Wearable biosensors for human fatigue diagnosis: A review

Abstract Fatigue causes deleterious effects to physical and mental health of human being and may cause loss of lives. Therefore, the adverse effects of fatigue on individuals and the society are massive. With the ever‐increasing frequency of overtraining among modern military and sports personnel, timely, portable and accurate fatigue diagnosis is essential to avoid fatigue‐induced accidents. However, traditional detection methods require complex sample preparation and blood sampling processes, which cannot meet the timeliness and portability of fatigue diagnosis. With the development of flexible materials and biosensing technology, wearable biosensors have attracted increased attention to the researchers. Wearable biosensors collect biomarkers from noninvasive biofluids, such as sweat, saliva, and tears, followed by biosensing with the help of biosensing modules continuously and quantitatively. The detection signal can then be transmitted through wireless communication modules that constitute a method for real‐time understanding of abnormality. Recent developments of wearable biosensors are focused on miniaturized wearable electrochemistry and optical biosensors for metabolites detection, of which, few have exhibited satisfactory results in medical diagnosis. However, detection performance limits the wide‐range applicability of wearable fatigue diagnosis. In this article, the application of wearable biosensors in fatigue diagnosis has been discussed. In fact, exploration of the composition of different biofluids and their potential toward fatigue diagnosis have been discussed here for the very first time. Moreover, discussions regarding the current bottlenecks in wearable fatigue biosensors and the latest advancements in biochemical reaction and data communication modules have been incorporated herein. Finally, the main challenges and opportunities were discussed for wearable fatigue diagnosis in the future.

However, these methods cannot continuously monitor the changes of fatigue biomarkers during exercise, which cannot provide timely feedback on the abnormal state of human. In conclusion, efficient, continuous, and timely fatigue analysis is still limited. 19 With the development of artificial intelligence (AI) and advanced material science, the concept of the Internet of Everything has emerged gradually. Methods, which can eliminate the limitations of laboratory environment for real-time detection of biomarkers are explored, and wearable sensing technology, composed of medical monitoring and wireless signal transmission, has emerged as the times require. 20,21 In this context, the advanced technology companies, such as Apple and Huawei, have already introduced integrated heart rate, pressure, and blood oxygen saturation modules into their wearable Wearable biosensors are biosensing devices, which can detect biomarkers accurately, instantly, and portably. 22 Early wearable biosensors predominantly focused on detecting physiological signals, such as blood pressure, steps, heart rate, and sweating rate, which could not reflect the abnormal human state changes accurately and comprehensively. 23,24 With the development of insight of health care and disease prevention, the quantitative detection of biochemical biomarkers in biofluids has shown more possibilities. 25 In recent years, with the development of flexible material technology and wireless communication technology, modules of wearable biosensors, including sample collection, biochemical reaction, and signal readout, are integrated in wearable biosensors, which noninvasively analyze the wearer's biofluids and output the signals in the form of electrical and optical. 26 Then, the wearable biosensors feeds back the health data and physiological changes through wireless communication technologies, such as WiFi, Bluetooth modules, and near-field communication (NFC). 27 Although wearable biosensors were initially widely applied in the field of medical monitoring, but with the increasing emphasis on exercise fatigue, wearable biosensors are expected to play the vital role in military operations and sports monitoring.
As an auxiliary means of medical monitoring, wearable biosensors mainly focus on the analysis of medical diseases, such as diabetes 28 and Parkinson 29 for improving the efficiency of medical diagnoses and health management, and the application of wearable biosensors has been summarized and reported in detail, which proved that the wearable biosensors were suitable for medicalcare. 30 However, advantages, such as easy portability, sweat collection, and a wide range of applications of wearable biosensors, are not fully exploited. The application of wearable biosensors in fatigue diagnosis may provide the possibility to exert their advantages fully. In the traditional diagnosis methods, the analysis of fatigue biomarkers in serum samples still occupies a significant position in fatigue diagnosis that require painful blood sampling and the larger time. 31 Whereas wearable biosensors can fit close to body surface and therefore can stably detect fatigue in both motion and resting states with both flexibility and portability in the most instances. However, few reports focused on customized wearable biosensors for fatigue diagnosis because of the limitations in technology and lack of specific diagnosis. In the following discussion, we will discuss about applications and prospects of wearable biosensors in fatigue analysis by introducing the development in them. First, biomarkers in invasive and noninvasive biofluids were introduced, and the advantages and disadvantages of wearable biosensors in different biofluids have been explored.
Thereafter, the improvement of biochemical response elements in multichannel detection, improved sensitivity, biometric recognition, and groundbreaking reports with significant impact have been reviewed. After that, the development of data modules based on different signal readout in recent years has been expounded, and their potential in continuous fatigue diagnosis has been discussed.
Finally, we have outlined the existing bottlenecks and challenges of wearable biosensors in the field of fatigue diagnosis and emphasize their development prospects.

| BIOFLUIDS FOR WEARABLE BIOSENSORS AND THEIR PROSPECTS IN FATIGUE DIAGNOSIS
Biological metabolism, including inorganic salts, hormones, enzymes, and biological macromolecules vary with the state of physical function. 32,33 Some biomarkers, which change significantly during exercise fatigue, are considered fatigue biomarkers. Biofluids are easy to obtain with few preprocessing steps and the presence of fatigue biomarkers are prerequisites for wearable fatigue diagnosis. Blood is a transit station for physiological metabolism and contains a rich variety of biomarkers. 34 Compared to other biofluids, there are a wide range of biomarkers in serum with higher concentration that is an ideal sample for fatigue diagnosis. Some biosensors for the detection of fatigue biomarkers in blood were summarized in Table 1, and their sensitivity and detection performance were feasible within its applicable range.
However, serum analysis is accompanied by a complicated sample collection and purification process, which requires professional technical operators. The painful blood specimen collection not only causes harm to the subjects but also causes infection and safety issues because of the pollution caused by surroundings and secreted sweat, 45 which is not suitable for users with exercise fatigue. In addition, although some colorimetric-based biosensing methods have been developed, strict experimental conditions are still required. To achieve user-friendly fatigue diagnosis, fatigue analysis of noninvasive biofluids is essential.
Noninvasive biofluids include tears, saliva, and sweat, where some biomarkers reflecting human fatigue are also secreted. For instance, tears are secreted by lacrimal glands and contain biological components, such as proteins, lipids, glucose, and electrolytes. 46 In particular, the contents of biological components in tears correlate with blood and is therefore regarded as excellent biofluids for noninvasive diagnosis of local or systemic disease. 47,48 Again, saliva is secreted by the salivary glands, and the antibodies present in saliva have proven to be an effective way to diagnose HIV 49 and intestinal infections. 50 Sweat is secreted and excreted to the skin surface by sweat glands, which is easy to obtain noninvasively. 51 Rich types of metabolites, such as water, electrolytes, and others, closely related to blood concentration in sweat, 52,53 reflect changes in physiological metabolism. Fatigue biomarkers, such as lactic acid, uric acid, and cortisol, are present in these noninvasive biofluid, based on which abnormal state of humans from multiple aspects can be analyzed. [54][55][56] The relationship between noninvasive biofluid and blood biomarkers has been explored gradually 57,58 that is the technical basis for efficient and accurate analysis. Based on the noninvasive biofluids, which facilitates sampling, wearable biosensors can avoid painful and dangerous blood collection processes and noninvasively extract samples for analysis, friendly to wearer's daily life. 7,51,59 By collecting sweat, saliva, blood, interstitial fluid (ISF), wearable glucose detection with easy operation could be realized. 60 Except for glucose, Figure 1 shows the physiological composition of biofluids and their representative wearable biosensors. Based on the understanding of physiological metabolites in different biofluids, biosensors targeting biofluid characteristics have been developed, such as wearable masks for breathing tracking, 61 fingertip blood sensor for measuring blood pressure, 62 and wearable tattoos for medical monitoring. 63 As presented in Table 2, some wearable biosensors have been developed in recent years. By collecting noninvasive biofluids, wearable biosensors can perform sensitive detection of both physiological biomarkers and biochemical biomarkers. However, deep understanding of the potential of biomarkerrich biofluids in wearable fatigue diagnosis is still unclear. In the following sections, we will introduce advances in wearable biosensors of different biofluids and explore their potential in military operation and exercise-induced fatigue diagnosis. Polydimethylsiloxane (PDMS) onto poly(MPC-co-DMA), could detect the glucose content of tears ( Figure 2a). 83 However, the strip was difficult to fix the iris and hence was hard to adapt to daily exercise.

| Application and prospect of wearable saliva biosensors
Recently, wearable saliva biosensors have attracted researchers for real-time monitoring of special groups of people. For instance, the wearable pacifier 89 can adapt infant fragility and hyperactivity characteristics, providing a feasible development direction for infant physiological state monitoring without blood collection. The prototype of wearable saliva biosensors for health care was the traditional paperbased saliva test strips, which collected saliva from subjects, followed by analyzing the responding targets with visual colorimetry or electrochemistry. 90,91 The detection of glucose and cortisol content in saliva was relatively easier by the test strips based on the colorimetric reaction. To achieve continuous monitoring, paper-based saliva test strips were replaced by wearable 3D-printed microfluidic paper-based silicone braces (μPADs). 84 Such 3D-printed uPADs braces integrated test strip into braces, followed by fitting the oral cavity stably and detecting targets through enzyme-substrate colorimetric reaction ( Figure 2d). However, the brace required repeated removal from mouth for colorimetric analysis, which limited the widespread increased that changed electrical signals gradually. Based on the continuous detecting-abled wearable biosensors, wearable saliva biosensors for detecting uric acid and glucose were also developed, whose research and application results showed that the wearable saliva biosensors could adapt to strenuous exercise to a certain extent. 79 It is worth mentioning that a further miniaturized "cavitas sensor" was later developed by Akiyoshi and his group (Figure 2f). 80 In this work, GOx electrode was modified on the surface of polyethylene terephthalate, which could fit the teeth well and detect changes in saliva glucose (5-1000 μM). Compared to wearable tears biosensors, wearable saliva biosensors, especially the miniaturized biosensors, are more user-friendly with better safety and sample collection efficiency, which are more advantageous in detecting the fatigue or physiological state of athletes and soldiers during their training and operation. 92

| Application and prospect of wearable epidermal biosensors
In addition to saliva and tears, noninvasive biofluids are also present on the skin surface. For example, sweat is vigorously secreted to the skin surface during exercise containing many biomarkers related to fatigue, and hence can be a satisfactory basis for fatigue diagnosis.
What's more tempting is that wearable sweat biosensors do not need contact with mouth and eyes, thereby reducing the sensation of foreign bodies and improving comfort. 93 Besides sweat, ISF, secreted from the superficial layer of the skin, is another biofluid with better potential in fatigue diagnosis. Unlike sweat, ISF surrounds tissue cells, directly obtaining biological metabolites diffused from the capillary endothelium without filtering, 94 and the biological composition in ISF is reliably similar to those in blood. Recent reports have also revealed that ISF can be extracted and analyzed noninvasively by iontophoresis and ultrasonic electroosmosis. 95 The first wearable wristband working on extracted ISF from skin surface was proposed by Tierney and his group 86 ( Figure 2g). In this work, glucose was extracted from skin by iontophoresis, and the entire module was integrated into a wristband with high accuracy and repeatability, whose relative error was considered after an appropriate interval.
The wearable wristband has been commercialized and applied in a

| Multichannel detection of wearable biosensors
Except for fatigue, diseases, such as diabetes 96 and gout 97

| Sensitivity of wearable biosensors
The quantitative analysis of biomarkers that reflect physiological changes is a prerequisite for specific and accurate detection. However,

| Selective elements for biomacromolecules detection
The development of multichannel detection and sensitivity of wearable biosensors provides hope for flexible fatigue diagnosis. However, the present biometric element can only detect the content of ion or small biomolecules, which cannot fully and specifically reflect human fatigue. For example, cortisol, a biomarker present in noninvasive biofluids, is increased significantly during exercise fatigue. However, traditional biological enzyme-substrate biorecognition methods cannot detect cortisol, which poses a challenge to reliable fatigue diagnosis. 115,116 In recent years, antibodies and molecularly imprinted tech- Another common biometric element is the aptamer. The aptamer is an oligonucleotide sequence obtained by systematic evolution of ligands by exponential enrichment (SELEX) and form a specific secondary structure after recognizing the target. 124 Based on the advantages of low cost and stability, aptamers were widely applied in toxin and biomarker detection. 125 However, because of the complicated modification process on flexible substrates, aptamer-modified elec- trodes was yet to be developed. The simplest method was to modify the aptamer generating easy-to-prepare paper-based aptamer biosensors (Figure 6a). In this work, aptamer modified with redox label MB was fixed on the three-electrode paper-based substrate. After recognizing cortisol, the structure of aptamer changed, shortening the distance from the surface of electrode to redox label MB. Then the electrical signal changed because of the difference in distance between redox label MB and electrodes. 126

| SIGNAL MODULES FOR WEARABLE BIOSENSORS
To continuously detect the concentration of secreted fatigue biomarkers during exercise, rapid response and signal readout are essential. The signal modules of wearable biosensors can be mainly divided into two types: optical signal and electrical signal. In the first two sessions, we briefly discussed the development of sample collection and biochemical reaction modules of wearable biosensors. The following discussion will summarize the advanced signal readout modules along with their prospects and challenges in wearable fatigue diagnosis.

| Optical signal module
Optical biosensors convert biochemical reactions into optical signals, including colorimetric, fluorescent, and Raman signals. [128][129][130] The wearable colorimetric biosensor designed by Rogers and his group 131 was considered to be the pioneering research on wearable optical biosensors ( Figure 7a). Herein, a microfluidic system was designed for sweat extraction and collection, in which the color changed according to the specific biological reactions in microfluidic channel via monitoring changes in multiple markers, such as lactate, glucose, pH chloride, and sweat flow rate. The chromaticity data in the channel could be transferred to the integrated NFC module and quantified by software.
The device did not require power supply equipment, which undoubtedly contributed to miniaturization and improved comfort of wearing. 134,135 The other advantage was that optical signals, especially the color changes, 132 (Figure 3a). Another optical signal-based wearable plasmonic-metasurface sensor has been developed by Ying and his group. 68 The wearable plasmonicmetasurface sensor has been based on surface-enhanced Raman scattering (SERS), which has almost realized the "universal" detection of molecular fingerprint. Each molecular own its specific "molecular fingerprint," which could be faintly observed by Raman spectroscopy.
With the help of gaps between rough-surfaced nanoparticles or metal nanoparticles close to each other, the "molecular fingerprint" could be

| Electrochemical signal module
Unlike optical biosensors, electrochemical biosensors output biological reactions as electrical signals. Typically, a pair of conductive silver/ silver chloride electrodes are integrated into flexible three-electrode system to prepare wearable "tattoos" and "patches" based on screen printing technology. Then, wireless transmission devices, such as WiFi, 139 Bluetooth, 27 and NFC module, 140 are integrated and transmit signals to the terminals (e.g., laptops, computers, and mobile phones) for continuous tracking and monitoring changes in electrical signals.
According to the principle of signal change, the electrochemical response signal includes current, potential, and electrical impedance.  74 and lactic acid detection. 143 The first wearable current biosensor was used for glucose detection. 86 When the biologically recognizable element specifically recognized the substrate, GOx reacted with glucose to generate current change in reaction chamber. The signal could be read through wireless transmission equipment. Wearable biosensors, based on current monitoring can also monitor changes in multiple samples simultaneously by iontophoresis. The classic wearable current biosensors of single-target and multitarget detection were developed by Joseph Wang and his group, which have been explained before. 63,74 In conclusion, although wearable electrochemical biosensors require an external power supply, they perform well in monitoring efficiency and practical applications.

| CONCLUSIONS AND PERSPECTIVES
Here, the latest advances of wearable biosensors and their application potential in fatigue diagnosis have been summarized. First, classical wearable biosensors for the detection of various biofluids and their application potential toward fatigue diagnosis have been reviewed. Emphatically, to diagnose the fatigue accurately and conveniently, saliva, sweat, and ISF seemed to be better substrates.
Thereafter, the focus has been shifted to the core of wearable biosensors and improvement of the biochemical reaction module in multitarget detection, sensitivity, and biometric recognition ability.
The development of biochemical reaction modules and their contri- Funding acquisition (lead); project administration (lead); supervision (lead); writingreview and editing (lead).

CONFLICT OF INTEREST
The authors declare no competing financial interest.

DATA AVAILABILITY STATEMENT
The data used in this paper are all available.