. 8).
To determine the degree of potential EGaIn leakage and assess
safety of the device in vivo, we immersed QUILT in simulated gastric fluid (pH 2) at 37 °C for 1 h, which represents one of the harshest
conditions these devices will encounter in the GI tract. No change in
colour of the solution was found after 1 h. We then dried the devices
overnight and found the weight changes before and after immersion
to be only 0.9 ± 0.2%, indicating that almost no substance exchange
has occurred with the surrounding medium under these extreme
testing conditions.
Experimental and numerical approaches to enhanced sensor performance. The notion of the ability to assemble the entire QUILT
with basic bench tools is tempting to achieve maximum simplicity
and frugality, but is not without challenges. For instance, the resulting percentage uncertainties in the measured sensitivity can be up
to ~43% for certain knot types (Fig. 1d). To address this issue, we
developed a mechanical stretching system with an integrated force
gauge (Fig. 2a) for precise control over the knotting process. For
our case, it was determined that a tensile force of ~0.1 N yielded
high-quality knots in terms of consistency and sensitivity (Fig. 2b).
Another potential issue that can be detrimental in certain applications is the change in sensitivity after the QUILT has been longitudinally stretched, as the knots will further tighten due to the applied
tensile force. We addressed this by applying ~0.2 ml of UV curing
adhesive (Loctite Si 5055) on the knots (Fig. 2b) that, upon curing,
locked its shape against stretching, at the expense of some sensitivity (Fig. 2c). We observed a considerable drop in percentage uncertainty, ~5 times at 150 mm Hg applied pressure, after the machine
and/or UV curing adhesive treatment (Fig. 2c,d). Through a cyclic
stretching test, we verified that although UV curing adhesive treatment reduced the initial sensitivity by half, it successfully preserved
the value upon 500 stretching cycles with 50% tensile strain, a considerable improvement compared with the untreated ones (Fig. 2e).
Next, we employed FE simulations to characterize the mechanical
response of elastic overhand knots with different design parameters.
Fig. 2 | Experimental and numerical approaches to enhance sensor performance. a, Photograph of the customized mechanical stretching system used to
form consistent elastic knots. b, Optical images of machine-tied knots with or without UV curing adhesive. Comparison of the identical yellow reference
crosses showed good qualitative geometric consistency. UV curing adhesive resulted in slightly expanded knot volume. Scale bars, 1 mm. c, Resistive
changes (ΔR/R0) of hand-tied, machine-tied, and machine-tied with UV curing adhesive samples as a function of applied pressure. Data reported as
mean ± s.d. for n > 5 measurements for each group. d, Percentage uncertainty of the 3 knot samples reported in c at 150 mm Hg pressure. Machine-tied
samples (6.1%) had ~5× lower percentage uncertainty than hand-tied samples (31.6%). e, ΔR/R0 of hand-tied, machine-tied, and machine-tied with
UV curing adhesive samples at 150 mm Hg pressure as a function of stretching cycle with 50% tensile strain. The values for hand-tied and machine-tied
samples increased by ~3-fold after the first 100 cycles, whereas that for samples with UV curing adhesive remained almost constant after 500 cycles.
f, FE models of an elastic tube, initially knotted into an overhand shape (top) via pulling the extremities, ∆x, and then compressed through application of
normal displacement, ∆z, using a rigid plate (bottom). T and F are the corresponding tensile and compression forces, and H0 is the initial height of the knot
in the z direction. g–k, The evolution of required T as a function of ∆x and normalized displacement, ∆x/L0, for an overhand knot obtained from FE models
(black) and the experiment (blue) (g). Experimental data reported as mean ± s.d for n > 5 measurements. Numerical snapshots showing the configuration
of the knot at levels of ∆x/L0 = 1.18 (I), 1.38 (II), 1.79 (III) and 2.0 (IV) with the corresponding von Mises stress distribution are presented in h. The effect of
∆x/L0, initial elastic modulus E0 and wall thickness t0 on F as a function of ∆z/H0 are reported for knots with the same initial tube diameter D0 = 0.64 mm,
t0 = 0.17 mm and E0 = 470 KPa (i); knots with identical D0 = 0.64 mm, t0 = 0.17 mm and ∆x/L0 = 1.75 (j); and knots with identical D0 = 0.64 mm, E0 = 470
KPa and ∆x/L0 = 1.75 (k), respectively.
Nature Biomedical Engineering | www.nature.com/natbiomedeng
Articles
NATure BIomeDIcAl EnGIneerInG
The elastic tube assumed an initial outer diameter D0 = 0.64 mm,
length L0 = 32 mm, wall thickness t0 = 0.17 mm and elastic modulus
E0 = 470 KPa (vinyl polysiloxane silicone-based rubber, Zhermack).
In Fig. 2g, we report the evolution of uniaxial tensile force, T,
required to create an overhand knot as a function of the corresponding applied displacement between the extremities, ∆x, and its
normalized value, ∆x/L0, demonstrating a monotonic increase in T.
To confirm the numerical predictions, we experimentally measured
In particular, we first developed 3D FE models of an elastic tube in
the loop configuration (Supplementary Fig. 9, left panel), and performed dynamic explicit analysis to evaluate the behaviour of the
tube by pulling the extremities (that is, knot formation; see Fig. 2f
top panel, Methods and Supplementary Video 1 for details). Then,
the response of the knots under normal compression was assessed
by subsequent compression of knots using a rigid plate (see Fig. 2f
bottom panel, Methods and Supplementary Video 2 for details).
a
b
Machine-tied 1
Machine-tied 2
Machine-tied 3
UV curing adhesive
Knot
d 50
5
Hand-tied
Machine-tied
UV curing adhesive
∆R/R0
3
20
-ti
ne
10
hi
ac
H
0
M
an
0.0038 mmHg–1
0
0
75
150
225
300
Pressure (mmHg)
U
ad V c
he uri
si ng
ve
ed
0.0082 mmHg–1
30
d
1
0.012 mmHg–1
150 mmHg
4
tie
∆R/R0
3
2
5
40
d-
4
e
150 mmHg
Uncertainty (%)
c
1
0
Type of knotting procedures
375
Hand-tied
Machine-tied
UV curing adhesive
2
0
100
200
300
400
500
Stretching cycles
f
g
h
∆x/L0
0.25
T
0
0.6
T (N)
1.8
2.4
3.0
FEA
Experiment
0.20
∆x
1.2
IV
0.15
III
I. ∆x/L0 = 1.18
II. ∆x/L0 = 1.38
III. ∆x/L0 =1.79
IV. ∆x/L0 = 2.0
II
0.10
I
0.05
Y
Z
0
T
X
i
F
0.4
∆z
Z
0.3
F (N)
H0
0.2
10
20
30
40
Displacement, ∆x (mm)
j
∆x/L0 = 1.18
∆x/L0 = 1.38
∆x/L0 = 1.79
∆x/L0 = 2.0
0.4
0.3
0.2
Higher
sensitivity
0.1
0.1
X
50
Von Mises stress
0.32
(MPa)
k
E0 = 32 kPa
E0 = 80 kPa
E0 = 180 kPa
E0 = 470 kPa
0.3
0.2
Higher
sensitivity
0.2
Nature Biomedical Engineering | www.nature.com/natbiomedeng
0.4
∆z/H0
0.6
t = 0.19 mm
t = 0.18 mm
t = 0.16 mm
t = 0.15 mm
Higher
sensitivity
0.1
0
0
0.53 0.96 1.17 0.75 1.39 1.60
0
0.2
0.4
∆z/H0
0.6
0.2
0.4
∆z/H0
0.6
Articles
the force-displacement response during knot tying of silicone tubes
(D0 = 0.64 mm, length L0 = 32 mm) filled with EGaIn using Instron.
The FE results (black) were in close agreement with the experimental data (blue), therefore validating our FE results. In Fig. 2h,
we present numerical snapshots obtained from the nonlinear FE
simulations of the overhand knot at different levels of ∆x/L0—1.18
(I), 1.38 (II), 1.79 (III) and 2.0 (IV), clearly showing that the knot
became tighter with the applied tension, while the local stresses
increased throughout the knot.
We further examined the compression behaviour of knots
through the application of normal displacement, ∆z, and monitored
the compression forces in the normal direction, F, as a function of
normalized vertical displacement, ∆z/H0, where H0 is the height of
the undeformed knots. Particularly, we numerically investigated the
effects of ∆x/L0 and a range of tube parameters, including E0 and
t0, on F. In Fig. 2i, we report the evolution of F at different levels of
the aforementioned ∆x/L0, showing that different values of F were
required to deform the knots for a given applied ∆z/H0. Moreover,
these results suggest that the rates of variation in F against ∆z/H0
(that is, the slope of the curves) are higher for the tighter knots (that
is, the knots with higher ∆x/L0), a signature of enhanced pressure
sensitivity of the tighter knots. Finally, we report the evolution of
F as a function of ∆z/H0 for the knots made of various elastomeric
silicones with E0 = 32, 80, 180 and 470 KPa (Fig. 2j) and t0 = 0.15,
0.16, 0.18 and 0.19 mm (Fig. 2k), formed by applying ∆x/L0 = 1.75.
The results demonstrated a considerable larger F corresponding to
higher pressure sensitivity for the knots made of stiffer tubes (that
is, with higher E0), while the effect of t0 on F and pressure sensitivity
remained almost unchanged for the range examined. Together,
these results showcase the ability to quantitatively optimize the
knot configurations to allow for enhanced pressure sensitivity and
customizations to meet specific application needs.
Strategies for multiplexed measurements. We next investigated
strategies for multiplexed measurements using QUILT, which are
important for assessing mechanical activities along the length of
the GI segment being evaluated, up to 80–100 cm in some cases.
Due to the high degree of sensor reconfigurability, three modes of
multiplexing can be devised, each with different levels of fabrication challenges, total numbers of channels and functions that can be
tailored for targeted application needs. Mode 1 is the most common
approach, where each knot occupies one channel (Fig. 3a), similar
to conventional manometry. This mode allows for maximum spatial
resolution, which is suitable for identifying pathological oesophageal or antroduodenal motility24,25, or coordinated movements26
that demand simultaneous measurements at distinct GI locations,
but at the expense of being fabrication-heavy, bulky and expensive
on the recording hardware. Mode 2, where multiple knots are tied
onto a single tubing (Fig. 3a) that require only one multimetre as the
recording hardware, is the most economical in terms of device fabrication and data recording. Although it cannot spatially resolve the
signals if they occurred simultaneously, this mode may still be useful
in some cases such as evaluating multiple rapid swallow responses27,
where the pressure triggering of each knot along the path is known
to occur sequentially and directionally. In this case, a second channel would be placed at the upper oesophageal sphincter (UES) for
simultaneous monitoring of the UES contraction/relaxation. It may
also be possible to realize spatially resolved monitoring in mode
2 through a time-domain reflectometry approach28, although the
resulting increase in complexity and cost of the recording hardware
may deter its use in resource-limited environments. Finally, mode
3 exploits different combinations of knots at a given spot, inspired
by the binary number system (Fig. 3a and Supplementary Fig. 10).
In this case, n channels can resolve up to 2n – 1 sensory knots, and
temporally overlapping pressure responses may be resolved by
de-coupling the linear combinations of the resulting signals if the
NATure BIomeDIcAl EnGIneerInG
amplitude and frequency of each signal are known to be similar.
For example, if channel 1 showed an overlapping signal as channels
2 and 3 but with twice the amplitude, it can be resolved as two
signals from positions 100 and 111, respectively. This mode may
find utility in haptic or keyboard sensing14,29 where the amplitude
and frequency of each load are similar.
We designed two in vitro tests to validate the three modes of
multiplexing. QUILT containing eight (modes 1 and 2) or seven
(mode 3) sensory knots with ~5 cm spacing were placed inside a
small intestine simulator (~40 cm long, ~1.5 cm inner diameter;
SurgiReal) made of ultra-soft silicone rubber. The first test involved
rolling a solid cylinder (~100 g) from one end of the simulator to
the other, mimicking in vivo oesophageal swallowing under healthy
conditions (Fig. 3b). In all cases shown in Fig. 3c, the passage of the
cylinder through each sensory node was registered as a spike in the
multichannel resistance recorder (BK Precision, Model DAS240).
In mode 3, the total pressure at a given spatial coordinate can be
reasonably estimated by summing the resistive changes (ΔR/R0)
across all channels with overlapping temporal coordinates. The
second test involved dropping weights (~100 g) at random knot
positions along the simulator (Fig. 3d), mimicking the spatially
random high-pressure events that may indicate GI motility disorders. As shown in Fig. 3e, evaluating through mode 2 was unable to
resolve spatial information, whereas pressure recording at a given
time was the sum of all weights accumulated on the sensor. Mode 1
showed the comprehensive spectrum by displaying both magnitude
and position information for each sensory node at any given time.
The total pressure (or number of weights in this case) at a given
knot position in mode 3 can be estimated by summing the resistive changes (ΔR/R0) across all channels with overlapping temporal coordinates. In both tests, the knot positions deduced from the
measurements in mode 3 using the binary algorithm agreed well
with the actual experiments (Fig. 3c,e), demonstrating its potential
for spatially resolved measurements with reduced total number of
channels. The signal-to-noise ratios in all cases were more than 10,
which were adequate for evaluating real GI motility.
In vivo demonstration of QUILT for gastrointestinal manometry.
We further validated the utility of QUILT using a porcine model
(Yorkshire swine, 40–80 kg weight, see Methods for details) due
to its anatomical similarity to humans30. Specifically, oesophagus
and rectum were chosen to evaluate the system by measuring the
oesophageal pressure during the passage of artificial food bolus
and the rectoanal pressure during the rectoanal inhibitory reflex
(RAIR), respectively (Fig. 4a). In the first study, we designed a multichannel, ribbon-shaped manometry device (Fig. 4b) for recording oesophageal motility. Spatial information will likely be crucial
in the diagnosis of an oesophageal motility disorder, as dysmotility may present in the form of absent contraction, spastic or premature contractions, or simultaneous pressurization. In addition,
a high-pressure junction exists at the UES ~10 cm from the oral
cavity, which holds important diagnostic values for skeletal muscle
disorders, head and neck radiation, stroke and neurodegenerative diseases such as Parkinson’s disease4. As such, sensing mode 1
(Fig. 3a) was selected here for maximum spatial resolution, where
eight knots were assembled using a ~45-cm-long soft and flexible
medical-grade silicone gel tape (AWD Medical) as the substrate,
and a ~13-µm-thick low-density polyethylene film as the encapsulation (Fig. 4b). A ~5 cm spacing between each knot was chosen
to match the sensor spacing of conventional manometry catheters.
Detailed fabrication procedures for the device are illustrated in
Supplementary Fig. 11. A battery-powered wireless multichannel
resistance-analysing circuit was designed and manufactured by
Linkzill (Hangzhou, China) to allow for low-cost, portable (weighs
less than 60 g) and real-time recording and display of the data onto
an Android mobile application with a sampling rate of 14 Hz for up
Nature Biomedical Engineering | www.nature.com/natbiomedeng
Articles
NATure BIomeDIcAl EnGIneerInG
a
Mode 1
Mode 2
Mode 3
Ch 1
...
Ch 1
Ch 2
Ch 3
100 010
001 110
011 101 111
Ch 8
Channels needed for n sensors: n
Channels needed for n sensors: 1
Channels needed for n sensors: ~log2(n)
Spatially resolved:
√
Spatially resolved:
×
Spatially resolved:
Multiplexing:
√
Multiplexing:
×
Multiplexing:
√
×
b
1
∆R /R0
c
2
4
3
5
6
Mode 2
Mode 1
Ch 1
Ch 2
Ch 3
Ch 4
Ch 5
Ch 6
Ch 7
Ch 8
7
Mode 3
Ch 1
Ch 2
Ch 1
Ch 3
5
Binary code 100 010 001 110
Knot position 1
1s
2
3
011 101 111
4
5
6
7
Time (s)
Mode 1
Mode 2
Mode 3
1
2
3
4
5
6
7
Time (s)
Time (s)
Time (s)
d
1
8
2
3
4
5
6
7
1
8
2
3
4
5
6
7
e
∆R /R0
Ch 1
Ch 2
Ch 3
Ch 4
Ch 5
Ch 6
Ch 7
Ch 8
Ch 1
Ch 2
Ch 1
Ch 3
5
Time (s)
5s
Binary code
101
011
Knot position
6
5
110+
110 010+ 010
111
4+
4 2+ 2
7
001
3
Fig. 3 | Strategies for multiplexed measurements. a, Schematics of modes 1, 2 and 3 of multiplexing. Inset: optical images of the total device cross-section
of each mode. Scale bars, 1 mm. b, Schematics of the rolling test. c, Multichannel resistance recording of each mode from the rolling test. The dashed lines
in mode 3 indicate spikes with overlapping temporal coordinates. The binary codes and knot positions deduced from the measurements are listed at the
bottom, and agree well with the actual experiments. d, Schematics of the random drop test. Each block and cross represent dropping and removal of one
weight at a knot position, respectively. e, Multichannel resistance recording of each mode from the random drop test. The dashed lines in mode 3 indicate
spikes with overlapping temporal coordinates. The binary codes and knot positions deduced from the measurements are listed at the bottom, and agree
well with the actual experiments.
to eight channels, suitable for use in resource-limited settings (for
example, at home or outdoor). During the procedure, the device
was wrapped onto a thin, stiff supporting tube (for example, temperature probe or polyurethane feeding tube, ~ 3 mm in diameter)
and inserted via the oral route into the oesophagus until the channel
closest to the oral cavity displayed a jump in pressure, indicating
the correct positioning of the first sensor at the UES; X-ray imaging
(Supplementary Fig. 12) confirmed the proper device deployment.
The porcine swallowing reflex was considerably depressed under
Nature Biomedical Engineering | www.nature.com/natbiomedeng
anaesthezia, so we simulated food swallowing by attaching ~5 ml
of artificial food bolus made from mixtures of alginate and gelatin
solutions (see Methods for details)31 onto the tip of the endoscope
(Supplementary Fig. 13) and sliding it through the oesophagus. The
bolus was unlikely to alter or damage the knots during sliding due
to its smooth, edgeless surface finish. In the first experiment, we
slightly retracted and held the bolus after reaching the end of the
manometry device to simulate the backflow and retention of bolus,
respectively, which may be found in oesophageal motility disorders
Articles
NATure BIomeDIcAl EnGIneerInG
a
b
Kn
ot
1
1
Food bolus
...
Endoscope
Knot sensor
2
Kn
ot
2
8
Water
inflation
1
To stomach
c
46
Pressure
(mmHg)
d
150
30
Time (s)
100
Time (s)
50
0
Knots 1–3
Knots 4–6
g
80
60
40
20
0
40 cm
6
4
2
0
Ba
Ba
se
lin
e
10
To rectum
30 cm
50
10
0
15 cm
20 cm
30
f
10 cm
e
10
e
0 cm
lin
40 cm
se
30 cm
Recovery velocity (mmHg s–1)
20 cm
0
Esophagus
50
10
0
10 cm
Oral
cavity
Residual pressure (mmHg)
0 cm
0
Esophagus
30
Oral
cavity
Inflation volume (ml)
Inflation volume (ml)
Fig. 4 | In vivo demonstration of QUILT for gastrointestinal manometry. a, Schematics of the two studies on evaluating the oesophageal pressure during
the passage of artificial food bolus attached to the tip of an endoscope and the rectoanal pressure during RAIR using a porcine model. b, Optical image of
the ribbon-like manometry device containing eight knots designed for oesophageal pressure evaluation. Inset: zoomed-in view of one knot-based sensor
node. Scale bar, 5 mm. c, Pressure colour plot from a representative (n = 1) event of passage, retraction and retention of artificial food bolus in the porcine
oesophagus, generated from the raw data in Supplementary Fig. 12. d, Pressure colour plot from a representative (n = 1) event of simultaneous passage
of two artificial food boluses in the porcine oesophagus, generated from the raw data in Supplementary Fig. 13. e, Optical image of the manometry device
with 6 knots, designed for RAIR evaluation. f, Averaged (n = 3) residual pressure as a function of inflation volume, showing a decreasing trend as inflation
volume increased. g, Averaged (n = 3) recovery velocity as a function of inflation volume, showing an increasing trend as inflation volume increased. Data
in f and g reported as mean ± s.d. for measurements for each group.
(Supplementary Fig. 14a). The multichannel pressure recording
(Supplementary Fig. 14b) was converted into a pressure colour plot
in adherence to modern data representations for HRM4,24 (Fig. 4c),
where both events corresponding to the backflow and retention
of bolus were clearly registered and displayed. In another experiment, we attached two separate food boluses, ~5 cm apart, onto
Nature Biomedical Engineering | www.nature.com/natbiomedeng
NATure BIomeDIcAl EnGIneerInG
the tip of the endoscope and passed them down the oesophagus
(Supplementary Fig. 15a). The multichannel pressure recording
(Supplementary Fig. 15b) and the resulting pressure colour plot
(Fig. 4d) demonstrated well-distinguished simultaneous pressure
events due to the presence of two boluses, implying its utility in
detecting dysmotility with simultaneous contractions (for example,
achalasia, diffuse oesophageal spasm) and those that may lead to
food bolus retention. We noted that the overall measurement range
and sensitivity were more than adequate to identify UES hypotension or hypertension, with pressure differentials up to 180 mm Hg
for additional diagnostic values4.
In the second study, we performed the standard RAIR measurement where a Foley catheter (18 Fr) was inserted ~13 cm proximal
to the anal verge and inflated with water to induce a transient involuntary relaxation of the anal sphincter32. Sensing mode 1 (Fig. 3a)
was used here to distinguish simultaneous pressure changes in the
rectum and the anal canal that were ~10 cm apart. The 6-channel
QUILT consisted of 3 knots at the front with ~2 cm intervals for
recording in the rectum, and 3 knots at the rear with ~1 cm intervals
for recording in the anal canal (Fig. 4e). UV curing adhesive was
used to bond 6 channels into 1, yielding an overall device diameter of ~2 mm. The device slid easily ~15 cm proximal to the anal
verge with endoscopic assistance, which was further confirmed with
X-ray imaging (Supplementary Fig. 16). No twisting or entanglement of the device was found, a common issue for small-diameter
flexible catheters4. A Foley catheter was then inserted into the
rectum and rapidly inflated with different amounts of water (10, 30,
50 or 100 ml) followed by deflation, which was repeated three times
for each volume. The resistance of each channel was measured
simultaneously using the customized resistance-analysing circuit
(designed and manufactured by Linkzill, Hangzhou, China), with
a sampling rate of 14 Hz per channel. During a typical trial, a rise
in rectal pressure was registered by sensor 5 (~12 cm from the anal
verge) immediately after inflating the Foley catheter with water, and
a gradual drop in anal pressure was recorded by sensor 2 (~2 cm
from the anal verge) with a temporal delay. The anal pressure completely recovered to its baseline ~5–15 s after the Foley catheter was
deflated. The pressure responses of sensors 2 and 5 derived from
their resistive changes during all trials are plotted in Supplementary
Fig. 17. All three major phases (relaxation, plateau and recovery)
that describe the dynamic nature of the RAIR can be identified
(Supplementary Fig. 17), from which diagnostic information such
as the resting anal pressure, residual pressure and recovery velocity can be deduced32,33. For example, the residual pressure, defined
as the final pressure value of the relaxation phase32, reduced from
49.7 ± 2.5 mm Hg to 11.7 ± 1.2 mm Hg (Fig. 4f), while the recovery
velocity, defined as the linear slope of the recovery phase, climbed
from 0.71 ± 0.1 mm Hg s−1 to 4.75 ± 0.3 mm Hg s−1 as the inflation
volume increased from 10 ml to 100 ml (Fig. 4g).
Benchmarking QUILT against clinically available pressure
sensors. As a final demonstration, we benchmarked QUILT against
clinically available pressure sensors (EndoFLIP, HRM) for GI
motility evaluations. We first compared the in vivo performance
Articles
of QUILT with EndoFLIP, a well-established technology for sensing endoluminal distensibility that has clinically demonstrated
correlation with HRM34,35. EndoFLIP exploits a single solid-state
pressure transducer inside a balloon catheter that is inflated with
diluted saline to record the intra-balloon pressure, which can
precisely evaluate the endoluminal pressure with a resolution of
0.1 mm Hg, but lacks spatial resolution. We performed the above
in vivo oesophageal and RAIR measurements using QUILT and
EndoFLIP, and repeated the measurements on two porcine models.
The screen clips of EndoFLIP measurements are summarized in
Supplementary Fig. 18. During the oesophageal experiment, we
first identified the location of the UES using an endoscope, placed
QUILT and EndoFLIP at the UES to record the resting UES pressure (P1), and then passed the bolus-attaching endoscope through
the UES to record the peak pressure as bolus passed (P2). We found
that P1 obtained from EndoFLIP was greatly affected by the choice
of initial inflation volume of the balloon catheter; larger inflation
volumes resulted in higher P1 values. We used a clinically recommended inflation volume of 20 ml, which yielded lower P1 and P2
compared with those recorded by QUILT. The averaged pressure
difference, P2 – P1, was within 10% between results from QUILT
and EndoFLIP (Fig. 5a), supporting the capacity of QUILT to evaluate relative pressure changes. During the RAIR evaluation with an
inflation volume of 10 ml using a Foley catheter in the rectum, we
noticed that the EndoFLIP balloon catheter (160 mm in length) was
too long to be placed entirely inside the anal canal, which may be
responsible for both lower resting anal pressures (P3) and residual
pressures (P4) compared with results from QUILT that sat completely within the high-pressure region of anal canal. The averaged
pressure difference, P3 – P4, was within 30% between the two systems (Fig. 5b). The percentage errors from multiple (n = 3) measurements were comparable (Fig. 5a,b), demonstrating good overall
consistency of the two systems in evaluating common GI motility
behaviours.
Next, we benchmarked QUILT against HRM, the current gold
standard for GI motility evaluations, both in vitro and in vivo using
porcine models. The system we acquired was a used Medtronic
ManoScan 360 Manometry equipped with oesophageal catheters.
The catheter had 36 solid-state pressure sensors spaced at 1 cm
intervals, with an overall diameter of ~4 mm. Each sensor was
further divided into an array of 12 circumferential solid-state
micro-transducers, and the final pressure recording in each channel
was an averaged value from all 12 circumferential transducers
during a total time span of ~2 s.
To better understand the performance of QUILT, we first conducted benchtop comparison with HRM (Supplementary Fig. 19a).
We ran two tests using (1) a set of different calibration weights (20,
50, 100 and 200 g; Supplementary Fig. 19b) placed onto the pressure sensors, and (2) a rolling test similar to Fig. 3b using a 50 gram
calibration weight. The total number of knots in QUILT was set
to 8, with an average spacing of ~2 cm for this test. As depicted
in Supplementary Fig. 19c, both HRM and QUILT were able to
distinguish different weights with good repeatability, while QUILT
performed better in recognizing the smaller weights (20 g and
Fig. 5 | Benchmarking QUILT against clinically available pressure sensors. a, Averaged (n = 3) difference between peak pressure as bolus passed (P2)
and resting UES pressure (P1) recorded by QUILT (blue) and EndoFLIP (red). Data reported as mean ± s.d. for measurements for each group. b, Averaged
(n = 3) difference between resting anal pressure (P3) and residual pressure (P4) recorded by QUILT (blue) and EndoFLIP (red). Data reported as
mean ± s.d. for measurements for each group. c, Photograph of QUILT attached onto a polyurethane tube next to an HRM catheter, demonstrating the
geometric and mechanical similarity. d, Schematic of the experimental setup during evaluation of DIEP. e,f, A representative (n = 1) multichannel pressure
recording of DIEP by HRM (e), with zoomed-in view in f. g, Pressure colour plot generated from the raw data in e. h, Schematics of the layout of knots and
channels for measuring DIEP using QUILT. i, A representative (n = 1) multichannel pressure recording of DIEP by QUILT. j, Pressure colour plot generated
from the raw data in i. Despite the poorer spatial resolution than in g, the essential features such as antegrade and retrograde peristalsis and LES
contraction were recognizable. k, Averaged (n > 3) peak peristaltic pressure and averaged (n > 3) peak LES pressure recorded by QUILT (blue) and HRM
(red). Data reported as mean ± s.d. for measurements for each group.
Nature Biomedical Engineering | www.nature.com/natbiomedeng
Articles
NATure BIomeDIcAl EnGIneerInG
50 g) than HRM. The pressure magnitudes recorded by QUILT
were larger than those by HRM, which made sense as QUILT had
a notably smaller diameter than HRM (Supplementary Fig. 19a),
and each applied pressure was estimated by the weight divided
b 25
80
20
P3 – P4 (mmHg)
100
P2 – P1(mmHg)
a
by the cross-sectional area of the catheter. In the second test, we
rolled a 50 gram calibration weight across a fixed length (~15 cm)
on both HRM and QUILT in a total time of ~2 s. Notably, as shown
in Supplementary Fig. 19d, the temporal resolution of QUILT
60
40
15
10
5
20
0
Yorkshire 1
QUILT
0
Yorkshire 2
Yorkshire 1
Yorkshire 2
EndoFLIP
c
d
Air inflation
QUILT
LES
HRM catheter
Stomach
Knot sensor
e
g
f
80
Ch 23
LES
cm
A
pe nte
ris gra
ta de
ls
is
(6
Antegrade
peristalsis
100
Ch 10
50
200 mmHg
R
pe etro
ris gr
ta ad
ls e
is
(9
cm
)
200 mmHg
Balloon
Ch 1
150
Retrograde
peristalsis
0
0 cm
0.5 cm
A
R
LE
pe etro pe nte
S
ris gr ris gra
ta ad ta de
ls
ls e
is
is
0.5 s
6
Pressure
(mmHg)
Antegrade
peristalsis
Time (s)
Retrograde
peristalsis
100
50
LES
0
Ch 2
Ch 1
10 cm
Stomach
20 cm
200
Pressure (mmHg)
j
100
mmHg
Ch 1
1.5 cm
17 cm
Ch 3
150
Ch 3
Oral cavity
k
Ch 2
Balloon
Antegrade
peristalsis
0
i
LES
Retrograde
peristalsis
Ch 1
5s
20 s
h Stomach
Time (s)
Ch 10
Pressure
(mmHg)
)
4
cm
L
(3 ES
cm
)
Ch 23
0 cm
Stomach
150
100
50
0
Peak peristaltic
pressure
QUILT
Peak LES
pressure
HRM
0
17 cm
Nature Biomedical Engineering | www.nature.com/natbiomedeng
NATure BIomeDIcAl EnGIneerInG
(4 Hz dictated by the recording hardware) was better than that of
HRM (0.5 Hz) in revealing detailed pressure profiles during fast
travelling of the object at a speed of ~7.5 cm s−1
Furthermore, we performed in vivo evaluations of HRM and
QUILT using a female Yorkshire swine (weight, 39 kg). To mimic
the overall dimensions and mechanical stiffness of HRM, we
attached QUILT onto a polyurethane tube with similar outer diameter (~4 mm) and radius of curvature (Fig. 5c). This time, instead
of passing the artificial bolus, we were able to induce secondary
peristalsis of the oesophageal body in lightly anaesthetized swine
(1.8–2% isoflurane in oxygen for under 30–60 min) by placing a balloon catheter transorally into the oesophagus at a depth of ~60 cm
and inflating with ~15 ml of air as illustrated in Fig. 5d. Using the
HRM catheter placed transorally into the oesophagus at a depth
of ~85 cm, we recorded symmetric, bidirectional peristaltic waves
initiated at the location of balloon inflation (Fig. 5e). Both antegrade
and retrograde peristaltic waves were observed as the oesophagus
reflexively tried to push the inflated balloon upward or downward
in efforts to dislodge it. The antegrade peristaltic wave culminated
with a contraction in the lower oesophageal sphincter (LES) ~10 cm
away from the site of balloon inflation (Fig. 5e), probably representing the rebound contractions often seen in the human LES
immediately at the end of a peristaltic swallow. Although details
were elusive in the time domain (see zoomed-in view in Fig. 5f)
due to relatively poor temporal resolution (0.5 Hz), the fine spatial
resolution (~1 cm) of HRM enabled pressure colour plotting of
the oesophageal motility (Fig. 5g) that resolved the essential
components, such as the antegrade and retrograde peristalsis, and
the LES contraction.
Knowing that the distention-induced peristaltic wave was bidirectional, we configured QUILT as illustrated in Fig. 5h where three
independent channels spanning a total length of ~17 cm recorded
the antegrade, retrograde and LES pressure activities. Pressure spectra during a typical peristaltic event recorded by QUILT (Fig. 5i)
revealed more features than those by HRM (Fig. 5f) due to the better temporal resolution (~4 Hz dictated by the recording hardware).
The spacing between each sensory node of QUILT (~1.5 cm) was
slightly larger than that of HRM (~1 cm), which resulted in lower
spatial resolutions; however, the resulting pressure colour plot
(Fig. 5j) was able to capture the essential information shown in
Fig. 5g as obtained from HRM. Finally, we compared the peak pressure during the secondary peristalsis and during the contraction of
LES as recorded by HRM and QUILT (Fig. 5k) and found them to
be within 20% difference, with comparable error bars. These findings provide preliminary experimental support for benchmarking QUILT against the commercial HRM in evaluating certain
GI motility patterns, although extensive future experimentations,
developments and clinical trials are necessary before QUILT could
be deployed clinically.
Discussion
Together, these results show that QUILT can operate under in vivo
conditions for at least 2 h and extract real-time, spatially resolved
information on GI motility across the dynamic range consistent with
human readings. The liquid metal-infused, knotting-based nature
of sensor fabrication allows for a high degree of sensor reconfigurability tailored specifically for different application needs and budget considerations. For example, a variety of sensor configurations,
such as number of independent channels (1, 3, 6 and 8), number of
knots per channel (1, 3, 4, 7 and 8), knot spacing (0.5, 1, 1.5, 2 and
5 cm) and total catheter lengths (15, 17, 40, 45 cm and more) has
been exploited in this work to facilitate and accommodate different
scenarios of pressure sensing both in vitro and in vivo, highlighting
the ability of QUILT to customize and reconfigure. The incorporation of multiple knots does not show interference or deterioration of
signal quality, suggesting the possibility of further increasing node
Nature Biomedical Engineering | www.nature.com/natbiomedeng
Articles
density to <1 cm spacing that may surpass the spatial resolution
of state-of-the-art HRM (~1 cm). Such level of spatial resolution
may be needed to help identify physiologically distinct anatomical
segments, such as the proximal (skeletal muscle) portion versus
the distal (smooth muscle) portion on oesophageal manometry,
although there is currently limited clinical evidence showing further
improvements in diagnostic value as sensor spacing decreases below
2 cm36. It is useful, however, from a machine-learning perspective, as
the classification accuracy generally improves with the density of
sensors37. This implies that our approach may offer a simple and
scalable approach towards unsupervised data analyses and diagnoses by achieving higher packing density of the sensors through
knotting configurations.
From a sensor and design perspective, we highlight two key
innovations that may be appealing to the biomedical community.
The first innovation is the discovery of using basic knotting configurations to transform the otherwise insensitive silicone/EGaIn
composites into devices capable of detecting small pressure changes
within the human GI tract. This finding allows us to use low-cost,
accessible materials and fabrication schemes while avoiding any
complex materials or architectures that require lengthy preparation or expensive equipment, thereby providing cost-effective
and disposable solutions that are in sharp contrast to the existing
manometry systems. The second innovation is the introduction of
multiplexing strategies by tying multiple knots onto a single conductor to minimize the number of channels without the need for
complex and expensive multiplexer circuits, in a manner analogous
to the historical quipu devices. For conventional manometry cathe­
ters where each solid-state pressure transducer needs to be individually addressed using multiple wires, the total number of wires
quickly multiplies to a degree that limits the maximum number of
sensors that can be packed within one catheter, thereby increasing
the overall manufacturing challenge. The wire bundle also increases
the diameter and mechanical stiffness of the catheter, which can
lead to increased patient discomfort and risk of injury. We expect
that our approach can inspire solutions to address the above challenges by exploiting liquid-based conductors that are intrinsically
soft and configurable, and by exploring the concept of using shared
conductors for multiple pressure transducers to increase the overall
efficiency of multiplexing.
However, a notable limitation of QUILT is the vulnerability
to pressure variations due to sensor orientations and placements
in vivo, which contrasts with HRM’s ability to measure pressure
from up to 12 radial directions at each sensor position, with the
resulting reading being an average of these measurements4. We
believe that the issue can be addressed by optimizing the knot
geometries to minimize directional heterogeneity, pursuing further
miniaturization of the elastic tubing38 and incorporating a robotic,
knitting-based manufacturing process39 to allow circumferential
packing of multiple knots into an integrated device.
An important consideration in potentially translating QUILT to
clinical use is the choice of the most appropriate sensing mode for
each clinical indication. For example, spatial information is important for evaluating oesophageal motility disorders, but may not be as
crucial in the assessment of anal sphincter tone or squeeze pressure.
The choice of the appropriate testing medium is also important for
oesophageal motility evaluations. In the in vivo portion of this study,
an artificial food bolus using a composite of alginate and gel was
chosen. It is unclear whether the measurements by QUILT resulted
from pressure exerted by the food bolus as it passed by each sensor,
or from the actual lumen-occluding contractions of the oesophageal
smooth muscles. The performance of QUILT in the setting of abnormal motility remains to be evaluated. Whether the sensitivity and
pressure resolutions remain similar at the extreme ranges of pressure measurements (that is, hypomotility and hypermotility) needs
to be further validated. Nevertheless, QUILT appear to be capable of
Articles
generating data on pressure amplitude and temporal pattern similar
to that demonstrated by the existing manometry system. This represents a potentially cheap, convenient and simple option for the evaluation of GI motility. From a practical standpoint, the availability of
an ultra-low-cost manometry device for one-time use would allow
expansion of this technology to regions with limited resources. Even
in resource-rich regions or institutions, such disposable devices can
further minimize the risk of cross-contamination and increase the
throughput of scheduled cases, as there would not be a need for
built-in time in between cases for device decontamination. Other
advantages of disposable catheters include avoiding the recurring
maintenance and repair costs of the current expensive catheters, as
well as minimizing potential delay in clinical care when multi-use
catheters are out for repair.
Methods
Fabrication of QUILT. The step-by-step fabrication flow is illustrated in Fig. 1b.
Fabrication started with trimming silicone tubing (Dow Corning Silastic laboratory
tubing) into the desired length, followed by injecting eutectic gallium-indium
(EGaIn, Sigma-Aldrich) using a needle syringe. Copper wires were inserted into
both ends of the tubing to establish electrical connections, followed by sealing
both ends using fast-drying silicone sealant (kwik-cast silicone sealant from WPI,
or sil-poxy rubber silicone adhesive from Smooth-on, or any other commercially
available adhesives that have good adhesion to silicone and a reasonable curing
time). Tying knots at designated positions directly by hand, or first by hand loosely,
then by a mechanical stretcher (Mark-10, model ES20) using designated tensile
force (for example, ~0.1 N in this study) and holding for ~30 s (Fig. 2a) completed
the fabrication of QUILT. For applications that require resistance to tensile
stretching, a drop of UV curing adhesive (~0.2 ml, Loctite 5055) was applied onto
the knots, followed by curing using a hand-held UV lamp (Loctite EQ CL32 LED
spot 365 nm) for ~120 s.
To assemble multichannel QUILT for in vivo rectoanal evaluations, each
channel was fabricated individually using the above procedures, then aligned and
positioned closely. UV curing adhesive was applied at multiple locations along the
device to secure individual tubing into one.
To fabricate the ribbon-like manometry device for oesophageal motility
evaluations, slightly different procedures were used and are illustrated in
Supplementary Fig. 11. Briefly, a metre-long silicone tubing was first filled with
EGaIn, followed by cutting into ~6 cm segments. A knot was tied onto each
segment using a mechanical stretcher, each segment was connected with copper
wires at both ends and sealed using UV curing optical adhesive (Norland NOA 81).
The step was repeated multiple times and the resulting segments were placed and
adhered onto a ~45 cm long medical-grade silicone gel tape (AWD Medical),
with ~5 cm spacing between adjacent knots. A ~13-µm-thick low-density
polyethylene film was used to encapsulate the top surface, which completed the
fabrication process.
Customized setups for characterizing pressure sensitivity. A manual mechanical
testing stage (Mark-10, Model ES20) coupled with a force gauge (Mark-10, Model
M4-05) was used to apply precisely controlled compressive force onto QUILT,
which was then converted to pressure using contacting area. The channels were
connected to a source metre (Keithley-2450) for 2-point resistance measurement,
with the source direct current voltage set to 0.5 V (see Supplementary Fig. 1 for an
image of the entire testing setup).
Numerical simulations. All simulations were carried out using the commercial
FE package Abaqus 2017 (SIMULIA). The Abaqus/Explicit solver was employed
for the simulations. 3D FE models of the elastomeric tubing were constructed
using 8-node linear brick with reduced integration and hourglass control (Abaqus
element type C3D8R). The material behaviour of the elastomers was captured
using a nearly incompressible Neo-Hookean hyperelastic model (Poisson’s ratio
ν0 = 0.499 and density of 1,000 kg m−3) with directly imported uniaxial tension test
data. The Dynamic Explicit solver (DYNAMIC module in Abaqus) with a mass
scaling factor of 10,000 (to facilitate convergence) was used. To ensure quasi-static
conditions, we monitored the kinetic energy and introduced a small damping
factor. A simplified contact law (General Contact type interaction) was assigned to
the models, with a penalty friction coefficient of 0.3 for tangential behaviour and
hard contact for normal behaviour. Two sets of analyses were performed:
Elastic overhand knot formation. First, we created FE models of the bent tube
similar to the loop configuration shown in Supplementary Fig. 9 (left panel).
We kinematically tied the extremities of the tube to a pair of reference nodes
positioned at the centre of each extremity. To establish a knot configuration,
we pulled the extremities of the tube via application of displacement, ∆x/2, to
the reference points along the ±x direction. The identical reaction forces at the
reference points were recorded as a function of ∆x. The sequence of progressively
NATure BIomeDIcAl EnGIneerInG
deformed shape of the elastic tube to form a knot are shown in Supplementary
Fig. 9 and Video 1.
Normal compression of elastic knots. Following the elastic knot formation
simulations, the response of the knots under normal compression was evaluated
by subsequent compression of knots at different levels of ∆x across a range of
elastic moduli and wall thicknesses using a rigid plate. The plate was meshed
using 4-node 3D bilinear rigid quadrilateral (Abaqus element type R3D4) and was
initially positioned slightly above the knots, which were obtained from the previous
elastic knot formation simulations. We performed dynamic explicit analysis
by lowering the plate in the z direction until it compressed the knots down to
∆z/H0 = 0.6, where H0 is the initial height of the given knot (Fig. 2f). The reaction
force of the plate was recorded as a function of applied displacement in the normal
direction, ∆z (Supplementary Video 2).
Synthesis of the artificial food bolus. The protocol for making artificial food
boluses was adapted from ref. 31. Briefly, a 3 wt% solution of sodium alginate
(Sigma-Aldrich, CAS 9005-38-3, medium viscosity grade) was mixed with a 5 wt%
solution of gelatin (Sigma-Aldrich, CAS 9000-70-8, gel strength ~300 g Bloom,
type A) at a 7:3 wt:wt ratio and solidified at room temperature over 12 h. The
resulting gel was subsequently immersed in a 20 wt% CaCl2 solution with a volume
equivalent to that of the alginate solution for 24 h before use.
In vivo anaesthetized animal study. All studies were conducted in accordance
with the protocols approved by the MIT Committee on Animal Care.
Randomization of the animals was not performed. Female Yorkshire swine
(Cummings Veterinary School at Tufts University in Grafton, MA) weighing
40–80 kg were used for the in vivo evaluation of QUILT. The swine were placed on
a liquid diet 24 h before the study and fasted on the day of the procedure. On the
morning of the procedure, the swine were sedated using intramuscular injection of
either 5 mg kg–1 Telazol (tiletamine/zolazepam), 2 mg kg–1 xylazine and 0.04 mg kg–1
atropine, or 0.25 mg kg–1 midazolamand and 0.03 mg kg–1 dexmedetomidine. After
intubation, anaesthesia was maintained with isoflurane (2–3% in oxygen). The
anaesthetic level was adjusted to 1.8–2% in oxygen during the DIEP experiments.
While under anaesthesia, vital signs were monitored and recorded every 15 min
throughout the study. After the study, swine were woken up using intramuscular
injection of 0.1 mg kg−1 atipamezole and monitored closely during the recovery
process until reaching full recovery.
Reporting Summary. Further information on research design is available in the
Nature Research Reporting Summary linked to this article.
Data availability
The main data supporting the results in this study are available within the paper
and its Supplementary Information. Source data for the figures are available from
figshare with the identifiers https://doi.org/10.6084/m9.figshare.14544453 (Fig. 4)
and https://doi.org/10.6084/m9.figshare.17434874 (Fig. 5).
Code availability
The code used to generate the plots in Figs. 4c,d and 5g,j is available from the
corresponding author on reasonable request.
Received: 25 June 2021; Accepted: 19 January 2022;
Published: xx xx xxxx
References
1. Fox, M. R. et al. Clinical measurement of gastrointestinal motility and
function: who, when and which test? Nat. Rev. Gastroenterol. Hepatol. 15,
568–579 (2018).
2. Peery, A. F. et al. Burden and cost of gastrointestinal, liver, and pancreatic
diseases in the United States: update 2018. Gastroenterology 156, 254–272
(2019).
3. Anderson, P. et al. Survey of digestive health across Europe: final report. Part 2:
the economic impact and burden of digestive disorders. United European
Gastroenterol. J. 2, 544–546 (2014).
4. Moosavi, S., Rezaie, A., Pimentel, M. & Pichetshote, N. Atlas of
High-resolution Manometry, Impedance, and pH Monitoring (Springer, 2020).
5. Saltman, R., Bankauskaite, V. & Vrangbaek, K. Decentralization in Health
Care. Strategies and Outcomes (McGraw-Hill, 2007).
6. Aas, I. M. Organizational decentralization in radiology. J. Telemed. Telecare
12, 1–3 (2006).
7. Al-Alusi, M. A., Ding, E., McManus, D. D. & Lubitz, S. A. Wearing your heart
on your sleeve: the future of cardiac rhythm monitoring. Curr. Cardiol. Rep.
21, 158 (2019).
8. Isakadze, N. & Martin, S. S. How useful is the smartwatch ECG?
Trends Cardiovasc. Med. 30, 442–448 (2020).
9. Ascher, M. The logical-numerical system of Inca quipus. IEEE Ann. Hist.
Comput. 5, 268–278 (1983).
Nature Biomedical Engineering | www.nature.com/natbiomedeng
Articles
NATure BIomeDIcAl EnGIneerInG
10. Ascher, M. & Ascher, R. Mathematics of the Incas: Code of the Quipu
(Courier Corporation, 2013).
11. Dickey, M. D. et al. Eutectic gallium‐indium (EGaIn): a liquid metal alloy
for the formation of stable structures in microchannels at room temperature.
Adv. Funct. Mater. 18, 1097–1104 (2008).
12. Cooper, C. B. et al. Stretchable capacitive sensors of torsion, strain, and touch
using double helix liquid metal fibers. Adv. Funct. Mater. 27, 1605630 (2017).
13. Yeo, J. C., Yu, J., Koh, Z. M., Wang, Z. & Lim, C. T. Wearable tactile sensor
based on flexible microfluidics. Lab Chip 16, 3244–3250 (2016).
14. Gao, Y. et al. Wearable microfluidic diaphragm pressure sensor for health and
tactile touch monitoring. Adv. Mater. 29, 1701985 (2017).
15. Ma, Z. et al. Permeable superelastic liquid-metal fibre mat enables biocompatible
and monolithic stretchable electronics. Nat. Mater. 20, 859–868 (2021).
16. Holloway, R. H. Esophageal manometry. GI Motility online https://www.
nature.com/gimo/contents/pt1/full/gimo30.html (2006).
17. Liem, O. et al. Solid‐state vs water‐perfused catheters to measure colonic
high‐amplitude propagating contractions. Neurogastroenterol. Motil. 24,
345-e167 (2012).
18. Lu, Y. et al. Transformable liquid-metal nanomedicine. Nat. Commun. 6,
10066 (2015).
19. Yi, L. & Liu, J. Liquid metal biomaterials: a newly emerging area to tackle
modern biomedical challenges. Int. Mater. Rev. 62, 415–440 (2017).
20. Kim, K. et al. Highly sensitive and wearable liquid metal-based pressure
sensor for health monitoring applications: integration of a 3D‐printed
microbump array with the microchannel. Adv. Healthc. Mater. 8,
1900978 (2019).
21. Chen, L. et al. Icing performance of superhydrophobic silicone rubber
surfaces by laser texturing. Mater. Res. Express 6, 1250e2 (2020).
22. Araromi, O. A. et al. Ultra-sensitive and resilient compliant strain gauges for
soft machines. Nature 587, 219–224 (2020).
23. Han, M. et al. Catheter-integrated soft multilayer electronic arrays for
multiplexed sensing and actuation during cardiac surgery. Nat. Biomed. Eng.
4, 997–1009 (2020).
24. Bredenoord, A. J. et al. Chicago classification criteria of esophageal
motility disorders defined in high resolution esophageal pressure topography.
Neurogastroenterol. Motil. 24, 57–65 (2012).
25. Hansen, M. B. Small intestinal manometry. Physiol. Res. 51, 541–556 (2002).
26. Pfeifer, J. & Oliveira, L. in Constipation (eds Wexner, S. D. & Duthie, D. S.)
71–83 (Springer, 2006).
27. Shaker, A. et al. Multiple rapid swallow responses during esophageal
high-resolution manometry reflect esophageal body peristaltic reserve.
Am. J. Gastroenterol. 108, 1706–1712 (2013).
28. Leber, A. et al. Soft and stretchable liquid metal transmission lines as
distributed probes of multimodal deformations. Nat. Electron. 3,
316–326 (2020).
29. Yu, X. et al. Skin-integrated wireless haptic interfaces for virtual and
augmented reality. Nature 575, 473–479 (2019).
30. Bellinger, A. M. et al. Oral, ultra-long-lasting drug delivery: application
toward malaria elimination goals. Sci. Transl. Med. 8, 365ra157 (2016).
31. Hosotsubo, M., Magota, T., Egusa, M., Miyawaki, T. & Matsumoto, T.
Fabrication of artificial food bolus for evaluation of swallowing. PLoS ONE
11, e0168378 (2016).
32. Cheeney, G., Nguyen, M., Valestin, J. & Rao, S. S. C. Topographic and
manometric characterization of the recto‐anal inhibitory reflex.
Neurogastroenterol. Motil. 24, e147–e154 (2012).
Nature Biomedical Engineering | www.nature.com/natbiomedeng
33. Lee, T. H. & Bharucha, A. E. How to perform and interpret a high-resolution
anorectal manometry test. J. Neurogastroenterol. Motil. 22, 46–59 (2016).
34. Gourcerol, G. et al. Do endoflip assessments of anal sphincter distensibility
provide more information on patients with fecal incontinence than
high‐resolution anal manometry? Neurogastroenterol. Motil. 28,
399–409 (2016).
35. Zifan, A., Sun, C., Gourcerol, G., Leroi, A. M. & Mittal, R. K. Endoflip vs
high‐definition manometry in the assessment of fecal incontinence:
a data‐driven unsupervised comparison. Neurogastroenterol. Motil. 30,
e13462 (2018).
36. De Schepper, H. U. et al. Impact of spatial resolution on results of esophageal
high‐resolution manometry. Neurogastroenterol. Motil. 26, 922–928 (2014).
37. Sundaram, S. et al. Learning the signatures of the human grasp using a
scalable tactile glove. Nature 569, 698–702 (2019).
38. Xi, W. et al. Soft tubular microfluidics for 2D and 3D applications. Proc. Natl
Acad. Sci. USA 114, 10590–10595 (2017).
39. Luo, Y. et al. Learning human–environment interactions using conformal
tactile textiles. Nat. Electron. 4, 193–201 (2021).
Acknowledgements
We thank A. M. Hayward, K. Ishida and J. Jenkins for supervising and performing the
in vivo experiments using Yorkshire swine models; M. Kolle for insightful discussions on
knot mechanics; and H. Luan for helping with finite-element modelling. This work was
supported in part by the Karl van Tassel (1925) Career Development Professorship and
the Department of Mechanical Engineering, MIT.
Author contributions
K.N. and G.T. conceived and designed the study. K.N. fabricated and characterized the
QUILT. S.B., C.M.P. and A.M.J. performed the mechanical analysis and the finite-element
modelling. J.L.P.K., K.N., S.S.S. and V.R.F. performed the in vivo experiments on porcine
models. V.R.F. synthesized and optimized the artificial food bolus. W.W.C. advised on
the clinical aspects of this project and wrote a substantial portion of the manuscript.
All authors discussed and interpreted the results, and participated in writing and editing
the manuscript.
Competing interests
The authors report a patent application (U.S. Provisional Application No. 63/301,491)
describing the system described for manometric evaluation. Complete details of all
for-profit and not-for-profit relationships for G.T. are included in the Supplementary
Information. The other authors declare no competing interests.
Additional information
Supplementary information The online version contains supplementary material
available at https://doi.org/10.1038/s41551-022-00859-5.
Correspondence and requests for materials should be addressed to Giovanni Traverso.
Peer review information Nature Biomedical Engineering thanks Michael Dickey, Pankaj
Pasricha and the other, anonymous, reviewer(s) for their contribution to the peer review
of this work. Peer reviewer reports are available.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2022
Last updated by author(s): Dec 23, 2021
Reporting Summary
Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.
nature portfolio | reporting summary
Corresponding author(s): Giovanni Traverso
Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested
A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)
For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.
For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.
Software and code
Policy information about availability of computer code
Data collection
DC resistance of QUILT was measured using a single-channel source meter (Keithley, Model 2450), or a multi-channel resistance recorder (BK
Precision, Model DAS240), or a customized, multichannel, wireless circuit board designed and manufactured by Linkzill, Hangzhou, China.
Data analysis
All data visualization and analysis, except for Figs. 4c, 4d, 5g and 5j were performed using the commercial software Origin Pro Version 2020.
Figs. 4c, 4d, 5g and 5j were plotted using a custom Python script.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.
Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
March 2021
- Accession codes, unique identifiers, or web links for publicly available datasets
- A description of any restrictions on data availability
- For clinical datasets or third party data, please ensure that the statement adheres to our policy
The main data supporting the results in the study are available within the paper and its Supplementary Information. Source data for the figures are available from
figshare with the identifiers 10.6084/m9.figshare.14544453 (Fig. 4) and 10.6084/m9.figshare.17434874 (Fig. 5).
1
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences
Behavioural & social sciences
Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf
Life sciences study design
All studies must disclose on these points even when the disclosure is negative.
Sample size
Multiple female Yorkshire swines (40–80 kg) were used in the study. Each measurement was repeated at least three times on a single swine.
Data exclusions
No data were excluded.
Replication
Multiple devices were tested on different swine models, at separate times, to check for repeatability.
Randomization
The Yorkshire swines tested were randomly selected by animal-care staff 1 week before the experiments.
Blinding
No blinding was performed. The data and plots from the tests were analysed and checked by multiple authors.
nature portfolio | reporting summary
Field-specific reporting
Reporting for specific materials, systems and methods
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.
Materials & experimental systems
Methods
n/a Involved in the study
n/a Involved in the study
Antibodies
ChIP-seq
Eukaryotic cell lines
Flow cytometry
Palaeontology and archaeology
MRI-based neuroimaging
Animals and other organisms
Human research participants
Clinical data
Dual use research of concern
Animals and other organisms
Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research
Laboratory animals
Female Yorkshire swines, 40–80 kg.
Wild animals
The study did not involve wild animals.
Field-collected samples
The study did not involve samples collected from the field.
Ethics oversight
All studies were conducted in accordance with the protocols approved by the MIT committee on Animal Care.
Note that full information on the approval of the study protocol must also be provided in the manuscript.
March 2021
2