Optical biosensing systems for a biological living body

With the development of technology, biosensors are increasingly used in the biomedical field. Due to the high sensitivity of optical signals, good stability, high signal‐to‐noise ratio, and different spectral characteristics of different analytes, optical biosensors can achieve direct and real‐time detection of analytes with high specificity compared with traditional analytical techniques. In view of the rapid development of optical biosensors for in vivo applications and their promising future, we have attempted to summarize the working principles and recent advances in application in humans, with the aim of helping readers to gain an in‐depth and detailed understanding of optical biosensors developed in recent years for applications in the biological living body. In this review, we focus on some of the current widely used and promising optical biosensors, including colorimetric biosensors, fluorescence biosensors, and surface‐enhanced‐Raman‐scattering‐based biosensors. The working principles for each type of optical biosensor are concisely described and the application of the biosensor in the living body is summarized by category. We conclude by looking at the prospects of in vivo applications of optical biosensors.

to their own limitations. For example, water in the body can excessively absorb terahertz waves, resulting in the degradation of terahertz-based biosensors' image quality. 10 In view of the rapid development of optical biosensors for in vivo applications and their promising future, we have attempted to summarize the working principles and recent advances in application in humans, with the aim of helping readers to gain an in-depth and detailed understanding of optical biosensors developed in recent years for applications in the biological living body. In this review, we focus on some of the current widely used and promising optical biosensors, including colorimetric biosensors, fluorescence biosensors, and SERS-based biosensors. At the beginning of each section, we will briefly describe how the optical biosensor works, and then we will summarize the application of the biosensor in the living body by category. We conclude by looking at the prospects of in vivo applications of optical biosensors.

COLORIMETRIC AND FLUORESCENCE BIOSENSING
Colorimetric biosensing systems can be divided into three main categories depending on the strategies that influence the change in color response. The first category not involving nanoparticles (NPs) is based on changes in external physical or chemical factors. The colorimetric biosensor of this type responds to changes in external factors such as pH values [11][12][13][14][15] and temperature [16][17][18] by changing the color of the visual readout. A momentous factor affecting the sensitivity of this sort is the type and concentration of the external indicator.
The second category relies on chromogenic reactions catalyzed by internal enzymes. [19][20][21][22][23] This kind of colorimetric biosensing system requires only the intrinsic catalytic reaction of enzymes and corresponding substrates, simplifying and reducing the cost of the chromogenic process. It should be noted, however, that the enzyme reaction may be affected by factors such as temperature and oxidative stress, which will influence the sensitivity and accuracy of the colorimetric biosensor. In order to overcome the shortcomings and deficiencies of internal enzymes, functional enzyme-like NPs have been developed for use in colorimetric biosensors. [24][25][26][27][28] These enzyme-like NPs function like internal enzymes and bind to their corresponding substrates for efficient catalysis in colorimetric biosensing systems. However, compared to natural internal enzymes, synthetic enzyme-like NPs exhibit better stability to changes in external factors such as pH values and temperature, with a more controlled catalytic activity.
The third category is on the ground of metal-based nanomaterials such as Au-based or Ag-based NPs. [29][30][31][32][33] AuNPs change interparticle distance by aggregation or dispersion in response to different external stimuli with low cytotoxicity and acceptable biocompatibility, which are widely used in colorimetric biosensors. When AuNPs are dispersed, the interparticle distance is larger than the average particle diameter, and the solution will appear red; while when AuNPs are aggregated, the reduction of interparticle distance will make the solution change from red to blue. 34 This distinct color change makes the visual readout of the biosensor easily readable by the naked eye without the assistance of optical instruments. In addition to interparticle distance, the form of NPs plays a crucial role in maintaining the readability and stability of colorimetric biosensors. Although many forms of NPs including nanopillar, 35 nanostar, 36,37 and core-shell structure, [38][39][40] are used in colorimetric biosensing systems, nanosphere is the most common form because of its isotropy. 41,42 NP accumulation reactions are also applied in point-of-care devices including lateral flow assay (LFA) and paper-based colorimetric biosensing systems. [43][44][45][46][47] In order to overcome the disadvantage of the narrow analytical spectrum of colorimetric biosensors, it is an excellent extension to introduce the application of fluorescence biosensing systems. Unlike colorimetric biosensors, fluorescence biosensors reflect changes in the concentration of the analyte mainly by differences in the intensity of fluorescence-containing optical-sensing molecules, with the advantages of high sensitivity and fast real-time monitoring. [48][49][50] Fluorescence biosensors can be divided into direct, "signal-on", and "signal-off" biosensing systems depending on the generation or bursting of fluorescence in the assay. Currently the most common is the direct fluorescence biosensors, which contain a specific ligand with a fluorescent tag that selectively binds to the target and produces a fluorescence intensity proportional to the concentration of the analyte. [51][52][53][54][55] Regardless of the classification of the fluorescence biosensors, labeling fluorescent probes is essential for the construction of fluorescence biosensing systems.
Although colorimetric and fluorescence optical biosensors are now widely utilized to measure physicochemical properties or biomarkers of body fluids such as blood, sweat, and tears, by external sampling techniques, these applications are beyond the extent of this review as they fall within the scope of ex vivo detection. Real-time health status monitoring and clinical diagnosis via wearable devices is the most common application of colorimetric and fluorescence biosensing systems in the biological living body (Table 1)

Sweat detection
Sweat, a body fluid secreted by the skin, is rich in electrolytes, metabolites, and proteins and can be applied as a non-invasive alternative approach to identify physiological status compared to traditional invasive blood analysis. [56][57][58] Koh and his colleague used specially formulated variants of polydimethylsiloxane to develop skin-mounted devices for the capture, storage, and colorimetric sensing of sweat ( Figure 1A), establishing foundations for microfluidic systems to identify the components in sweat via colorimetric reactions. 59 Compared to traditional tattoo patches that come into direct contact with the skin, microfluidic systems can calculate total sweat volume and rate more accurately and are compatible with subsequently developed sensors and electronic devices. Owing to the microfluidic system, athletes can precisely determine the regional sweat loss rate and the concentration of chloride in sweat through colorimetric biosensors for personalized electrolyte replenishment 60 ( Figure 1B).
To overcome the shortcomings of earlier colorimetric biosensors, which could only measure a certain point in time, valves were used in colorimetric biosensing systems, where the visual readout of the sensors could demonstrate a color proportional to the concentration of the sweat analyte, to realize the chrono-sampling of sweat [61][62][63][64] (Figure 2A-D). Further, Bandodkar's group completed an accurate time recording of the sweat utilizing galvanic cells as "stopwatches", to avoid unintentional contamination of sweat samples via chronometric collection 65 ( Figure 2E). Another way to improve the accuracy of the detection is to exclude other factors such as pH values that may interact with the test substance, whereas this seems difficult to achieve by the colorimetric approach alone.
In clinical practice, the most direct relation to the colorimetric biosensor is the diagnosis of cystic fibrosis (CF). CF is an autosomal recessive disease of the exocrine glands and quantitative assessment of chloride concentration in sweat remains the gold standard for confirming the diagnosis of CF. [66][67][68] Traditional methods rely on a Macroduct collection system to collect sweat from infants, 69 which has the disadvantage of not collecting sufficient sweat for subsequent biochemical analysis. 70,71 Zhang et al. developed a smartphone-operated chloridometer using F I G U R E 1 Epidermal microfluidic colorimetric biosensors for sweat detection. (A) Schematic illustration of an epidermal microfluidic sweat monitoring device integrated with the near-field communication (NFC) system and the optical image of the fabricated device mounted on the forearm. Reprinted with permission. 59 Copyright 2016 American Association for the Advancement of Science. (B) Schematic drawings and optical images of wearable microfluidic colorimetric biosensing patch for athletes to detect sweat. Reprinted with permission. 60 Copyright 2020 American Association for the Advancement of Science.
citrate-derived sensor materials that have been optimized to detect chloride in sweat efficiently. 72 A smartphone captured measurable changes in fluorescence emission as sweat chloride levels changed. Recently, Ray et al. developed a soft, epidermal microfluidic device, with skin-safe adhesive and optimized multi-layer geometry, to collect and analyze sweat rapidly. A silver-chlorinate-based colorimetric reaction displayed color change proportional to chloride concentration in sweat with real-time image analysis. 73 The accuracy of this system for CF diagnosis still needs to be validated in a large clinical cohort to compare with traditional external analysis.
In another study, researchers build a soft microfluidic system for passive sweat collection and colorimetric analysis of biomarkers relevant to kidney disorders. 74 In order to determine the concentrations of creatinine and urea in sweat, enzymatic chemistries, and colorimetric readout approaches were applied in the biosensing system. This system creates a potential opportunity to use sweat for kidney disease screening or monitoring with an accuracy comparable to standard methods.
Determination of glucose levels in sweat with a colorimetric biosensor is also used for non-invasive diabetes monitoring 64,75-77 ( Figure 3A-C). Although the upper and  lower limits of glucose levels in sweat can vary up to 1000-fold 78 and are much less stable than blood glucose, the results of several recent studies have shown a positive correlation between glucose levels in sweat and blood glucose, 79-81 giving a basis for diagnosing diabetes by monitoring sweat glucose levels. Due to the low glucose content in sweat, the detection limits of the colorimetric biosensors pose higher requirements. Recent studies have achieved higher sensitivity of sweat glucose colorimetric biosensing by changing the enzyme delivery, increasing the degree of enzyme aggregation, and optimizing the sweat intake and storage 82,83 (Figure 3D,E).
It is worth noting that the current optical biosensors for in situ detection of sweat are mainly limited to colorimetric biosensors. Although fluorescence biosensors are more sensitive compared to colorimetric biosensors, only a few studies have involved fluorescence biosensors, which deserve further expansion. Sekine et al. integrated the fluorescence biosensor into a microfluidic system that allows in situ measurement of ion concentration in sweat through the imaging capabilities of a smartphone 84 ( Figure 4A). Fluorescent probes in the biosensing system can selectively bind chloride ions, sodium ions, and zinc ions in sweat, and the change in fluorescence intensity can achieve rapid and accurate quantitative analysis. Unlike Sekine's work, which could only detect ions, Ardalan et al. presented a wearable patch based on different selective fluorescence biosensing strategies, capable of in situ measurement of sweat volume and pH values as well as chloride, lactate, and glucose concentrations in sweat with the assistance of a smartphone 85 ( Figure 4B). However, as the product is a disposable patch, there is no way to achieve real-time sweat monitoring. How to achieve in situ real-time monitoring effectively and easily may be a future research direction for fluorescence biosensors.
Some fluorescence biosensors rely on only one fluorescent probe with a specific emission wavelength, which makes the biosensors vulnerable to different factors such  Figure 4C). During the oxidation of sweat glucose catalyzed by glucose oxidase, the dual fluorescence nanohybrid of luminescent porous silicon loaded with carbon quantum dots undergo a ratiometric fluorescence transformation from red to blue, and there is a strong correlation between the transformation ratio and sweat glucose concentration. Xu et al. developed an alternative fluorescence strategy. They introduced a fluorescence biosensor for in situ detection of chloride ion concentration in sweat by loading two lanthanide metal-organic frameworks fluorescence materials onto a flexible cotton sheet 90 ( Figure 4D).
Other physicochemical properties and contents of sweat such as pH values, cortisol, lactic acid, vitamin C, calcium, zinc, iron, and uric acid, can also be measured in situ using colorimetric and fluorescence biosensors. 59,63,83,[91][92][93][94] Bandodkar et al. designed a battery-free, wireless electronic sensing platform inspired by biofuel cells for simultaneously monitoring lactic acid. 91 Considering that nutrients play critical roles in maintaining core physiological functions, Kim and his colleagues introduced a miniaturized system allowing simple, rapid colorimetric assessments of the concentrations of multiple essential nutrients in sweat including Vitamin C, calcium, zinc, and iron. 92 He et al. presented a flexible and skin-mounted band that combines superhydrophobic-superhydrophilic microarrays with nanodendritic colorimetric biosensors toward in situ sweat calcium detection. 93

Tears detection
Similar to sweat, the physicochemical properties and enrichment of tears can be detected non-invasively with colorimetric biosensors to determine the health status of the living body. [95][96][97] Contact lenses combined with colorimetric sensors have been widely used for in situ monitoring of tears. Riaz et al. employed Anthocyanin pigments to functionalize the polymeric matrices of contact lenses to continuously monitor the ocular pH values. 98 Functionalized contact lenses exhibited a change in color from pink to purple to blue as the pH values change from 6.5 to 7.5. Another colorimetric contact lens for measurement of tears pH values was developed based on methyl red, bromothymol blue, and phenolphthalein that exhibit changes in reflected light in the visible spectrum 99,100 ( Figure 5A,B). Chen et al. improved the process of making microchannels in contact lenses by using 3D printing technology to create poly(2-hydroxyethyl methacrylate) hydrogels with the integrated colorimetric pH biosensing system 101 ( Figure 5C). Previous studies have demonstrated that glucose concentrations in tears are higher in diabetic patients than in normal subjects and that there is a positive correlation between glucose concentrations in tears and blood, [102][103][104][105] suggesting the potential of glucose in tears as a biomarker for diabetes. The two-step enzymatic approach involving glucose oxidase and peroxidase is the most common method utilized for glucose colorimetric biosensors. Depending on the amount of oxidation products, the biosensors exhibit a color change that matches the glucose concentration 99,100,106 ( Figure 5D). Recently, cerium oxide nanoparticle-based contact lenses based on a two-step enzymatic method have been designed for colorimetric determination of glucose in tears 107 ( Figure 5E). The oxidation product of glucose, hydrogen peroxide, could rapidly reduce colorless Ce 3+ to yellow Ce 4+ with high sensitivity. However, due to the slow reduction from Ce 4+ to Ce 3+ , this biosensor cannot perform continuous tear glucose monitoring.
Colorimetric biosensors can also be used to detect proteins and ions in tears. Moreddu et al. used the reaction of 3′,3′,5′,5′-tetrachlorphenol-3,4,5,6tetrabromsulfophthalein with free amino groups in proteins to display a color change from beige to light blue for different protein concentration. The same group also assayed the concentration of l-ascorbic acid and nitrite in tears based on the redox properties of phosphomolybdic acid and diazonium salt. 99,100 Chloride ions are also a critical biomarker in tears, reflecting tear osmolality. In the presence of mercuric 2,4,6-tripyridyl-s-triazine, Fe 2+ , and Hg 2+ , the color changed from colorless to blue based on the concentration of chloride ions in tears. Yang et al. used this colorimetric reaction to construct a contact lens capable of in situ detection of chloride ion concentration in tears. 106 Tear analysis can also be performed quickly and efficiently using contact lenses with fluorescence biosensors. Back in 2003, Badugu et al. developed contact lenses loaded with boronic acid-containing fluorophores for tear glucose monitoring. 108,109 The subsequent improvements were mainly based on the difference in the fluorescent groups used for detection. March et al. designed contact lenses using liquid hydrogel nanospheres loaded with tetramethyl-rhodamine isothiocyanate concanavalin and fluorescein isothiocyanate dextran. 110 Fluorescence intensity increases with increasing concentration of glucose in tears. In addition to monitoring glucose, Yetisen et al. described a multiplexed contact lens biosensor containing multiple fluorescence probes such as benzodicarboxylic acid and crown ether derivatives to quantify pH values, potassium, sodium, calcium, magnesium and zinc ions in tears 111 ( Figure 6A). The fluorescence signal can be read by a customized handheld device and quantified by a deconvolution strategy. This biosensor is useful for grading the severity of dry eye and identifying subtypes. In order to achieve continuous non-invasive measurement of tears, a strong bonding of the fluorescent probes to the lens is required. Badugu et al. strengthened the adhesion of the probes to the contact lenses by changing the material of the lens and improving the way the lens and the fluorescent probes are attached [112][113][114] ( Figure 6B-D).

Exhalation products detection
The coronavirus disease 2019 (COVID-19) pandemic, which began in 2019, has caused a public health crisis, infecting a large number of people to date, and imposing a significant social and economic burden. [115][116][117] Early diagnosis of COVID-19 allows for timely treatment of patients, reduces psychological burden and sequelae, and is critical in preventing its widespread transmission. [118][119][120] Currently, the detection of viral nucleic acid titers using reverse transcription polymerase chain reaction (RT-PCR) is the gold standard for the diagnosis of COVID-19, and viral antigen testing is also utilized for early self-screening of COVID-19 infection. 121 However, both methods are based on invasive and uncomfortable nasopharyngeal swab sampling. On the one hand, it may lead to false negative results due to inadequate sampling, and on the other hand, sample transfer during the sampling to testing process may make the sample contaminated and thus make the test results less accurate. 122 Due to the low cost of surgical masks and the ease of capturing aerosols produced during breathing, integrating a colorimetric biosensor into surgical masks appears to be an acceptable option for detecting COVID-19. Nguyen et al. at Harvard University achieved a colorimetric response to rapidly collected splashes by integrating a freeze-dried-cell-free-based colorimetric biosensor into a flexible, resilient cellulose substrate. Subsequently, they designed a mask with four modules (a reservoir for hydration, a sample collection pad, a three-step μPAD, and an LFA strip) to enable accurate detection of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in exhaled aerosols at ambient temperature, without the need for additional power. The three-step μPAD module allows for a customized isothermal amplification reaction, enabling simple and reliable colorimetric visual readout via LFA 123 ( Figure 7A). Vaquer et al. used a colorimetric biosensor based on gold NPs for the diagnosis of COVID-19. The AuNPs integrated into the mask can bind specifically to viral antigens in the hydrophobic polypropylene layer of the mask, producing a colorimetric signal that can be quantified with a smartphone. According to an analysis of a cohort of 27 patients with mild to no symptoms, the area under the curve for this test was 0.99, with 96.2% sensitivity and 100% specificity 124 ( Figure 7B). In addition to the direct detection of SARS-CoV-2 viral nucleic acids in the mask, exhaled breath metabolites can be assessed with colorimetric biosensors to assist in the diagnosis of COVID-19. Previous findings suggest that the concentrations of certain exhaled metabolites such as ethanal, methanol, iodobenzene, and 1-propanol, are different in COVID-19 patients than in healthy individuals and other respiratory diseases. [125][126][127] Bordbar et al.
designed the optical sniffer devices which are mounted on thin masks and exposed directly to exhaled breath streams. Exhaled metabolites can interact directly with colorimetric elements, such as porphyrazines, modified organic dyes, porphyrins, inorganic complexes, and AuNPs, causing a color change in the biosensing system within the mask, thus aiding in the diagnosis of COVID-19. In terms of assay reliability, it is calculated that patient, healthy, and cured samples can be differentiated with approximately 80%-84% accuracy. A linear correlation exists between the color changes of the colorimetric biosensing system and the viral load determined by RT-PCR 128 ( Figure 7C).

Wound infection monitoring
With the number of surgical procedures and the prevalence of chronic diseases such as diabetes increasing every year, acute and chronic wound healing is gaining importance. 129,130 The occurrence of infection is a major complication of acute as well as chronic wounds, negatively impacting the effectiveness of wound healing, the quality of life of patients, and social economics. 131 Early diagnosis of wound infection is critical to accelerate wound healing and improve the quality of life of patients. The easiest and most direct way to monitor wound healing is to measure the pH value of the wound in situ in real-time. Tamayol's group introduced the pH-sensitive hydrogel fibers that can be used to monitor epidermal wounds over time 132 ( Figure 8A). Mesoporous microparticles are loaded with self-developed pH-sensitive dyes and incorporated into hydrogel fibers obtained through microfluidic spinning. Smartphones can be used to capture images of pH-sensing fibers during real-time pH measurements for convenient on-site reading. It is possible to extract a quantitative pH map of the hydrogel fibers and the underlying tissue through image processing. Other groups used different pH-sensing indicators to prepare the wound dressings separately. In Pan's study, they developed a fibrous material containing curcumin, a biocompatible color-changing substance, that allows for in situ visual monitoring of wound pH 133 ( Figure 8B). A pH change from 6 to 9 causes the curcumin-loaded fibrous mat to display an obvious and visible color change from yellow to red-brown. Similarly, this fibrous mat can also be used to determine the pH value of the wound specifically by analyzing the image from the smartphone camera. As fibrous materials have a high degree of flexibility, they have been processed from 1D to 3D to fit irregular wounds. Alsahag et al. introduced a new aerogel wound dressing based on anthocyanin probes that can sense the progress of wound healing. 134 In addition to general wound monitoring, Zhu et al. also developed a hydrogel dressing specifically for diabetic patients. Their work involves the development of a multifunctional zwitterionic hydrogel that detects both pH and glucose levels simultaneously to monitor diabetic wound status in real time. 135 The anti-biofouling zwitterionic poly-carboxybetaine hydrogel matrix encapsulates two glucose-sensing enzymes, glucose oxidase, and horseradish peroxidase as well as the pH indicator dye. A smartphone collects visible images and converts them into RGB signals in order to quantify the wound pH value and glucose levels. While providing a moist healing environment that can promote diabetic wound healing, this wound dressing also monitors pH and glucose levels effectively ( Figure 8C).

SERS-BASED BIOSENSING
Raman scattering was first theoretically predicted in 1923 by Smekal. 136 It is a spectroscopic technique in accordance with the inelastic scattering of photons, characterized by the vibration of chemical bonds in molecules when light interacts with them. 137 In 1928, C. V. Raman experimentally confirmed the existence of Raman scattering, for which he was awarded the Nobel Prize in Physics in 1930. 138 The ability of Raman spectroscopy to calculate the composition and structure of analytes based on vibrational information has led to its subsequent widespread use in biomedicine, analytical chemistry, and other fields. [139][140][141][142][143] The weak Raman scattering cross section leads to inefficient Raman scattering, which limits its application in analysis practice due to the lack of sufficient signal intensity. Fleischmann et al. found that the Raman scattering signal was significantly enhanced when certain molecules were adsorbed on the surface of silver electrodes, 144 which was named SERS. Jeanmaire's group then confirmed that metal matrices possessing nanomaterial characteristics are the main cause of SERS. 145 It is now generally accepted that there are two main mechanisms of SERS: electromagnetic enhancement and chemical enhancement mechanisms. The SERS electromagnetic enhancement mechanism is based on the presence of surface plasmonic resonance, which usually occurs when electrons on the surface of metal nanostructures resonate with incident light. Since noble metals such as gold, silver, and copper nanostructures have free electrons on their surfaces, the electron oscillations lead to an enhanced electric field in the surrounding space, and therefore SERS electromagnetic enhancement is mainly based on noble metal NPs. [146][147][148][149][150] Since the increase in intensity to the observed Raman scattering signal typically exceeds 10 orders of magnitude, electromagnetic enhancement is thought to be the main mechanism of SERS. Chemical enhancement relies mainly on charge transfer processes between surfaces and adsorbed molecules, which occurs mainly on semiconductor surfaces or non-metallic surfaces such as carbon nanomaterials. 151,152

In vivo diagnosis and imaging
Due to the high sensitivity of SERS, researchers have combined SERS probes and image reconstruction techniques for in vivo imaging applications. 153 SERS signal is much more stable to photobleaching compared to fluorescence. 154 Moreover, in Raman scattering, the spectral signatures are unique, making it possible to distinguish target signals from background signals with unmistakable ease, demonstrating a high signal-to-noise ratio. 155,156 In addition, SERS probes with different spectral characteristics can be activated by a single wavelength, presenting a satisfactory multiplexing effect. SERS-based in vivo imaging is currently most commonly used in oncology research due to the enhanced permeation and retention effect, which allows SERS NPs to passively target and aggregate at tumor sites, thus enabling tumor imaging through SERS. [157][158][159] Although it is possible to recognize certain molecules in vivo such as DNA and proteins directly as Raman reporter molecules, so-called label-free SERS detection, the different molecular structures and the complexity of the human body environment make this direct recognition approach lack stability and is, therefore, less used in in vivo imaging. [160][161][162][163][164] Currently, the use of SERS nanosubstrates loaded with Raman reporter molecules with large Raman cross sections as SERS tags is the main approach for SERS detection, which provides a stable SERS signal intensity with high sensitivity and specificity. [165][166][167][168] Qian et al. pioneered the application of SERS tags for in vivo tumor detection and imaging. They designed a pegylated-AuNP-based SERS tag with a small molecule dye as a Raman reporter molecule and a single-chain variable fragment as a ligand to achieve targeting of EGFR-positive tumors in an animal model, which can be used for tumor imaging based on photothermal effects, with higher brightness compared to fluorophores. An interesting finding is that thiol-modified PEG can increase the stability of Raman reporter molecules such as organic dyes. 169 Since multi-labeling methods may improve the sensitivity and specificity of detecting certain diseases, Zavaleta et al. tried to spectrally separate and correctly identify up to 10 different types of SERS tags in animal models. They successfully multiplexed four different Raman labels in vivo and observed a linear relationship between increased Raman signal and increased concentration of SERS multiplexed labels, allowing semi-quantitative prediction of SERS NPs accumulation in vivo 170 ( Figure 9A). Following the pioneering works, SERS is increasingly being used for diagnostics and imaging in vivo, primarily with improvements to the SERS labels. Yin et al. introduced a novel SERS tag with waxberry-like AuNPs containing lipid bilayers, SiO 2 coats, and near-infrared (NIR) dye. Lipid bilayers play a stabilizing role for Raman reporter molecules 171 ( Figure 9B). Bock's group precisely controlled the size of AuNPs to synthesize AuNP-assembled SiO 2 NPs with small nanogaps for the development of NIR SERS tags 172 ( Figure 9C). AuNP-assembled SiO 2 NPs with 14 different Raman label compounds demonstrated distinct SERS signals upon subcutaneous injection, F I G U R E 9 Surface-enhanced-Raman-scattering (SERS)-based biosensors for in vivo imaging in cancer. (A) Evaluation of multiplexing exhibiting the potential in multiplex imaging in vivo. Other improvements for SERS probes include the application of nanorods, [173][174][175] nanostars, [176][177][178] nanoflowers, 179 and core-shell nanostructures. [180][181][182][183] Improving the quality of SERS in vivo imaging by optical clearing is also an acceptable option. 184,185 Since visible light is scattered or absorbed in large amounts, 155 limiting its penetration into the tissue and thus poor imaging, most early SERS detection was based on the first window of NIR with wavelengths of 650-900 nm, mostly 785 nm. Although the first window of NIR provides more excellent tissue penetration than the visible wavelength, the second window of NIR actually has further reduced scattering, absorption, and background noise compared to the first window due to the longer wavelength. [186][187][188] In recent years, there have been several studies that performed in vivo SERS detection within a second window of NIR 189,190 (Figure 9D,E). Li et al. designed a nanostructure of porous-cubic-AuAg nanoshells with hollow interiors, exhibiting excellent NIR-II plasmonic properties for in vivo SERS imaging. By controlling the features of the pore structures in the nanoshells, it is possible to tune the plasmonic properties of designed AuAg nanoshells over the broad NIR-II range. They then validated the AuAg-nanoshells-based SERS probe for its non-invasive and highly accurate in vivo imaging of tumor tissues, especially microscopic tumors, in the second window of NIR on a mouse model, demonstrating its great promise for microtumor identification and accurate tumor tracing 191 ( Figure 9F).
Although most of the current in vivo SERS detection and imaging is focused on oncology, SERS imaging has applications in other diseases as well. Noonan et al. designed a string of antibody-functionalized SERS-based Au nanoprobes to identify ICAM1, VCAM1, and P-selectin in vitro and ex vivo. Multiple imaging of these molecules in vivo was further confirmed by intravenous injection of antibody-functionalized SERS-based Au nanoprobes in a mouse model, suggesting multiplexed SERS-based in vivo imaging can demonstrate vascular inflammation, laying the groundwork for future applications in cardiovascular disease of SERS in vivo imaging 192 ( Figure 10A). Jiang et al. achieve ultrasensitive in vivo detection of bone crack with PDA-capsulated SERS NPs. The bone crack can be specifically labeled using PDA-capsulated SERS tags due to the high affinity of PDA towards calcium exposed to the damaged bone. After intravenous injection of a PDAcapsulated SERS probe in the mouse model, bone cracks can be effectively labeled and generate a strong Raman signal which can be easily detected, indicating promise in the ultrasensitive detection of the bone crack in clinic 193 ( Figure 10B).

SERS-guided surgery
For most patients with solid tumors, surgical resection remains the best treatment option. However, underresection of tumor tissue may lead to residual lesions and increase the chance of cancer recurrence, while overresection may affect the function of normal tissues. This requires the surgeon to precisely determine the extent and boundaries of the tumor during surgery to achieve the most accurate and complete resection. Given the ability of SERS to target tumors in vivo and its excellent imaging capabilities, the use of SERS imaging to guide surgery in real time appears to be an option. Mohs et al. introduced a hand-held spectroscopic pen device, which was called SpectroPen, for intraoperative detection of tumors, depending on fluorescence and SERS signals. They further found that the tumor borders can be accurately detected preoperatively and intraoperatively in the breast cancer mouse model. Optical signals and tumor bioluminescence also demonstrate strong positive correlations. After surgery, the SpectroPen device permits further  Figure 11A). A variety of different SERS nanotags were subsequently used for intraoperative image-guided surgery [195][196][197][198] (Figure 11B,C). Kircher et al. combined SERS and magnetic resonance (MR) and photoacoustic to construct a novel MR-photoacoustic -SERS nanoparticle to assist clinicians in accurately mapping the edges of brain tumor tissue preoperatively and intraoperatively. Experiments in glioblastoma-bearing mice confirmed that this nanoparticle accumulated in tumor tissue and was not expressed in healthy tissue. 199 This combination of SERS applications with macroscopic imaging may become a future direction. Other studies have used SERS probes to make improvements to surgical instruments. They loaded SERS NPs in conventional endoscopes as molecular imaging contrast agents to provide multiplexed clinical data with high sensitivity to identify microscopic lesions that are difficult to detect or easily missed by ordinary white light endoscopes 200,201 (Figure 11D,E).

Sweat detection
Similar to colorimetric and fluorescence biosensors, SERSbased biosensing can be used for in situ sweat detection. Mogera et al. developed a wearable paper-based microfluidic system containing Au nanorods for in situ sweat analysis. The paper microfluidics enables rapid and accurate determination of sweat rate and loss. The integrated Aunanorods-based SERS biosensor can detect and quantify the concentration of uric acid in real time based on labelfree SERS 202 ( Figure 12A). Koh et al. designed a patch with a three-layer structure containing a sweat collection layer, an Ag-nanowire-based SERS biosensor layer, and a protective layer for the SERS biosensing of sweat. The protective layer allows direct in-situ Raman spectroscopy without the need to remove the patch from the skin 203 ( Figure 12B). Chung's group fabricated a flexible SERS active nanosubstrate based on a combination of electrostatic spinning of thermoplastic polyurethane and gold sputtering coating, exhibiting fast and reliable sweat absorption properties. The biosensing system can be used for in situ pH determination of sweat after functionalizing the SERS substrate with two common pH-responsive molecules 204 ( Figure 12C).

Wound infection monitoring
Similar to colorimetric and fluorescence biosensors, some SERS biosensors are also available for in situ wound monitoring. SERS biosensors are effective tools for identifying and monitoring wound surface bacteria Mei et al. designed highly sensitive internal-standard SERS microneedles for direct detection of bacterial metabolite in dermal interstitial fluid. 205 Perumal et al. developed a biofunctionalized flexible, cost-effective, scalable, and easy-to-fabricate plasmonic SERS substrate using cellulose fiber, which could be incorporated into wound dressings in the future for routine monitoring of wound healing status. 206 He and his colleagues achieved non-invasive, highly-sensitive detection and long-term tracking of residual bacteria in wounds using gold-silver-nanoshells-based SERS. 207

CONCLUSION AND PERSPECTIVE
In this review, we provide a comprehensive and detailed description of the research progress of optical biosensors in the field of health monitoring and clinical diagnostics in the biological living body. We focused on the application of colorimetric biosensors, fluorescence biosensors, and SERS-based biosensors in health monitoring and diagnosis in vivo, such as in situ detection of pH value, glucose, ions, and proteins in sweat and tears, real-time detection of exhalation products, wound healing monitoring, and in vivo imaging. Despite the significant advances in the use of these optical biosensing systems in vivo, there are still considerable challenges to overcome to make further progress. Colorimetric biosensors offer the advantages of low cost, ease of operation, and easy readout. Although the ability to read distinct color changes with the naked eye seems to be an advantage of colorimetric biosensors, the fact that each person's visual perception is different and can be influenced by the environment, lighting, and emotions, may result in less than an appreciable determination of the colorimetric biosensor results. An alternative solution is to combine a smartphone and a colorimetric biosensor to quantitatively calculate color changes using pictures taken by the smartphone camera. But this will also bring some potential problems. Each brand of the smartphone takes pictures with different modes and algorithms, which will lead to differences in the contrast and saturation of the pictures taken. How to unify the quality of pictures on different smartphones is a problem that needs to be solved in the future. Another major challenge for colorimetric biosensors is the innovation of colorimetric indicators.
The color change pattern of the currently existing colorimetric indicators is relatively homogeneous, often shifting from one color to another, lacking discernible intermediate colors. The development of new colorimetric indicators capable of displaying different colors in sharp contrast to determine the process of analyte or disease change deserves further research in the future. In addition, the improvement of materials for constructing colorimetric biosensors is also a future research direction. For example, most of the colorimetric biosensors currently used for in vivo detection are wearable devices, and the material can be changed from rigid to flexible considering the wearing comfort.
Compared to the narrow analytical spectrum of colorimetric biosensors, fluorescence biosensors have high sensitivity and specificity. At the same time, the fluorescence biosensor has excellent multiplexing capability because different substances have specific emission wavelengths. However, similar to colorimetric biosensors, a challenge with fluorescence biosensors is the analysis of the results. Early fluorescence biosensors were not simple to quantify the results and often required special instruments and software for analysis, which limited their daily use. Although the popularity of smartphones has led to improvements in readout devices, there is still the inevitable problem of distortion in images taken by smartphones. Further development of portable and easy-to-use readout devices or integration of readout devices into biosensors deserves deeper research. On the other hand, fluorescence biosensors rely on fluorescence intensity or fluorescence decay time to associate with the analyte, which requires the stability of the fluorophore. However, most existing fluorophores have short lifetimes and are highly susceptible to responding to changes in confounding factors, as well as to photobleaching, which generates background noise to the fluorescence signal. New stable fluorophores such as inorganic fluorophores should be developed, or existing fluorophores should be improved to enhance the accuracy and stability of the fluorescence biosensors. In addition, current fluorescence biosensors are limited to the detection of glucose and specific ions in body fluids. Considering the multiplexing potential of fluorescence sensors, the possibility of simultaneous detection of more biomarkers such as immunoglobulins and inflammatory markers can be explored.
SERS-based biosensors allow convenient, rapid, and sensitive quantitative analysis of target molecules. SERS biosensing-based in vivo imaging and image-guided surgery reduces the cost of clinical diagnosis, aids clinicians in more complete lesion resection, and improves patient prognosis and quality of life. The in vivo application of SERS-based biosensors is only a decade old and in its infancy, with challenges and opportunities for future research. The plasmonic nanostructures in the exist-ing SERS-based biosensors have various shapes, including nanostars, nanorods, nanoflowers, and core-shell structures. The different shapes have different requirements for the optimal particle size, spacing, and morphology of NPs.
There are relatively few studies on how to precisely control nanoparticle synthesis, which deserves more effort. Moreover, high-quality complex nanosubstrates for SERSbased biosensors often require advanced and sophisticated instruments for preparation, with high processing and equipment maintenance costs, limiting their large-scale production and further commercial applications. Due to the inconvenience of using Raman spectrometers, it is also particularly essential to develop SERS readouts that are convenient, portable, and capable of accurate and fast readings.

A U T H O R C O N T R I B U T I O N S
Tao Yan and Changfa Guo carried out a literature search. Kai Zhu and Chunsheng Wang conceptualized the study, managed the project, and acquired funding. All authors contributed to the scientific planning, discussions, and writing of the final manuscript.

A C K N O W L E D G M E N T S
The authors acknowledge that this work was supported by grants from the Science and Technology Commission of Shanghai Municipality (20ZR1411700).

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare no conflict of interest.

R E F E R E N C E S
surgery, the design of aortic smooth muscle chips for drug screening, and the development of biosensors for cardiovascular diseases.
Kai Zhu received his PhD Degree from Fudan University. He is currently working as an Associate Professor at Department of Cardiovascular Surgery, Zhongshan Hospital, Fudan University. His research interests include the development of biosensors and high-throughput chip for biomedical applications.