A review of wearable biosensors for sweat analysis

Abstract

Recent advances in the skin-interfaced wearable sweat sensors allow a personalized daily diagnosis and prognosis of the diseases in a form of a non-invasive, portable, and continuous monitoring system. Especially, the soft microfluidic system provides robust quantitative analysis platforms that integrate sweat sampling, storing, and various sensing capabilities. This review systematically introduces the sweat collecting mechanism using soft microfluidic valves, including calculation of sweat storage and loss. In terms of sweat analysis, colorimetric (e.g. enzymatic, chemical, or their mixed reactions), electrochemical (e.g. voltammetric, potentiometric, amperometric, or conductometric), and multiplex measurements of sweat contents facilitate diagnosis of diseases via analysis of combined multiple data, such as vital signals (e.g. ECG, EMG, EEG, etc.) and information from the skin (e.g. temperature, GSR, etc.). The integration of wireless communication with the microfluidic systems enables point-of-care health monitoring for disease and specific physiological status.

Introduction

Human respiration primarily regulates temperature for thermal homeostasis in response to environmental factors and physiological conditions dynamically. The eccrine sweat is an unexplored biofluid and is of interest as it contains a rich blend of electrolytes, metabolites, micronutrients, hormones, proteins, and exogenous agents that can be analyzed in a non-invasive manner. Recent works establish applications of skin interfaced systems that include the capabilities to readout those sweat contents and physiology changes (i.e., sweat loss, ion concentrations, etc.) with various sensing mechanisms in the context of sports performance, military activities, and healthcare diagnostics. Here, we highlight the advanced wearable biosensor studies, and discuss the key technologies, concepts and their particular feature that provides insights toward future wearable biosensor research that aim practical, field-used devices.

Skin-interfaced, soft microfluidic system

Despite the non-invasive and easy-to-collect advantages, quantitative analysis of sweat has practical difficulties in direct sampling and detection of multiple biomarkers without evaporation. Conventional sweat researches and analysis relied on absorbent pads or fabric substrates that adhere to the skin to accumulate sweat and require professional handling to retrieve the deployed patch for precision analysis after sweat sampling. Koh et al. introduced a skin-interfaced soft microfluidic system that captures and stores sweat on the surface of the subjects' skin [1'. Figure 1a illustrates the soft, wireless microfluidic device to capture and store sweat to the channel.. The polydimethylsiloxane (PDMS) microfluidic channel layer with sweat harvesting areas (~ 10 mm²) and conformal contact of adhesive layer allow volumetric sweat harvesting (~ 1.2 to 12 µL/hour) in human studies [1]. The microfluidic system includes five separated channels that the one serpentine channel (width, 1.0 mm; height, 300 µm) is to measure how much sweat is excreted, and other channels are for quantitative colorimetric assays that occur in the 4 mm diameter and ~ 200 μm depth of cylindrical chamber in the middle of the channel. The unconventional concept of soft, microfluidic system design and subsequent human testing in real world activities with instant analysis of sweat in situ demonstrate the device enables to get rid of the possibilities of contamination and evaporation issues of the conventional research manner exploiting absorbent pads. Furthermore, the soft, microfluidic device provides a practical quantitative analysis platform enabling stoichiometry of the conversion reactions as the quantitative analysis and their calculations mostly rely on the basis of captured sweat volume. The subsequent researches have focused on overcoming of fundamental challenges, providing ideas to devising to response to the demands in the field where the broad range of environmental conditions (i.e. various humidity, temperature, aquatic sporting activities, etc.) [2, 3] may hinder the performance of designed sweat sensors.

Fig. 1.

Microfluidic devices and control system. a Optical image of soft, wearable microfluidic device. b Illustration of top viewed microfluidic, chrono-sampling device for sweat. c Detailed schematic illustration of a collection chamber, extraction chamber, sampling outlet, and three capillary bursting valves (CBVs). d SEM image of CBVs with indicated channel width and diverging angles that determine bursting pressure of the valve. e, optical image of in vitro chrono-sampling with dyed water in a device. Each reservoir requires 1 h to fill in and progresses to a counter-clockwise direction. f Optical image of a microfluidic, chrono-sampling device with super-absorbent polymer valves. The bended device shows its flexibility under mechanical distortions. g Before and e after appearance after sweat collection. The red box indicates swollen SAP valves within the reservoir after activation. i Digital images showing chrono-sampling of artificial sweat introduced at a rate of 10 µL min⁻¹. j Illustration of a representative wearable bioanalytical platform for integrated programmable microfluidic valving system. k A schematic illustration of thermo-responsive, active valving system controlled by commands (automated and manual) communication for protection of other compartments. l Optical image of assembled FPCB with integrated microfluidic actuation device placed on a human subject’s forearm. The device exploits alternating current electrothermal (ACET) flow phenomenon for on-demand, programmable, and precise microfluidic actuation. m Experimental setup for the ACET mixing of two fluids (blank and dyed solution). n Optical images corresponding to mixing indices of 0, 0.4, and 1 (Reproduced with permission, John Wiley and Sons, Springer Nature, and Royal Society of Chemistry)

Microfluidic network and valving system

The invention of skin interfaced soft microfluidic system derived a number of spin-off research projects such as various passive valving system development to classify sampled sweat in time order, wireless communication protocols for a real-time data process, and various capabilities of chemical, enzymatic, electrochemical, and immuno-reaction systems for the various sweat contents that can be integrated with the soft, microfluidics platform. Because of physiological regulation in the process of perspiration, the sweat contents change at excretion time. Especially, the differences in the contents of electrolytes (e.g., sodium ion) and metabolites (e.g. lactate) are critical to know the physiological condition of the subjects in the sweating scenario. The mechanical valve systems of Choi et al. (Fig. 1b–e) and Kim et al. (Fig. 1f–i) enabled passively programmed sweat flow logic in the microfluidics on the skin, thereby the effects of valving enabled time-sequenced sweat sampling [4, 5] and analysis of the discrete sweat samples that are separated into the well-defined volume of reservoirs at designed time points. Choi et al. [4] established a chrono-sampling system for sweat collection with capillary bursting valves (CBV; Fig. 1b). Figure 1c shows a schematic illustration and SEM images of the system, where three bursting valves (#1,2,3) with different bursting pressures (BP) can direct the sweat to flow into a network of microchannels in a sequential manner [4]. The BP of the elaborated microfluidic system (400 µm thickness; 200 µm channel width; 300 µm channel height) in a circular overall design (3 cm diameter) can be designed under operational pressure of ~ 1000 Pa for mechanical stability of the system. Figure 1d is the diverging angle (β) of the channel (13°, 90° and 120° for CBV of #1, #2, and #3, respectively) that determines the BP. When the sweat's flow reaches the inlet of the collection chamber, the fluid's sequence is determined by the CBVs' BPs. The channel width and diverging angle can adjust the theoretical BPs of valves. As a result of the CBV, the sweat's flow is guided to the collection chamber first, as the BP of CBV #1 is lower than that of valve #2. The extraction of the sweat can only be done through centrifugation (5000 rpm) of the CBV #3, which has a theoretical BP of 3035.7 Pa [4]. Figure 1e exhibits the chrono-sampling manner with the completed device.

Another microfluidic network, that has more complicating structure (Fig. 1f) with the combined effect of the different mechanism of a valving system through a double layer of microfluidics (Fig. 1g and h) enables time-sequential sweat sampling with instant quantitative assay eliminating any possibilities of mixing the sampled sweat with sweat flow in the microfluidics. Single time point has triplet reservoirs labeled clockwise T1-T5 [5] in the research. From T1 to T5, there are two valves used to induce the sweat to flow in a well-defined sequence: super absorbent polymer (SAP) valve and hydrophobic valve. The SAP valves, as shown in Fig. 1g, are activated when the reservoir is filled with sweat, then block the channel to the reservoir by swelling (Fig. 1h). Hydrophobic valves existing between the adjoining triplet reservoirs block the sweat flow to the next triplet reservoir before activating the SAP valve through the hydrophobic surface. Figure 1i shows how the sweat fills the complicating microfluidic network based on the combination of the passive valving system.

On the other hand, Lin et al. [6] devised a programmable valving system that the microfluidic network includes individually addressable microheater-controlled thermo-responsive hydrogel valves. Figure 1j exhibits the completed device that includes pressure-regulated valving system for sweat in six electrochemical sensing compartments which are connected with an inlet at the center of the compartments array. Use of PNIPAM-based hydrogel that is synthesized from a N-isopropylacrylamide (NIPAM) monomer and N,N′-methylenebis(acrylamide), BIS crosslinker, enables the valving at each compartment, specifically, shrinking and expanding in response to the temperature increment and decrements, respectively. The hydrogel valves at the compartments close the channel toward the ventilation duct, and the sweat flow cannot process unless it activates the first heater. Once the first valve is activated by opening the ventilation, the sweat flow is guided to compartment #1, where one pair of electrodes on the surface is modified for biomarker sensing (Fig. 1k) [6]. The sweat flow goes over to the valve and deactivates the first valve to touch the electrode pads that activate next hydrogel valve simultaneously. In this way, the series of compartments are activated to store and analyze the sweat samples.

Figure 1l shows another devising for an actuation system that is for engineering sweat flow profiles on the body [7]. The system exploits the alternating current electrothermal effects that induce a stirring-like fluid flow profile depending on the temperature. Figure 1m is a schematic illustration to show the design of the channel to generate laminar flow and a pair of electrode pads inducing the effect of mixing, as shown in Fig. 1n.

Storage of sweat in the skin-interfaced soft microfluidics

Flexible and stretchable features of soft-microfluidics enable to capture and store the excreted sweat. Given the nature of perspiration, sweat is evaporative and easily contaminated. Simple polymer film based or PDMS microfluidics provide basic platform in bench top researches in lab scale but practical applications in situ requires more robust and versatile platform for the sensing capabilities, various valving systems, and wireless communication protocols.

Reeder et al. introduced materials and designs to enhance the hydrophobicity by using poly(styrene–isoprene-styrene) (SIS) that can encapsulate and capture sweats more effectively in the aquatic or arid environments that may result contamination of the fluids or significant sweat evaporation [8]. The designs consisted of microchannels in circular serpentine geometries with 40 turns (depth: ~ 220 µm, 1 turn volume: 1.5 µL, total volume: 60 µL). Figure 2a shows the fabrication process of SIS microfluidics, and Fig. 2b shows the cross section of the completed device. Figure 2c demonstrates the fully operational sweat patch collecting the sweat in aquatic environments such as a scenario that a subject is in indoor swimming pool, which was achieved using small outlet geometries (r = 0.25 mm) and SIS polymer [8]. The research shows an experiment (n = 3) to compare the storage and waterloss of devices made with SIS and PDMS with open outlets at 37℃ for 4 h. The SIS devices demonstrated higher performance in sweat storage with less than 20% waterloss, while that of PDMS devices with comparable geometry had ~ 100% waterloss within 3 hours [8].

Fig. 2.

Integrated systems in microfluidic devices. a Schematic illustration of microfabrication process for waterproof, microfluidic poly(styrene-isopropene-styrene) (SIS) system. b Cross-sectional micrograph of a microfluidic channel showing contoured geometry of the top surface. c A subject wearing an epifluidic device for sweat collection in aquatic athletes. d Optical image of a microfluidic device with skeletal system to measure local sweat loss and sweat chloride concentration. e Schematic drawing of the device showing the dimensions of rigid PU microchannel structure, top silicone encapsulation layer (Ecoflex), and a white PDMS substrate. f Optical images of a PDMS device and a skeletal device during sweat sampling for evaporation test (Reproduced with permission, The American Association for the Advancement of Science and John Wiley and Sons)

Another advanced microfluidic platform uses polyurethane as a material of the microfluidic channel. Choi et al. [9] introduced a set of materials and mechanics design concept devicing a skeletal microfluidics as shown in Fig. 2d. Polyurethane resin defines microfluidic channel representing a rigid and sufficient stiffness as a skeletal structure, and the layout of the microfluidics which is machined by laser is packaged with low modulus, soft polymers such as silicone rubber and PDMS. Figure 2e shows the cross section of the skeletal microfluidics, and Fig. 2f is the results of sweat filling test with human subjects that exhibits the polyurethane microfluidic channel minimizes the evaporation of captured sweat for ~ 24 h comparing same testing with PDMS microfluidic channel.

Sweat loss measurements

Sweat excreted from sweat glands consists of various analytes of interest to determine the subject's physiological or health status, and wearable microfluidic devices are a prominent method to collect sweats for further analysis. Therefore, sweat excretion functions significantly as a source of analytes and a diagnostic parameter for hyperhidrosis and real-time fluid dehydration [10, 11]. Skin-interfaced microfluidics enable to capture the sweat and the collected sweat volume can be calculated into sweat excrete rate by dividing the number of estimated eccrine gland on the sweat and collecting time [1, 12, 13]. The colorimetric (Fig. 1a) tracking of sweat volume exhibits how much sweat is collected in the microfluidic channel, and exploiting of electrochemical measurements (Fig. 3a, b) enable digital tracking that the data can be calculated and seen with data processing protocols and smart devices. The simple design of digital volume dispensing system (DVDS) shown in Fig. 3a detects sweat rate in the chamber by tracking electrodes placed at the inlet and ceiling of the chamber: contacting of both electrodes indicates a droplet forming a capillary bridge between the two electrodes, and a strong wicking event breaks and resets the droplet. The time difference of the signals rendered from the droplet shorts can detect the sweat flow rate of the DVDS device (25 ~  9 × 10⁵ nL/min) [12]. Integrating of a pair of continuous tracking electrodes system in the skin-interfaced microfluidic channel along with coupling of NFC electronic module provides more precision measurement of real-time sweat rate as shown in Fig. 3b [14]. Figure 3c is an illustration to shows the cross section of the electrodes embedded microfluidic channel. Filling of sweat can be recognized as the sweat acts as resistance and measuring of ohmic impedance enable to calculate how much sweat is filled. In same manner with Koh et al. the sweat volume filled in the channel can be calculated into sweat rate [14]. Bandodkar et al. [15] presented flexible galvanic cells in each reservoir to incorporate the measurement of times to fill out the reservoir. For each galvanic cells that are in contact with sweats, the galvanic cells are activated and the time. Sweats, which function as electrolytes, activate the galvanic cells in each reservoir, and the time of the elevated voltage corresponds to the time taken to fill out the reservoir's volume. Figure 3d delineates a flowchart of the time-sequenced sweat collection during physical activity (red box) and logging of time stamps using the integrated NFC electronics module (blue box). The skin-interfaced soft microfluidic systems in the wearable sweat researches is demonstrated to be it is versatile and robust platform in situ, and recently Baker et al. [16] experimentally probed the advances of the skin-interfaced microfluidics showing that there is a significant correlation between microfluidic sweat patch and conventional absorbent patch using natural pressure of sweat secretion and microfluidic channel made of hydrophobic polymer (n = 43, r = 0.83 and P < 0.0001 for sweating rate; n = 43, r = 0.84 and P < 0.001 for sweat [Cl⁻]). In the experiment, personalized sweating rates in on-field/court sports training (n = 43 subjects) and controlled laboratory conditions (n = 45 subjects) measured by microfluidic patch and absorbent patch. The systematic analysis of the field-study and laboratory experiment indicates the sweating rate from the microfluidic patch was significantly higher than that of the absorbent patch (1.99 ± 1.22 versus 1.55 ± 0.68 mg/cm2 per minute, P < 0.0001) [16]. Figure 3e illustrates the automated optical analysis of microfluidic patch for sweat rate and chloride measurements. The microfluidic patch aims to detect chloride and sweating rate by colorimetric measurements. Figure 3f shows detailed boundaries for each analysis and critical features. Another important parameter for predicting whole-body sweating rate is a regional measurement of sweating rate, which could be derived from microfluidic sweat patches (r² = 0.74, n = 312) [16]. The report spotlights systematic studies to compare the wearable microfluidic platform with conventional techniques for sweat testing. Furthermore, the study presented algorithms to predict whole-body sweating rate and whole-body sweat [Cl⁻] in a large clinical study (312 athletes). Figure 3g indicates the scatterplots of measured versus predicted whole-body sweating rates (n = 312) based on seven factors, including microfluidic regional sweating rate, body mass, sex, air temperature, type of sport, exercise duration, and energy expenditure. Figure 3h demonstrates whole-body sweating results based on six factors including all above except energy expenditure (n = 312). The studies demonstrated reliable measurement results for the microfluidic devices compared to the reference techniques with a 9% difference in sweating rate and 12 ~ 13% difference in sweat’s chloride concentration [16]. The technical advancement of the epidermal, soft-microfluidic sweat sensors and their practicality proven by the report further imply a more personalized and accurate point-of-care system used in broader applications.

Fig. 3.

Sweat rate measurements and wireless data transfer. a Optical image of two different devices for digital volume dispensing system (DVDS) coupled to a wearable sweat collector. b Schematic illustrations of a skeletal microfluidic channel for tracking sweat rate, sweat loss, and an immunoassay for cortisol. c Cross-sectional view of the main channel with channel dimensions and integrated electrodes. d Flowchart delineating device usage for time-sequenced sweat collection during physical activity (red box) and logging of time stamps using the integrated NFC electronics module (blue box). e Automated optical analysis of microfluidic patch for sweat rate and chloride measurements. Photographing the patch on user’s arm with the smartphone application provides analyzed results of the device. f Automated-detection of patch boundaries and critical features. Colored outlines denote boundaries of detected patch features. g Scatterplots of measured versus predicted whole-body sweating rates (n = 312 subjects) based on seven factors including microfluidic regional sweating rate, body mass, sex, air temperature, type of sport, exercise duration, and energy expenditure. h Whole-body sweating rate results with six factors including all above except energy expenditure (n = 312 subjects). i Image of a reversible coupling process of a NFC electronic system to the microfluidic platform. j Schematic diagram of the system and illustration of the electrodes layouts for the main (purple) and reference (orange) microchannel structures. k Digital image of results on the screen of a smartphone transferred by the integrated NFC interface. l Photographs of a subject wearing a laser-engraved wearable sensor patch at different body parts. m Schematic illustration of vector-mode laser cutting for microfluidic fabrication. n Image of the fabricated microfluidic device made up of polyimide, polyethylene terephthalate (PET), and graphene-based layers (Reproduced with permission, Royal Society of Chemistry, National Academy of Sciences, John Wiley and Sons, The American Association for the Advancement of Science, and Springer Nature)

Wireless communication for data process

Integrating a near field communication (NFC) into a microfluidic platform provides ultrathin construction, ultralow modulus, and battery-free configurations that attributes to the conformal, skin-like sweat sensors for intermittent real-time monitoring of sweat or physiology of the body [17–19]. Kim et al. [20] proposed a NFC protocol based (Fig. 3i), electrochemical readout of sweat rate, sweat loss, and electrolyte concentration in the soft microfluidic system with embedded ultra-thin electrodes system. Two separated channels serve as main tracking channel and reference channel. Photolithography technology enabled preparation of ultra-thin flexible Au-Cu electrodes that are laminated on PDMS substrate by bonding of SiO₂ on the bottom of the electrodes, and the exposed 5 pairs of pads in the assembled microfluidics track how much sweat filled in tracking channel by calculating of the sweat conductance from reference channel. Figure 3j shows schematic equivalent circuit design of the NFC electronics [20]. NFC technology enables converting of the analogue data from the electrodes embedded microfluidics to the digital data. Smart devices power the NFC system by transmitting radio frequency (RF) at 13.35 MHz and the NFC microcontroller generates an AC voltage (V_TMS = 1.5 V) at ~ 5 kHz to measure ohmic impedance of sweat in reference and main electrodes (Fig. 3j and k) [20].

In a way, some biomedical applications would require continuous profiling of physical and vital signals and change of biomarker levels in terms of data accumulation, and NFC protocol may have limitation in the applications such as internet of things (IoT) and point of care (PoC). Wireless, long-ranged and multiply connected features of Bluetooth technology provides active, multiplex and continuous monitoring capability that is required in IoT and PoC scenarios [21]. Yang et al. [22] used the Bluetooth-enabled mobile hand set that is combining of sensing capabilities for uric acid and tyrosine and data processing (Fig. 3l). Laser engraved microfluidics enabled multiple measurements of temperature, sweat rate, vital signals, and biomarkers (Fig. 3m and n) and these data are converted and transmitted through the Bluetooth protocol displaying data on the smart device [22]. Lee et al. [23] showed advanced and practical version of the miniaturized Bluetooth based wearable sensor to readout the acoustic vital signals with programming based on the algorithms that is consisted of specific thresholds and filters to figure out the vibrating signals.

Measurement technologies

The rich mixture of metabolites (e.g. ammonia, lactate, urea) [24–27], micronutrients, hormones (e.g. cortisol) [28, 29], proteins (e.g. cytokines) [30–32], exogenous agents (e.g. alcohol, drugs) [33, 34], etc., dissolved in eccrine sweat, is of interest as a potential non-invasive interstitial biofluid for monitoring of human physiology and biochemistry. Quantitative measuring of the contents along with physical and vital information has been a core of the efforts in wearable sweat sensor development, and here, we classified the mechanisms of quantitative measurements as colorimetric, electrochemical, immunoreactions and their hybrid manner to readout the various sweat contents, physical sweat information and signals from skin.

Colorimetric and fluorescent assay

Colorimetric methods, including enzymatic, chemical, or their mixed reactions, exploit the yield of conversion reaction in the quantities of target substance and final product that determine the intensity of color or fluorescence developments [1, 15, 20, 35, 36]. These mechanisms are relatively simple and economic as they require only few microliters of reagents for the sweat sample in the soft-microfluidics. The resulted colors can be analyzed with camera capturing on the smart devices extracting the color indexes (i.e. RGB index) followed by color normalization or filtering [5, 35, 37]. Combining of colorimetric or fluorescent assays and chrono-sampling microfluidics provides time-dependent information of the sweat contents [5, 35, 38]. Figure 4a–c shows typical colorimetric wearable sweat sensor that enzymatic color development reaction system is integrated in soft microfluidics for ammonia and ethanol measurement [39]. Ammonia (or ammonium) and ethanol levels in plasma are useful information to recognize DUI status and physiology researches in esophageal varices, cirrhosis and encephalopathy [39–43], and can be reduced with oxidization enzymes to produce proton and electron which reduces OxiRed, a probe chemistry with horse radish peroxidase (HRP). Figure 4b, c show the color developments and its kinetics depending on the target substances (i.e. ethanol) as the results of the enzymatic reaction [44].

Fig. 4.

Measurement technologies and collected data. a Schematic drawing of an microfluidic systems with integrated enzymatic assays for measurement of ammonia and ethanol in sweat. b Colorimetric analysis of the ethanol assay in varying concentrations per time. c Colorimetric analysis of ethanol assay as a function of time. d Monitoring pH of sweat and vaginal fluid from a subject exercising on a stationary bicycle. e Sweat pH values obtained from the subject over different phases of exercise intensity. f On-body nicotine monitoring via exercise-induced sweat. g Collected data of sweat nicotine concentrations for a subject cycling on an ergometer with smoking. h Schematic representation of glucohol (glucose + alcohol) biosensor with electrode layers for iontophoretic operation. i Glucohol biosensor sensing performance with sequenced eating and drinking. Meal and first alcohol intake were followed by second alcohol intake for stepwise observations. (i) Corresponding blood glucose concentrations and breath alcohol concentrations after each step. (ii) Amperometric measurements in phosphate buffer solution (pH = 7.4) for the duration of 60 s using a step potential of -0.2 V (vs Ag/AgCl). j Photograph of a subject exercising while placing a battery-free, skin-interfaced microfluidic/electronic system. k Optical image of complete system captured after a bout of cycling by subject #1. Real-time wirelessly acquired sweat concentration levels for l Lactate and m Glucose, respectively (Reproduced with permission, Royal Society of Chemistry, Elsevier, American Chemical Society, John Wiley and Sons, and The American Association for the Advancement of Science)

Electrochemical assay

Electrochemical sensing of the sweat aims real-time or continuous monitoring of dynamic changes of sweat contents and physical sweat information (Fig. 4d–g). The electrochemical mechanisms (i.e. voltametric, potentiometric, amperometric, conductometric, etc.) enables high sensitivity, rapid response assays of the sweat contents. Measurement of impedance along with phase and reactance enables to measure sweat conductance and basic properties such as pH [20]. Kim et al. [44] introduced ohmic impedance measurement to readout sweat conductance from the captured sweat sample, and Pal et al. developed a wearable pH sensor. Figure 4d, e shows the demonstration of the device with human subject, and the Ag/AgCl and carbon electrodes measured pH showing impedance measurement at 1 Hz–1 MHz. The electrodes are integrated in the waterproof electronic decals. Sweat conductance and pH can be utilized to predict dehydration and diagnosis of bacterial vaginosis, respectively. In most cases, electrochemistry sensing is affected by the volume of measuring sample as its sensitive response reflects overall changes in the measurement. Combination of electrochemical mechanisms with soft-microfluidics shows reliable and reproducible assay results as the compartments and reservoirs in the soft-microfluidics network define very well controlled volume and offer precision background of measurement that minimize or eliminate the errors in situ.

Modification of electrodes with proteins, enzymes etc. enhances the sensitivity and selectivity of the sensing system. Tai et al. [45] designed the electrodes system immobilizing cytochrome P4502B6 which is nicotine-oxidizing enzyme on the 11-mercaptoundecanoic acid modified Au working electrode. The enzymatic oxidization reaction enables amperometric measurement (Fig. 4f, g). Exploiting of the feature (i.e. specificity) of enzymatic reaction enables to sense various trace substances in sweat. The alcohol oxidase, a same enzyme that Kim et al. used in their colorimetric reaction system (Fig. 4a–c), enables quantitative electrochemical sensing of alcohol specifically on the skin with the tattoo type electrodes design. Figure 4h shows the electrodes design that has three electrodes system, the surface of working electrode is modified with oxidizing enzymes for alcohol and glucose sensing, and the device has capability of localized stimulation for sweating and drug delivery based on the iontophoresis mechanism [46]. Figure 4i shows the sequences of human test events, and the representative results. Furthermore, amperometric sensors enable to detect low concentration of substances in sweat. Hybrid form of electrochemical-colorimetric wearable sensor exploited oxidizing enzymes for lactate and glucose sensing in sweat. Figure 4j, k show the device deployment with RF transmitting antenna to operate the battery free NFC electronics [47]. The enzymes immobilized electrodes selectively readout the amperometric data from the captured sweat, and Fig. 4l, m show the results of human tests. Sweat glucose quantity reflects blood glucose in diluted form, and the glucose sensor responded at trace amount of glucose (170 μM of glucose). Cortisol, a steroid hormone, is known to be a stress hormone that receives great attention, and the level of free form of cortisol in sweat is well correlated with that of the blood [48]. But, the amount of cortisol is relatively low and needs specific strategies for quantitative analysis. Recent research effort exploiting enzyme-linked immunosorbent assay (ELISA) enabled precision analysis of sweat cortisol amplifying the small amount of cortisol associating with anti-cortisol antibody conjugated gold nanoparticle. However, the development in the colorimetric assay has limitations for application in real-time monitoring of sweat cortisol in situ. Figure 5a–e show cortisol sensing based on the concept of real-time monitoring [49, 50]. The lab-on-a patch for cortisol sensing shown in Fig. 5a is designed based on 3D structured Au electrode that is embedded in microfluidic platform which can capture sweat from forehead (Fig. 5b), in turn, the pre-loaded immunoassay reagent can be mixed by operating in one-touch [49]. The data collected from three subject in the morning and evening, respectively exhibit consistent trend comparing to the ELISA data (Fig. 5c). Another sensor that exploits carbon mediator to immobilize anti-cortisol antibody. In the mechanism, horse radish peroxidase conjugated with cortisol serves as a competitor of label free cortisol in sweat and enables oxidization of hydrogen peroxide to generate current. The sensing system shows high sensitivity of cortisol detection down to ~ 0.5 ng/mL concentration [50].

Fig. 5.

Measurement technologies for cortisol and collected data. a Schematic illustration of a stretchable microfluidic device for cortisol sensing. The microfluidic device contains four chambers, the sweat inlet (I), sensing (S), disposal (D), and reagent (R) chambers, and three valves in the channels. b LOP placed on a subject for cortisol monitoring during cycling. c Cortisol concentration results from an ELISA kit (n = 3) and LOP (forehead placed) assay for 6 samples, 2 each from subjects 1, 2, and 3. The error bars are indicated on the analysis from the ELISA kit (n = 3) and relative difference between the measured concentrations from the ELISA and LOP assays $\left(\left|C_{\mathit{ELISA}}-C_{\mathit{Patch}}\right|/C_{\mathit{ELISA}}\right)$ is marked above the bar graph. d Design of flexible microfluidic three-working-electrode (3WE) sensor array for cortisol detection and printed circuit board with graphene sensor patch. e Results of the measured cortisol from three physically untrained subjects (B1-B3) and one trained subject (B4) in a constant-load cycling exercise (Reproduced with permission, Elsevier)

Multiplexed measurements

Personal healthcare monitoring, PoC, and diagnosis of disease and specific physiological status (i.e. DUI, enforced monitoring of criminals etc.) requires multiplex measurement of sweat contents, vital signals (e.g. ECG, EMG, EEG etc.), information from skin (i.e. temperature, GSR etc.), combination of the multiple data, and analysis. Current wearable sweat sensors focus on simple measurement of discrete substances in sweat based on various sensing mechanisms. Advanced technologies of wireless data processing and transmitting along with the developing sensing mechanisms could be combined and hybridized with the detection mechanisms to address the complicating applications in healthcare monitoring and diagnosis. Recent convergence researches based on advanced technologies show the multiplex measurement system development to address more complicating physiologic behaviors.

Sweat is a mixture of various electrolytes, and it can serve as an electrolyte for Mg–Ag/AgCl battery cell that is called sweat activated cell (SAC; Fig. 6a) [51]. A microfluidic platform can collect sweat, and the filling of sweat electrolyte in the microfluidics facilitates the electrochemical reaction. Their discharge data shows high specific energy density of ~ 580 Wh/kg at ~ 1.6 V operational voltage, and the wearable sweat sensor includes a SAC powered NFC system and ECG electrodes to measure heart rate (Fig. 6b, c) along with the colorimetric measurement of chloride ion and pH that may represent the electrolyte condition [51]. These capabilities and system organization enable to monitor biophysical information and can analyze in real-time if the subject feels fatigue or if the subject requires re-hydration in the exercise and sweating situation. Kim et al. [14] suggested another multiplex measurement and combination of collected data to predict dehydration of the human subject. The three channels of integrated NFC system readout sweat conductivity, collected sweat amount and skin conductance (Fig. 6d) with the galvanic skin response (GSR) electrode pair and the tracking electrodes system embedded in their devised serpentine polyurethane channel (Fig. 3b, c). Accordingly, the embedded electrodes measure the length of sweat-filled channel by calculating the ohmic impedance with the reference sweat conductance. Here, the channel length can be converted into collected or excreted sweat volume (i.e. sweat excretion rate or sweat rate), and the derived sweat rate and comparison with GSR measurement provide important insights into sweat gland activity such as ion resorption. Prolonged sweating in the exercise of human induces high sweat rate that can increase ion excretion and resorption in terms of physiological regulation for homeostasis. Figure 6e shows the typical GSR data from human subject that the GSR data at initial phase of sweating is relatively plane at low sweat rate as there would be selectively resorption of sodium ion and simultaneous ion excretion, and the later phase of sweating phase exhibits the GSR data raised at high sweat rate which can be seen as the starting of dehydration [14]. Continuous measurement of key sweat contents with the vital signals enables to diagnose chronic diseases, and multiplex measurement platform. Figure 6f is 3D illustration to show the whole device that microfluidics part, in which measuring electrodes are placed, is combined with Bluetooth data processing part [22]. The laser engraved microfluidics consists of discrete compartments that has graphene based chemical sensor for measurements of uric acid and tyrosine and physical sensors for temperature, sweat rate and heart rate (Fig. 6g). The combination of multiplex data collection in the human study controlling diet exhibits the system can predict cardiovascular disease, type 2 diabetes, rental disease, and gout.

Fig. 6.

Multiplex measurement of sweat sensors. a Photograph of a sweat-activated cell (SAC)-powered hybrid microfluidic-microelectronic device for simultaneous monitoring of heart rate (HR), sweat chloride (Cl⁻), and sweat pH. b, c Real-time HR captured by a device on 2 human subjects during cycling. Green region indicates postcycling cool-down. d Optical image of microfluidic systems with integrated immunoassays, fluorometric sensors for colorimetric and impedance measurements for multiplex measurement capabilities. e Correlation between sweat rate and change in galvanic skin response (ΔGSR) after skin temperature stabilizes and sweating begins (placed on forearm, 18 to 20 ℃ temperature, and 15 to 30% humidity). f Setup of the flow test for wireless monitoring of the laser-engraved wearable sensor. A syringe pump is used to inject analyte solutions through an inlet. g Real-time, continuous in situ measurement of respiration rate (RR; measured in breaths per minute (b.p.m.)), temperature and sweat uric acid (UA) and Tyr levels from the neck of a healthy subject during a constant-load stationary cycling exercise (5-min warm-up, 30-min constant-load cycling, and 5-min cool-down). Insets in the RR region indicate 30-s strain sensor signals recorded at 5-min intervals (Reproduced with permission, National Academy of Sciences)

Summary and outlook

In summary, recent researches and developments for wearable sensors focus on devising advanced forms of complete devices, including various practical sensing mechanisms to read sweat contents and vital signals. Colorimetric and electrochemical mechanisms with well controlled enzymatic, chemical reactions can be involved in the design of wearable sensors upon the demands in the field. Especially, skin-interfaced soft microfluidics provide robust quantitative analysis platforms that integrate sweat sampling, storing and its various sensing capabilities. Baker et al. demonstrated the feasibility of the skin-interfaced microfluidics showing significant correlations of measuring sweat rate and chloride ion concentration with 312 athletes in the real-world application. Moreover, integrating of ultrathin electrodes, wireless data processing electronics, etc. in the microfluidic platforms enables real-time monitoring that provides continuous data collecting for long-term study of the subjects. In this way, connection of big data, and personal healthcare data accumulation could be followed for further prediction and complex diagnosis. The multiplex measurements and combination analysis of the collected data with programming based algorithm design, as shown in recent studies, demonstrated their practicality in understanding more complicating human physical status, disease, etc. in various environments of human activities and scenarios of healthcare monitoring in situ, and are expected to be employed in future researches. These efforts enable precise monitoring of diseases and physical conditions of the subjects that should be constantly monitored in daily life, such as fatigue control, neuropsychiatric monitoring, encephalopathy, cardiac disease, DUI, drugs monitoring and protection of elderly senior drivers in a way of simple operation.

Declarations

Conflict of interest

The authors declare that they have no conflict of interest.