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