Operational principles of the wearable valving system

Figure 1a illustrates a representative pressure-regulated six-compartment valving system—with a sweat collection inlet at the center and an electrochemical sensing interface within each compartment—interfacing a wireless flexible circuit board to form a fully integrated wearable bioanalytical platform. To construct the valve, we specifically use a PNIPAM-based hydrogel (Fig. 1b and Supplementary Fig. 1a, synthesized from a N-isopropylacrylamide (NIPAM) monomer and N,N′-methylenebis(acrylamide), BIS crosslinker), which significantly shrinks/expands in response to local temperature increments/decrements, above/below its lower critical solution temperature (LCST)33. By embedding this hydrogel within a microfluidic channel, the volumetric thermal responsiveness of the hydrogel can be exploited to effectively permit/block fluid flow via activation/deactivation of the heater. Previous efforts have already demonstrated the utility of thermo-responsive hydrogel-based valving for controlling fluid flow within conventional lab-on-a-chip devices16,17,21,25. However, they required manually operated and bulky external instrumentation to actuate the hydrogels, preventing their translation into wearable platforms. Here, we devised a circuit-controlled micropatterned heater (on a flexible substrate) to actuate the hydrogels. In this way, we formed a miniaturized programmable valve, which can be extended into an addressable array, and subsequently, exploited to realize a valve-gated multicompartment bioanalytical platform amenable for wearable applications.

Fig. 1: A fully integrated wearable valving system (concept and operational principle).

a Illustration of a representative wearable bioanalytical platform, consisting of an integrated programmable microfluidic valving system interfacing a FPCB. b Illustration of PNIPAM hydrogel shrinkage/expansion in response to temperature change above/below its LCST (induced by activation/deactivation of the microheater). c A schematic operation example of the programmable microfluidic valving system, demonstrating biofluid routing, compartmentalization, and analysis in the selected compartment and sensor protection in the nonselected compartments. d Illustration of control commands (automated and manual) communication for scheduled and on-demand biomarker data acquisition with the aid of user interfaces preloaded on smart consumer electronics.

An example operation of our valving system is shown in Fig. 1c. In this example, the valve (downstream of the microfluidic channel) in compartment 1 is first activated (while others remain deactivated) to route and sample biofluid. Then, it is deactivated to block the flow, allowing for biofluid compartmentalization and analysis (using an electrochemical sensor positioned upstream of the channel). Accordingly, sample analysis can be performed—without the confounding influence of flow rate variability—by the sensor(s) in the addressed compartment, while the sensors in the other compartments remain protected.

This addressable compartmentalization capability can be exploited to take biomarker readings at scheduled/on-demand timepoints, thus enabling contextual biomarker analysis. In the presented wearable bioanalytical platform, valve activation and sensor output signal processing are delivered with the aid of a circuit board, which is equipped with a multichannel programmable current source and analog front-end circuits. Through bilateral Bluetooth communication with personal smart electronics (e.g., smartwatch), preloaded with a custom-designed user interface, biomarker data acquisition timepoints (pre-scheduled/on-demand) can be programmed (via automated/manual commands) and biomarker data can be displayed in real-time (Fig. 1d).

Wearable valve-gated microfluidic networks

For fluid valving, ideally, a binary off/on valve operation is desired, where fluid flow is completely blocked with no leakage in the off-state (when the valve is deactivated), and fluid flow is permitted in the on-state (when the valve is activated). In the context of our thermo-responsive PNIPAM-based hydrogel, off/on transition is achieved upon decreasing/increasing the temperature below/above the LCST. The thermo-responsive property of PNIPAM stems from the coexistence of hydrophilic amide and hydrophobic propyl groups within its polymer structure34. When the hydrogel’s temperature is lower than its LCST, the hydrogen-bonding interactions between the amide group and the water molecules are dominant. Therefore, the hydrogel becomes highly hydrated, leading to its structural expansion. Conversely, when the hydrogel’s temperature is higher than its LCST, the hydrogen-bonding interactions become weaker and the interactions between the hydrophobic propyl group and the water molecules are dominant. As a result, the water is released from the hydrogel structure, leading to hydrogel shrinkage.

For robust on-body valving, the temperature at which the hydrogel’s volumetric transition occurs should be sufficiently above the skin temperature (~35 °C), such that the heat transfer from the skin to the valve does not result in significant hydrogel shrinkage and subsequent fluid leakage. By incorporating an ionizable monomer (MAPTAC) in the hydrogel structure35, the volumetric transition temperature of about 45 °C is achieved. As shown in Fig. 2a, the modified PNIPAM-based hydrogel exhibits about 40% shrinkage from its original size (based on the 2D imaged area) after ramping up its temperature above the LCST point. Reversibly, the hydrogel can recover back to its original volume, simply by deactivating the microheater (Fig. 2b). The observed asymmetry in the hydrogel shrinkage and recovery rates can be attributed to the difference between the outward and inward diffusion rates of the surrounding buffer solution that is leaving and entering the hydrogel, respectively36. Moreover, the corresponding shrinkage and recovery rates are found to be proportional to the hydrogel size as demonstrated in Supplementary Fig. 1b, c. In order to maintain a fast valve responsive time, we minimized the size of the hydrogel embedded inside the channel (circle-shaped with radius <1 mm). By setting up a pressure-controlled fluid flow configuration (Fig. 2c), the flow rate within a hydrogel-embedded and microheater-coupled microfluidic channel was monitored. As shown in Fig. 2d, upon deactivation/activation of the microheater, the flow rate within the channel correspondingly dropped to zero/recovered to its default value, illustrating the reversible, consistent, and periodic switching capabilities of the formed valve. The slower transient characteristic of the embedded hydrogel as compared to that of the standalone hydrogel (Fig. 2b vs.  2d) can be attributed to the surface contact forces acting on the embedded hydrogel. Furthermore, our device temperature characterization results show that operationally the valve opens at temperatures 45 °C (Supplementary Fig. 2).

Fig. 2: Fabrication and characterization of valve-gated microfluidic networks.

a Standalone PNIPAM hydrogel shrinkage percentage vs. temperature profile (polynomial fitted curve illustrates the trend). Microscopic images of the standalone hydrogel at the annotated temperatures are shown as insets. b Reversible hydrogel (standalone) volume transition upon activation/deactivation of the microheater (polynomial fitted curve illustrates the trend). c A microfluidic valving characterization setup with a feedback-controlled pressure configuration. d The measured flow rate profile through a valve-gated microfluidic channel upon the periodic activation/deactivation of the valve. e Hydrogel layer fabrication procedure and layer-by-layer device integration scheme to realize microfluidic valving systems with different architectures. f Optical images of the representative fabricated hydrogel layers with different numbers/arrangements of hydrogels (a black substrate background is used to visualize the transparent hydrogel features). g Sequential optical images of progressive microfluidic routing and compartmentalization through illustrative serial, parallel, and tree microfluidic networks (constructed through integration with the same arrangement of hydrogels).

To fabricate the valve interface in an array format and within a tape-based flexible microfluidic module, a simple and high-throughput fabrication and integration scheme is devised (Fig. 2e, Supplementary Figs. 3 and 4). The scheme involves fabricating the hydrogel array, microfluidic network structure, and electrode array on separate layers, followed by the vertical alignment and assembly of the layers37. We particularly positioned the microheater electrode array layer as the top layer (i.e., away from the skin, in which case intermediary layers serve as insulators), to minimize the heat conduction to skin. In our scheme, the hydrogel array and microfluidic network features are defined by a laser cutter, which can be programmed at a software level to rapidly render various arrangements and dimensions. The hydrogel arrays can be developed by simultaneously injecting PNIPAM precursor solutions into the respective defined features, followed by a one-step ultraviolet crosslinking procedure, altogether rendering the development process low-cost and highly scalable (Fig. 2f). Our vertical integration approach also allows the same arrangement of hydrogel arrays to form various microfluidic routing and compartmentalization networks, simply by integrating microfluidic layers with different architectures. For example, as shown in Fig. 2g, an arrangement of six hydrogels are used to gate microfluidic networks with serial, parallel, and tree-like architectures (for visualization purposes, a blue dye is embedded within the channels and the hydrogels are externally/locally heated).

Active biofluid sampling from pressure-driven sources

In order to adapt the demonstrated valving operation to actively sample, route, and compartmentalize epidermally retrievable biofluids from pressure-driven sources, pressure release mechanisms are necessary. Specifically, in the context of sweat as the target biofluid, a pressure release mechanism is devised to avoid excess pressure build-up from the sweat glands. Without such mechanism in place, valve breakage would occur, due to the high pressure caused by the accumulated sweat (as high as ~500 mmHg with an air-tight sealed interface)38. The problem at hand can be formulated with the aid of an electrical circuit-hydraulic analogy (Fig. 3a), involving a current source (delivering current level IS) and a transistor switch. Here, the minimum turn-on voltage for the transistor switch is denoted as Vmin and its maximum tolerable voltage is denoted as Vmax (corresponding to its breakdown voltage). When directly connecting the transistor (in its off mode) to the current source, the built-up high voltage difference across the transistor (V) inevitably leads to transistor breakdown (>Vmax). Similarly, as shown in Fig. 3b (left), when directly interfacing the air-tight closed valve (microfluidic transistor switch) with actively secreting sweat glands (with secretion rate QS), the built-up high-pressure difference (P) across the valve inevitably leads to the valve breakage (P > Pmax, where Pmax denotes the valve’s maximum tolerable pressure).

Fig. 3: Elaboration, characterization, and demonstration of pressure-regulated valving.

a An electric-hydraulic analogy. (Vmin: minimum turn-on voltage of the transistor switch; Vmax: maximum tolerable voltage of the transistor switch; Pmin: minimum required pressure to open the valve; Pmax: maximum tolerable pressure of the hydraulic valve). b Design rationale of the pressure regulation mechanism (assisted by the electrical circuit analogy). c Optical image of the implemented pressure-regulated valve. Real-time pressure recording for the characterization of the d maximum tolerable pressure, e minimum required pressure, and f regulated pressure. Input flow rate was set to 5 μL min−1. g Characterized accumulated pressure across pressure-regulated microfluidic channels at different flow rates. Error bars, mean ± s.e (n = 3 measurements from different devices). h Sequential optical images of progressive microfluidic routing and compartmentalization through an illustrative pressure-regulated six-compartment valving system (performed ex situ).

In both scenarios, the addition of a secondary parallel electric/hydraulic conductive path allows for redirecting the electrical current/fluid flow as a relief mechanism (Fig. 3b, center). However, the electric/hydraulic resistance of these paths must be tuned to ensure that the voltage/pressure across the respective switches is maintained above Vmin/Pmin (where the switches are turned on). Electrically, this can be achieved by adding a parallel resistor (Re). Hydraulically, here, we use a membrane filter incorporated within an auxiliary microfluidic channel to render the desired hydraulic resistance, which effectively serves as a pressure regulation mechanism (Fig. 3b right, c).

To characterize Pmax and Pmin for our pressure-regulated valving interface, the same setup as that of Fig. 2c is used (with a programmed input flow rate of 5 μL min−1). As shown in Fig. 3d, the direct injection of fluid through the closed-valve microfluidic device (using a syringe pump) resulted in pressure built-up on the order of 300 mmHg (corresponding to Pmax), beyond which the device failed (due to leakage), as evident from the annotated drop in the measured pressure. Furthermore, the injection of fluid through the opened-valve microfluidic device resulted in ~10 mmHg pressure (corresponding to Pmin) across the device (Fig. 3e). Characterization of a microfluidic pathway, with the pressure regulation mechanism in place (Fig. 3f), illustrates the mechanism’s ability to effectively maintain the operational pressure (P) within the permissible pressure range (Pmin < P < Pmax, Supplementary Fig. 5) for different input flow rates (Fig. 3g). In addition, Fig. 3h shows that a fully formed valving system (consisting of heater-coupled hydrogel valves and pressure regulating embodiments) can be successfully used to route and compartmentalize fluid in an addressable and electronically programmable manner.

Flow rate-undistorted biomarker analysis

To demonstrate the utility of the devised active biofluid management system, biochemical sensing interfaces are developed and incorporated in the sensing chamber of the valve-gated compartments (upstream of each compartment channel as shown in Fig. 4a), following the previously reported mediator-free enzymatic sensor development methodology39. We specifically adapted the sensing interfaces to target glucose and lactate as examples of informative metabolites. As illustrated in Fig. 4a, the corresponding sensing interfaces comprised of: (1) an enzymatic layer (glucose oxidase or lactate oxidase) to catalyze the oxidation of target molecules and generate hydrogen peroxide (H2O2) as a detectable byproduct; (2) a permselective membrane (poly-m-phenylenediamine) to reject interfering electroactive species; and (3) an electroanalysis layer (platinum) to detect the generated H2O2. The response of the glucose and lactate sensors were validated within the respective analytes’ physiologically relevant concentration range in sweat40,41. As shown in Fig. 4b, c, for both sensors, linear relationships were observed between the measured current responses and target analytes’ concentration levels (R2 = 0.99, for both sensors).

Fig. 4: Demonstration of flow rate-undistorted biomarker analysis.

a Reaction schematic of the developed sensor (embedded within a valve-gated compartment). Current response to target analytes for b a glucose sensor and c a lactate sensor. Error bars, mean ± s.e (n = 3 measurements from different sensors). d Simulated analyte concentration (gradient) profiles for relatively low and high flow rate conditions (low flow rate: Q = 1 µL min−1, resulting in Pe = 12.4, high flow rate: Q = 10 µL min−1, resulting in Pe = 124, assuming a channel transverse width of 2 mm and analyte diffusivity constant of 6.7 × 10−6 cm2 s−1). The annotated dashed lines tangent to the normalized concentration curves indicate the local analyte concentration gradient for the respective case. e Simulated local analyte concentration gradient at various flow rates (the values are normalized to that obtained for the case of 1 µL min−1). The curve fitted line indicates that simulated data points present a (root {3} of {Q}) relationship. f Measured amperometric current response of a glucose sensor to 200 µM glucose solution introduced at various flow rates. The inset figure shows the corresponding measured real-time amperometric current response in the presence of progressively increasing flow rate (from 0 to 10 μL min−1). The curve fitted line indicates that simulated data points present a (root {3} of {Q}) relationship. g Comparison of the estimated glucose concentration of a 200 µM glucose solution introduced at 5 µL min−1 (no valve) and 0 µL min−1 (corresponding to valve-gated condition). Error bars, mean ± s.e (n = 3 measurements from different sensors).

The active biofluid flow control achieved by the valving system can be leveraged to address sensor-level challenges relevant to wearable biomarker sensing. In particular, here, the valving capability was utilized to decouple the confounding influence of flow rate variability on sensor response, an issue which is well-reported in the context of conventional lab-on-a-chip platforms29,30,31,32, but overlooked by previously reported wearable sensors.

In a generalizable continuous microfluidic electrochemical sensing setting, the response of the sensor is flow rate-dependent, because of the central role of advective flow in transporting analytes to the sensor42. In the case of electrochemical sensing, the sensor current response (I) is proportional to the flux of analyte molecules onto the sensor surface, which in turn is directly proportional to the local concentration gradient (M = (frac{{partial c}}{{partial z}})). In that regard, determining the local concentration gradient requires the consideration of various coupled phenomena, including advective and diffusive analyte transport to the sensor surface, and the reaction rate at the sensor surface. As described in the Supplementary Note, the coupled problem at hand can be simplified by assuming the sensor has a high surface reaction rate, and that advection is the dominant form of analyte transport (manifested as Peclet number  1, due to the relatively high sweat rate Q ~ 1–10 µL min−1 during active secretion). The theoretical analysis based on these assumptions leads to the (I propto M propto root {3} of {Q}) relationship.

This relationship was validated through finite element analysis (FEA) (COMSOL), where we simulated the analyte concentration profile at the sensor surface in response to various continuous flow rates (within the physiologically relevant range of sweat secretion rate). As shown in Fig. 4d, e, the concentration gradient on the sensor surface increased along with the flow rate in the microfluidic chamber, in which M was proportional to (root {3} of {Q}) (R2 = 0.98). Similarly, the measured amperometric current of a representative glucose sensor presented a cube-root relationship with Q (Fig. 4f, R2 = 0.96), which is in agreement with our theoretical analysis.

Practically, without accommodating for the influence of dynamically varying flow rate (during on-body measurements), if conventional calibration methods are followed (which are performed at zero flow rate, ex situ), inaccurate biomarker measurements will inevitably be obtained. This problem can be resolved by leveraging the devised valving mechanism, as it allows for performing analysis in a sample-and-hold manner. To elaborate, in a valve-gated sensing chamber, the valve can be opened, to allow for the introduction of the sample into the sensing chamber, and closed, to allow for sample compartmentalization and sensing at zero flow rate, thus effectively decoupling the confounding influence of flow rate variability. To demonstrate the influence of flow rate variability, the response of a representative glucose sensor to an introduced sample (containing 200 µM glucose) was monitored at 5 µL min−1 (no valve) and 0 µL min−1 (corresponding to valve-gated condition), and the corresponding estimated concentrations were derived by referring to the calibration curve (obtained at 0 µL min−1). As shown in Fig. 4g, the conventional setup overestimated the glucose concentration by 114%, whereas the valve-gated condition accurately estimated the glucose concentration.

Contextually relevant on-body biomarker analysis

In order to apply the devised pressure-regulated valving system for on-body biofluid management and biomarker analysis, we first evaluated the system’s operational stability during prolonged use and in the presence of motion artifacts. In that regard, we applied the flow rate characterization setup (same as that used in Fig. 2c) to quantitatively monitor the performance of a pressure-regulated valve in an ex situ setting. First, to assess its stability during a prolonged testing period, we sequentially activated and deactivated the valve at set timepoints over a period of 6 h. Figure 5a shows the flow rate, injected by the pressure-driven syringe pump, was successfully reduced to zero and back to its default value upon deactivation and activation of the valve, respectively. In addition, Supplementary Fig. 6 illustrates that for our context, hydrogel dehydration does not affect the intended valving functionality, as evident from the maintenance of a relatively constant pressure—across a valve-gated channel—over an extended amount of time (8 h). The minimal impact of hydrogel dehydration can be attributed to the small size of the outlets, minimizing the evaporation rate. Furthermore, to evaluate the stability of the valving system against motion artifacts, its performance was characterized under oscillatory motion (amplitude: ~3 m s−2 at 5 Hz43, generated by a vortex mixer). The measured flow rate profile, shown in Fig. 5b, indicates the successful opening and closing of the valve. Further ex situ and in situ characterization results, shown in Supplementary Figs. 7 and 8, provide insight into the robustness of the valving interface in the presence of mechanical deformation and unconstrained body motion. Altogether, these characterization results illustrate the preserved functionality of the valve over the test periods/conditions, informing the robustness of the valving operation for on-body application.

Fig. 5: Integration and characterization for contextually relevant on-body biomarker analysis.

a Ex situ characterization of the prolonged operation of the pressure-regulated valve (performed over 6 h). b Ex situ characterization of the high-fidelity operation of the pressure-regulated valve in the presence of vortical vibration. The vibrational acceleration profiles are presented in the top half, and the characterized flow rate profile is captured in the bottom half. c Optical image of a representative fully integrated programmable epidermal microfluidic valving system applied on the back of a subject with a zoomed-in view of the FPCB electronic components. The block diagram details the circuit-level valve actuation and signal-processing operations. d Illustration of the planned study for scheduled/on-demand sweat sampling during physical activity (cycling). e Optical images of intermittently sampled, routed, and compartmentalized sweat on-body (visualized with the aid of blue dyes, embedded within the compartments). Three valves were sequentially activated/deactivated at programmed timepoints during a physical exercise. The inset figures show the characterized electrical current through the respective valves’ microheaters (activated for 4 min). f Measured sweat glucose and lactate concentrations based on-body sensor readouts (green data points). The corresponding calibration curves (dashed lines) were constructed by linear fitting the measured sensor responses to three reference samples with known analyte concentrations (blue/red data points in the case of glucose/lactate sensors). Sweat glucose readouts were obtained by the valve-gated sensing compartments 1 and 2, before and after a scheduled beverage intake event, respectively. The sweat lactate readout was obtained by the valve-gated sensing compartment 3 upon on-demand user activation.

To realize a wearable valve-enabled bioanalytical platform with seamless control command and biomarker data communication capabilities, the sensor array-coupled valving system is interfaced with a custom-developed wireless FPCB (Supplementary Figs. 9 and 10). Structurally, the FPCB module is 100-μm thick, and its base material is polyimide, the Young’s modulus of which is on the same order as those of the materials used in the microfluidic module’s structure (Supplementary Table 1). In case a higher degree of mechanical flexibility is needed (e.g., when interfacing high curvature areas), other base materials with lower Young’s modulus can be used to construct the circuit board44. Figure 5c illustrates the operational block diagram of the FPCB, which is capable of rendering multichannel valve actuation and signal processing. Depending on the context at hand and the desired mode of analysis, an activation signal for the designated valve-gated sensing compartment is transmitted to the FPCB’s microcontroller unit (MCU). This activation signal can be generated through a scheduled timetable or on-demand (initiated by the user). Upon processing the received command, and with the aid of a multiplexer unit, the MCU selects the appropriate actuation channel to power the corresponding microheater by a current source, subsequently opening the desired valve. Subsequently, the harvested biofluid is routed to the selected compartment. Then, following MCU-generated instructions, the valve closes, and the sensor response is recorded and processed by an analog front-end (consisting of potentiostat and low-pass filter units) via the multiplexer-selected sensing channel. The signal processed by the analog front-end is then translated to digital at the MCU level, and wirelessly communicated to a user interface. The user interface can be used to display the acquired biomarker information in real-time and to store it in the user’s database.

The devised wearable valve-enabled bioanalytical platform was deployed for sweat sampling at scheduled and on-demand timepoints, to illustrate the platform’s capability for contextually relevant biomarker analysis applications (Fig. 5d). Accordingly, the platform was mounted on the back of a subject engaged in cycling (with the aid of a skin-adhesive layer, which provides adequate adhesion force to maintain the platform on the skin, Supplementary Fig. 11). Prior to on-body deployment, we activated the microheaters and monitored the electrical current passing through them to verify their operation. As can be seen from the on-body experiment, shown in Fig. 5e, the secreted sweat, at set scheduled/on-demand timepoints, was routed to and compartmentalized within the desired compartments (following a 4-min microheater activation time-window), while other compartments were protected. This time-stamped biofluid acquisition capability can be exploited to take contextual biomarker readings. As shown in Fig. 5f, the platform was programmed to take glucose readings before and after a scheduled beverage intake (Trutol, containing 50 g per 296 mL of dextrose) event, and sweat lactate level was measured on-demand as per the user’s command. Specifically, the biomarker readouts indicated the subject’s sweat glucose level was elevated after glucose intake, and the measured sweat lactate level was within an expected range (in agreement with previously reported studies)45,46. Supplementary Fig. 12 further validates our solution’s suitability for compartmentalization and sensor operations on-body. Our human subject experiment results, captured by the final device (prior to COVID-19 pandemic), were limited to the ones presented in this paper—no further human subject experiments were performed due to safety concerns around COVID-19. To provide physiologically meaningful interpretations of such sensor readouts, future large-scale studies should be conducted, aiming to contextualize the measured sweat biomarker concentrations in relation to relevant inter/intra-individual physiological variabilities (e.g., gender, muscle density, and body hydration).


Source link

Leave a Reply

Your email address will not be published. Required fields are marked *