Wearable optical patches for non-invasive sweat sensing

Flexible substrates

Flexible substrates are essential for wearable optical sweat sensors^[46], providing mechanical compliance with skin deformations, long-term wearing comfort, and compatibility with optical sensing modalities. Various materials have been explored, including stretchable polymers, hydrogels, textiles, paper, and ultrathin or structured films [Figure 2]. These materials offer distinct advantages in flexibility, breathability, biodegradability, optical transparency, and integration with microfluidics and optical components.

Figure 2. Representative flexible substrates and their material characteristics for wearable optical sweat sensors. The illustrated substrates include PDMS, hydrogels, textiles, paper, and TPU. Created in BioRender. Chu Wang. (2026) https://BioRender.com/w58hd9a, Adobe Illustrator, and Maxon Cinema 4D R21. PDMS: Polydimethylsiloxane; TPU: thermoplastic polyurethane.

Paper-based substrates

Filter paper, a degradable cellulose natural material, offers an appropriate substrate for sweat sensors^[43]. Paper-based substrates are low-cost, biodegradable, and highly porous, facilitating sweat absorption and transport^[67,68]. For instance, a piece of Whatman Grade 1 filter paper can be embedded into a PDMS microfluidic device as a pre-loaded colorimetric reagent reaction zone, where its high porosity promotes efficient sweat absorption^[50]. Plasmonic paper substrates can be fabricated by adsorbing silver nanoparticles (Ag NPs) onto filter paper fibers. The resulting flexible substrate can be folded and bent without cracking and retains its plasmonic properties even after multiple bending cycles^[67,69]. Furthermore, cellulose-containing fluorescent sensors are prepared by linking fluorescent probes on paper substrates and integrating a cotton thread for sweat collection, making them compatible with portable imaging modules for in situ data capture^[43].

Sweat collection

Effective and reliable sweat collection serves as a cornerstone for accurate metabolite sensing in wearable optical platforms, directly influencing data integrity. A key advancement in this field is the introduction of epidermal microfluidic systems that enable simultaneous sweat collection and in-flow analysis^[39]. These systems address persistent challenges such as low sweat volumes, irregular sweating without stimulation, and potential biodegradation or evaporation^[3], which have limited earlier methods. By leveraging efficient, miniaturized microfluidics specifically designed for passive sweat collection^[39], such devices offer an integrated solution. They not only reduce evaporation and skin contamination^[39] but also isolate sweat from the skin surface, thereby preventing continuous leaching of chemicals into the sweat^[3]. Strategically designed microchannels direct older sweat away from the sensing zones while allowing freshly secreted sweat to flow in, which facilitates real-time analysis. For optical sensing specifically, these microfluidic channels act as dedicated conduits, delivering fresh sweat to sensing zones containing colorimetric or fluorescent reagents. This enables sequential analysis and volume measurement while protecting the sensing chemistry from the external environment, directly enhancing optical readout reliability and accuracy^[73,74].

Beyond fundamental capture, sophisticated design enhances temporal resolution and functionality. For chronological sampling and dynamic analysis, passive capillary bursting valves (CBVs) with precisely engineered bursting pressure gradients are integrated into microfluidic networks to guide sweat in sequentially filling a series of chambers, spatially separating samples over time^[71]. This enables time-sequenced capture and analysis, crucial for tracking metabolic dynamics. For instance, a patch with a 0.5 µL microchamber volume operating at 0.25 µL/min can fill each chamber within about 2 min, yielding a well-defined temporal profile^[71]. Another critical aspect for continuous operation is ensuring that newly produced sweat promptly interacts with the sensor and is subsequently transported away to allow a constant reaction process. This can be achieved by designing channels within the adhesive layer to divert analyzed sweat away from the sensor surface^[75]. Moreover, when natural passive sweat production is insufficient, localized sweat stimulation via transcutaneous iontophoretic delivery of agonists such as acetylcholine or pilocarpine provides an active means to generate adequate sample volume for subsequent detection and analysis^[75]. Collectively, these strategies such as passive collection, timed valving, flow management, and active stimulation ensure the acquisition of high-integrity sweat samples for reliable optical sensing.

Fabrication techniques

Beyond the design of sweat collection and fluidic management systems, the practical realization of wearable optical sweat sensors hinges on advanced fabrication techniques that translate conceptual designs into functional, reliable devices. These techniques must accommodate a wide range of flexible substrates from elastomers and hydrogels to textiles and paper, while enabling the precise integration of microfluidic networks, optical sensing elements, and sometimes electronic components. The choice of fabrication method directly impacts sensor performance, scalability, and cost. The diverse manufacturing approaches for various optical sensors were summarized in Table 1.

Several key fabrication paradigms have been developed. Soft lithography is a cornerstone technique for creating intricate microfluidic structures in elastomers such as PDMS. It involves casting and curing PDMS prepolymer against lithographically prepared silicon molds to produce solid elastomers with relief features, which are then bonded to form sealed microfluidic channels and reservoirs^[50]. For rapid prototyping and potentially high-throughput manufacturing, laser cutting and die cutting of thin polymer films have emerged as viable alternatives. These techniques can process multilayer stacks of materials in a roll-to-roll manner, enabling the fast and cost-effective production of complex microchannel and assay well networks^[76]. When working with fibrous materials, electrospinning is employed to create flexible, porous substrates such as thermoplastic TPU nanofibrous mats. This process involves ejecting a polymer solution under a high-voltage electric field to deposit fine fibers on a collector, which can subsequently be functionalized (e.g., with SERS-active gold coatings) for sensing applications^[70]. For substrates such as paper, laser cutting based on pre-designed patterns allows for the precise shaping of microfluidic channels^[67], while dip coating is used to uniformly modify these porous matrices with functional nanomaterials, such as adsorbing gold nanorods (Au NRs) onto chromatography paper for SERS sensing^[67]. Furthermore, techniques such as hydrogel photo-crosslinking under ultraviolet (UV) light enable the formation of tough, adhesive, and sensor-encapsulating hydrogel matrices from precursor solutions^[55], and solid-state dewetting of metal films on flexible supports facilitates the creation of large-area, uniform plasmonic nanostructures for SERS^[71]. The integration of biochemical recognition elements often relies on enzyme immobilization or immunoassay assembly protocols on these fabricated platforms^[77,78]. For more complex integration, additive manufacturing techniques such as 3D printing are utilized to produce custom components such as smartphone-based imaging modules housing light sources and filters^[79]. These diverse fabrication strategies collectively underpin the development of the multi-analyte detection platforms summarized in Table 1, enabling the transition from laboratory concepts to wearable devices.

Adhesion and encapsulation

The successful translation of these fabricated components into a robust wearable device requires effective strategies for skin adhesion and system encapsulation. Robust skin adhesion and effective system encapsulation are paramount for ensuring stable operation, reliable data acquisition, and user comfort in wearable optical sweat sensors^[80]. The choice of form factor and adhesion strategy is often application-dependent; for medical uses, patch-style formats that discreetly adhere to the skin are preferable^[3]. A common design for such epidermal devices involves a multi-layer structure, where a functional layer for sensing and fluidics is integrated with a dermal contact layer for skin adhesion, frequently utilizing a medical adhesive with a sweat collection port to ensure stable conformal contact^[71]. Wearable flexible microfluidic platforms use low-modulus elastomeric (e.g., PDMS) and biocompatible materials and hypoallergenic silicone adhesives to fabricate a watertight and strong interface for sweat collection^[9]. The general approach to achieving a strong packing around the skin sweat capture region, essential for accurate sampling, is using a robust seal to prevent leakage and contamination^[45,81]. A key issue is ensuring secure yet comfortable adhesion, particularly for users with sensitive skin. Hydrogel-based adhesives, offering biocompatibility and breathability, provide a promising solution by enabling firm yet gentle adhesion that adapts to skin movements^[15]. Furthermore, a robust skin adhesion can elevate signal reliability, but excessive adhesion makes removal difficult and causes skin irritation. Recently, some researchers reported dry adhesive electrodes, aiming to realize a reversible and robust adhesion on skin for sweat collection^[46].

Beyond skin-device attachment, effective encapsulation is critical for maintaining integrity of the internal sensing environment and managing mechanical stresses. A significant challenge in flexible sensors is the interfacial instability, where strain and/or stress concentration occurs at soft-hard interfaces, resulting in a main cause of failure via detachment/delamination^[82]. Generally, the principles in solving soft-hard instability are avoiding abrupt hardness/softness differences and improving interfacial adhesion^[82]. For systems that incorporate less stretchable functional components, a widespread design technique is to combine them in “islands” electrically and mechanically interconnected by stretchable conductors, thereby achieving overall system stretchability while protecting critical elements^[75]. The encapsulation of microfluidic and optical sensing components themselves is often realized through precise multilayer assembly and lamination. This can involve fabricating structures by stacking and laser-cutting layers of medical adhesives, with the assembly completed by aligning and attaching polymer films (e.g., PET) using additional adhesive layers to form sealed, functional devices^[71,81]. To achieve long-term and continuous sensing, a well-sealed detection chamber has been implemented by combining flexible plastic substrates (e.g., PET) with soft microfluidics (e.g., PDMS). The chamber effectively harvests sweat, reducing sample and contamination and evaporation^[45]. Even with the mitigating effects of highly multiplexed sensors and nuanced designs, developing robust encapsulation materials and packaging to protect against corrosion, moisture, and other noise factors remains important for reliable optical sensing, particularly for ultralow concentration species^[9]. The resulting mechanical robustness and flexibility of the fully encapsulated device ensure conformal dermal contact and enable it to securely preserve sweat under twisting, detachment, compression, and other physical stresses, thereby validating the overall reliability of the integrated system for real-world use^[71].

Colorimetric sensors

Colorimetric sensors function by generating visible color changes through enzymatic or chemical reactions, in which chromophore molecules interact with target biomarkers to change the absorption of light, producing a color change in proportion to analyte concentration^[84-87]. To systematically assess this modality in the context of wearable sweat sensing, we evaluate it across several critical dimensions: its sensing mechanism and operational simplicity, sweat collection and delivery requirements, on-body signal robustness, quantitative performance, and suitability for real-time continuous monitoring. This method offers significant advantages for integrating wearables, especially by eliminating complex electronics and allowing passive, flexible and inexpensive substrates such as paper, textiles or polymers^[87-89]. In terms of performance, they offer good sensitivity for major sweat metabolites and electrolytes [Table 1]; however, their quantitative accuracy is often semi-quantitative and can be compromised by uneven color development, intrinsic chroma error, and a dependence of signal intensity on both analyte concentration and sample volume^[77,86,88]. For reliable on-body operation, effective sweat delivery via integrated microfluidics is often crucial to control reagent mixing and enable sequential analysis^[87,89]. However, a major limitation is that chemical reactions are usually endpoint-based, which means that current wearable colorimetric systems do not allow for real-time and continuous monitoring of dynamic concentration changes^[50,90]. Moreover, the compatibility of these sensors with everyday devices such as smartphones for digital imaging enhances their capability for simple, rapid in-situ analysis of sweat constituents [Figure 3A]. Data analysis can range from semi-quantitative visual comparisons to sophisticated smartphone-based digital image processing techniques, including RGB (red-green-blue)/HSB (Hue, Saturation, Brightness) extraction or the use of CIE 1976 Lab color space (CIELAB) color space and deep learning algorithms to enhance accuracy and mitigate environmental variability^{[50,86,87,91]}.

A skin-attachable paper fluidic sensor patch has been developed by printing microfluidic patterns with ceramic-based ink on cellulose paper. The patch is equipped with detection zones modified by nanoparticles for colorimetric detection of sweat analytes and for simultaneous monitoring of the sweat rate [Figure 4A]^[87]. The ceramic ink creates impervious, thermally stable barriers to prevent leakage, while amine-functionalized nanoparticles enhance color homogeneity and sensor sensitivity. The device employs specific enzymatic reactions for glucose, lactate, and uric acid detection, along with a pH-sensitive dye, and the color changes are analyzed through smartphone imaging in HSB and RGB color spaces. This work establishes a scalable, low-cost, and robust paper-based platform that integrates multiplexed colorimetric sensing with sweat rate measurement, providing a practical and disposable solution for wearable non-invasive health monitoring. Similarly focused on basic sensing elements, but moving towards a more complex soft form factor, Kim et al. used PDMS doped with white pigment and specific colorimetric reagents to fabricate a soft, skin-interfaced microfluidic platform for the colorimetric detection of nutrients in sweat, enabling non-invasive nutritional monitoring and transdermal nutrient delivery [Figure 4B]^[89]. The device features deep microreservoirs to enhance optical path length and sensitivity, and integrated passive CBVs to direct sweat flow and prevent cross-contamination. The quantitative analysis was achieved via smartphone RGB imaging, validated against laboratory methods. A key innovation is the demonstration of a simple, electronics-free, and cost-effective colorimetric platform for multiplexed nutrient sensing, highlighting the potential of sweat-based microfluidics for personalized nutrition management and at-home health monitoring.

Figure 4. (A) Schematic layout of the wearable colorimetric sensing patch and depiction of sweat flow directionality and relative flow rates within the patch. Reproduced with permission^[87]. Copyright 2024, Wiley-VCH; (B) Schematic illustration of the microfluidic sensor and nutrient patch in an exploded configuration: the microchamber architecture for colorimetric analysis, the handheld device photograph, and the optical image of the assembled system situated next to a vitamin tablet. Reproduced with permission^[89]. Copyright 2024, Wiley; (C) Exploded view of a skin-interfacing microfluidic system for personalized sweat-rate and sweat-chloride analysis: patterned microfluidic layer and embedded sensing layer. Reproduced with permission^[76]. Copyright 2020, Science; (D) Overall workflow of the deep-learning-assisted programmable chromatic chip for evaluating sweat glucose, pH, and lactate, including an intelligent imaging device and interpretable convolutional neural network. Reproduced with permission^[86]. Copyright 2022, American Chemical Society.

Building on these basic sensing platforms, considerable efforts have been made to facilitate complex physiological monitoring in the real world and in ambulatory settings. For example, Baker et al. employed multilayer thin-film polymers to develop a flexible, skin-interfaced wearable microfluidic system for the real-time monitoring of regional sweating rate and chloride content in sweat in sports science applications [Figure 4C]^[76]. The device consisted of two primary microchannels: Microchannel 1 incorporated an orange dye that dissolved upon sweat entry for volumetric measurement, while Microchannel 2 was pre-loaded with a chloride-sensitive assay which produced a purple color with intensity proportional to sweat chloride concentration. A smartphone application captured images for automatic analysis. This work represents an innovation by seamlessly integrating a low-cost, manufacturable microfluidic platform with ubiquitous smartphone capabilities for in-situ colorimetric analysis. It successfully transitions sweat analytics from laboratory environments to real-world applications and serves as a foundational model for field-deployable optical sweat sensors. Advancing beyond traditional colorimetric methods, the incorporation of artificial intelligence (AI) has opened new avenues for highly accurate and automated data interpretation. Liu et al. utilized sodium alginate-calcium chloride gel capsules immobilized with various enzymatic/chemical indicators to fabricate a programmable colorimetric sensor, which was coupled with a deep learning algorithm designed for the precise distinguishing and quantification of pH levels, glucose, and lactate in human sweat [Figure 4D]^[86]. The gel capsules functioned as programmable sensing pixels containing specific reagents that induced distinct color changes. A convolutional neural network (CNN) model was trained on an extensive dataset of colorimetric images, achieving 100% accuracy and a Class Activation Mapping (CAM) analysis was applied to demonstrate the inner working mechanism of CNN operation^[86]. By integrating the customizable colorimetric platform with the interpretable deep learning framework, this end-to-end strategy becomes an example of AI to overcome the limitations of traditional technologies and provides an important theoretical reference for future sensor design.

The development of colorimetric sensing is further illustrated by its innovative integration with different sensing modes and readout mechanisms to create a multi-functional system. For instance, a wireless, skin-interfaced platform was developed by integrating a thermal flow sensor with a straight PDMS fluid passage and a Bluetooth Low Energy system-on-chip, for real-time analysis of sweat flow rate and cumulative loss. This platform allows for optional integration into advanced microfluidic systems designed for the colorimetric sensing of pH, chloride, creatinine, and glucose^[92]. Operating on a non-contact thermal sensing principle, the platform was engineered to be reversibly integrated with a microfluidic module containing reservoirs pre-loaded with colorimetric assays for sequential analysis. By integrating wireless digital readout of sweat parameters and advanced colorimetric chemistry into one device, this multifunctional, multimodal platform signifies a major advance in the development of fully integrated “lab-on-skin” systems.

While colorimetric sensors offer significant advantages such as suitability, low cost, and simplicity for in situ and multiplexed analyses, they face several challenges. These include the inability to perform real-time continuous monitoring, a typically semi-quantitative nature, susceptibility to environmental interference, and limitations in accuracy due to factors such as uneven color development and signal dependence on both concentration and sample volume^{[50,60,77,86,88]}. Colorimetric sensing provides a simple, low-cost, and electronics-free pathway for wearable sweat analysis, enabling immediate perceptibility and user-friendly operation^[93]. However, compared to more complex optical techniques such as fluorescence or SERS, it trades some quantitative accuracy and temporal resolution for unmatched simplicity and cost-effectiveness. This makes it particularly ideal for disposable, single-use patches aimed at periodic health screening or personalized hydration and nutrition management, where continuous data streaming is not the primary requirement^[76,85,89].

SERS seznsors

SERS technology offers higher specificity and multiplexing capabilities by acquiring the unique fingerprint information of molecules. SERS utilizes the localized surface plasmon resonance (LSPR) effect on the surface of noble metal particles and the gaps between them to enhance the Raman scattering signal of adsorbed molecules by approximately 6 to 9 orders of magnitude, even down to the single-molecule level^[67,70,94]. Compared to the colorimetric method, SERS provides a direct molecular identification without the need for chemical markers or enzymatic reactions and is less likely to interfere with complex sweat matrices^[94,95]. The basic principle of SERS relies on the enhancement of the Raman signal, generated by specific molecular vibrational/rotational modes through the high local electric field produced by LSPR. This enables label-free and highly specific identification of target analytes [Figure 3B]^[65,67,96]. The technique offers advantages such as ultra-high sensitivity, resistance to photobleaching, minimal interference from water, and the ability to analyze multiple components simultaneously in a single measurement^{[55,94,95,97]}. With the development of flexible substrates and portable Raman spectrometers, SERS is now being incorporated into wearable devices for noninvasive, real-time analysis of biofluids^[95,98]. When evaluated within the comparative framework established for colorimetry, SERS reveals distinct trade-offs: it offers higher specificity and multiplexing capability, but its on-board reliability is questioned due to the need for a uniform and mechanically stable plasmonic substrate, and its integration often relies on external optical equipment. For effective wearable integration, SERS substrates must not only provide dense and uniform plasmonic hotspots for signal enhancement but also maintain mechanical compliance and stability under skin deformation^[71,94]. This is often achieved through flexible materials such as paper, textiles, and nanostructured polymers, which can conform to skin curvature and withstand repeated bending^[66,70]. Furthermore, efficient sweat collection and delivery are essential for reliable SERS-based sensing. Microfluidic architectures are frequently incorporated to enable controlled fluid transport, prevent cross-contamination between samples, and facilitate sequential or time-resolved sampling^[67,99]. Nevertheless, the realization of robust in vivo SERS sensing continues to face significant challenges, including the scalable fabrication of homogeneous nanostructures, preservation of signal fidelity under mechanical strain, and mitigation of interference from the complex biochemical composition of sweat^[70,99]. Based on these fundamental considerations, innovative designs have been developed to translate SERS into practical wearable formats.

Building on these intrinsic advantages, SERS has been effectively adapted into immunoassays for sensitive biomarker detection. For example, a SERS-based immunoassay (SERSI) was developed to analyze cortisol in biological fluids using Au NRs as SERS nanotags and gold-coated glass slides or magnetic beads as capture substrates functionalized with cortisol antibodies [Figure 5A]^[78]. In the magnetically-assisted SERSI modality, competition between free cortisol and SERS nanotags for binding sites on antibody-decorated magnetic beads resulted in an increase in the SERS signal in the supernatant, which is proportional to cortisol concentration. This enabled quantification across a linear range of 12.5-400 ng/mL, with a detection limit (LOD) of 7 ng/mL. This method demonstrated high selectivity against structurally similar hormones, excellent repeatability, and was successfully validated against standard techniques. This work presents a highly reliable and sensitive SERS immunoassay platform, highlighting its potential as an effective platform for in situ and real-time detection of cortisol-related disorders and human stress levels, and establishing a crucial methodological foundation for subsequent SERS biosensing development. To transition toward wearable applications, research has focused on integrating SERS with flexible, low-cost substrates. Mogera et al. employed Au NRs immobilized on chromatography paper integrated with a flexible microfluidic system and SERS to fabricate a wearable sensor for continuous, label-free quantification of uric acid in sweat [Figure 5B]^[67]. The paper microfluidic layer enabled efficient sweat collection and transport via capillary action while simultaneously quantifying sweat loss and sweat rate, and the gold nanostars served as stable plasmonic substrates that provided significant SERS enhancement for highly sensitive (LOD: 1 µM) and specific uric acid detection using a ratiometric approach. This work innovatively combines label-free SERS detection with paper-based microfluidics in a flexible and stretchable wearable format, effectively addressing the environmental instability of traditional enzymatic/antibody-based sensors and demonstrating considerable potential for advancing reliable, continuous metabolite monitoring in point-of-care health applications.

Figure 5. (A) Illustration of two SERS immunoassay approaches: magnetic-bead suspensions bearing cortisol antibodies for magnetically assisted detection, and gold-coated glass substrates functionalized with cortisol antibodies for rigid-support detection. Reproduced with permission^[78]. Copyright 2020, Elsevier; (B) Conceptual, overhead, and stacked representations of the wearable plasmonic paper microfluidic system for sweat collection, storage, and in situ SERS measurement. Reproduced with permission^[67]. Copyright 2022, Science; (C) Schematic illustration depicting the fabrication process of the Au NSs/ion-track-etched polycarbonate membrane-based SERS substrate and its subsequent detection procedure. Reproduced with permission^[94]. Copyright 2023, American Chemical Society; (D) Schematic diagram of the stacked structure of functional layers, an integrated microfluidic plasmonic device enabling biofluid flow, storage and SERS analysis, alongside a system-level block diagram of the internal functional modules within the portable Raman analyzer. Reproduced with permission^[99]. Copyright 2022, Nature Publishing Group; (E) Schematic illustration of the fabrication and initiation of the flexible nanoplasmonic SERS patch comprising plasmonic fluid-channel layer and dermal-contact layer. Reproduced with permission^[71]. Copyright 2025, Nature Publishing Group; (F) Schematic diagram of the sweat collection principle of Janus-based wearable textiles and the generation of inverse Raman signals by microstructured optical fiber cores. Reproduced with permission^[66]. Copyright 2023, American Chemical Society. 3D: 3-dimensional; SERS: surface-enhanced Raman scattering; PDMS: polydimethylsiloxane.

In addition to disposable paper-based systems, enhancing the performance and reusability of SERS substrates has emerged as a key focus area for sustainable wearable sensing technologies. Gui et al. fabricated a flexible and nanoporous SERS substrate by synthesizing Au NRs in situ on an ion-track-etched polycarbonate (PC) membrane aimed at facilitating sweat enrichment and metabolite analysis [Figure 5C]^[94]. The nanoporous structure of the PC membrane enabled efficient analyte enrichment, while the Au NRs provided dense and uniform plasmonic hotspots for significant Raman signal enhancement. The substrate showed highly reproducible SERS signals for methylene blue, along with excellent mechanical flexibility (withstanding 2,000 bending cycles) and long-term stability. This work showcases the integration of high-performance SERS substrates with flexible porous membranes, providing a material design strategy for achieving stable and reproducible non-invasive sweat monitoring. The evolution of SERS sensing further extends to sophisticated system integration with microfluidics and intelligent data processing for dynamic and multi-parameter monitoring. For example, He et al. used Ag nanomushroom arrays integrated into a PDMS microfluidic platform to fabricate a wearable SERS sensing device for portable and refreshable detection of sweat analytes such as pH, lactate, and urea [Figure 5D]^[99]. This work integrated epidermal microfluidics with portable SERS analysis, offering a reusable, high-resolution sensing platform for personalized healthcare, and highlights the potential of microfluidic SERS systems for continuous, non-invasive sweat analysis.

Advancing beyond traditional sensing paradigms, the incorporation of chronological sampling and machine learning has opened new avenues for intelligent metabolic monitoring. Jeon et al. developed an all-flexible chronoepifluidic nanoplasmonic patch that integrated plasmonic nanostructures within a microfluidic sequential sampler for label-free and chronological analysis of sweat biomarkers [Figure 5E]^[71]. The sensor featured plasmonic Ag nanoislands formed on an ultrathin fluorocarbon-coated PDMS membrane via low-temperature solid-state dewetting, creating large-area, uniform electromagnetic hotspots. The microfluidic sampler utilized CBVs to collect and spatially separate sweat samples over time. Combined with machine learning, the patch enabled the quantitative analysis of lactate, uric acid, and tyrosine, and successfully captured dynamic metabolic changes induced by exercise and dietary intake. This work represents a significant advancement in manufacturing high-performance nanostructures on flexible substrates and provides new ideas for addressing signal stability issues in continuous, long-term monitoring using wearable SERS sensors. Innovation in SERS wearable technology also extends to novel architectures that decouple fluid handling from detection. Han et al. combined a Janus textile for unidirectional sweat collection with a grapefruit microstructured optical fiber decorated with Ag nanoparticles as a SERS probe for sweat analysis [Figure 5F]^[66]. The Janus textile features a superhydrophobic skin side and patterned superhydrophilic zones for directional sweat transport, while the optical fiber extracts nanoliter sweat volumes via capillary action and provides length-distributed plasmonic hotspots for SERS enhancement. The system detected urea and sodium lactate at very low concentrations with high uniformity and multiplexing capability, illustrating a creative merger of textile-based microfluidics and optical SERS for practical wearable operation.

Despite the successful integration of SERS with flexible platforms, there are still challenges in translating this technology into robust and widely applicable wearable systems. The main obstacle is the difficulty of producing scalable SERS substrates with dense, homogeneous and mechanically stable plasmonic hotspots^[71,94]. Signal reproducibility can be compromised by the oxidation of silver-based nanomaterials, variations in nanoparticle spacing under skin-induced strain, and interference from competing molecules present in sweat^[66,70,99]. For system integration, SERS sensing requires external optical excitation and spectroscopic analysis. The current reliance on portable Raman spectrometers imposes constraints in terms of device cost, overall size and easy integration in a separate, low-profile patch^[95,98]. In addition, detection efficiency may be limited in individuals with low sweating capacity and long-term considerations regarding the biocompatibility of nanomaterials and laser safety require careful consideration^[88,97]. Thus, SERS is particularly suited for applications demanding high-specificity, multi-analyte profiling, such as metabolic monitoring, hormone detection and drug screening, although its wide use in wearable formats is dependent on progress in portable spectroscopy and the stability of substrates^[95,98].

Fluorescence sensors

Fluorescence-based sensing is another powerful optical method for wearable sweat analysis, being distinguished by its high sensitivity, suitability for quantitative and multiplexed detection, and inherent compatibility with flexible, skin-interfaced platforms^[100,101]. The basic principle is to detect fluorescence signal changes, such as intensity, emission wavelength or lifetime, emitted by chemical probes when they are activated by light and when they interact with target analytes^[100,102]. Common fluorescence detection mechanisms include fluorescence quenching (e.g., dynamic quenching of quinine sulfate by chloride ions)^[79], fluorescence enhancement (e.g., sodium binding to crown ether derivatives)^[102], wavelength shifts, and changes in fluorescence lifetime^[79,102]. Energy transfer mechanisms such as Förster resonance energy transfer (FRET) and the inner filter effect (IFE), as well as enzyme-coupled reactions (e.g., glucose sensing via H₂O₂-mediated quenching of fluorescein)^[79], are also used to achieve high specificity and multiplexing capability for metabolites such as glucose, lactate, chloride, and pH^[54,79]. Compared to colorimetric approaches, fluorometric approaches offer higher intrinsic sensitivity for multiplexing, which results in lower LODs and broader linear ranges for different analytes [Table 1]^[79,100].

A prime example of this integrated approach is the smart wearable sweat patch developed by Ardalan et al., which utilized cellulose-based paper discs and cotton-thread microfluidics for the fluorimetric sensing of glucose, lactate, pH, chloride, and sweat volume [Figure 6A]^[79]. Each biomarker was measured via a distinct fluorescent mechanism, and a custom 3D-printed smartphone module with a UV-LED and optical filter captured the signals for quantification via a dedicated app. This work demonstrated a low-cost (~ $0.03 per patch), disposable platform, underscoring the potential of combining simple microfluidics with smartphone fluorimetry for accessible multi-analyte sensing. Extending the concept of multiplexed ion detection, Sekine et al. integrated specific fluorescent probes for chloride, sodium, and zinc into a soft microfluidic device for sequential sweat analysis during exercise [Figure 6B]^[102]. The device used CBVs to direct sweat into pre-loaded assay reservoirs. A smartphone attachment with appropriate optical filters enabled fluorescence imaging, with signals normalized to an internal fluorescent reference (Rhodamine 110). This system achieved accuracy comparable to standard laboratory techniques in human studies, establishing a practical, low-cost platform for fluorometric sweat ion analysis.

Figure 6. (A) Digital photograph of the IoT-integrated cellulose microfluidic fluorescent patch constructed from medical tape, cotton thread, and paper. Reproduced with permission^[79]. Copyright 2020, Elsevier; (B) Schematic of a microfluidic platform with in situ fluorescence analysis: collection and retention of chloride, sodium, and zinc with corresponding chemical-probe fluorescence signals. Reproduced with permission^[102]. Copyright 2018, The Royal Society of Chemistry; (C) Schematic representation of a soft microfluidic device for fluorescent measurement of amino acid concentrations in sweat via enzymatic reactions. Reproduced with permission^[103]. Copyright 2025, The Royal Society of Chemistry; (D) Schematic illustration of the fabrication of a wearable hydrogel sensing patch embedded with upconversion optical probes and its multichromatic response to urea by inner-filter effect. Reproduced with permission^[54]. Copyright 2023, American Chemical Society; (E) Schematic representation of the process of luminescent silicon-loaded carbon quantum dots and bimetallic nanoparticles for red-to-blue fluorescence detection of glucose in sweat. Reproduced with permission^[104]. Copyright 2020, American Chemical Society; (F) Schematic diagram of the principle of Tb-DPA MOF-based detection of human sweat pH. Reproduced with permission^[105]. Copyright 2024, Elsevier. CQD: Carbon quantum dot; DPA: 2,6-pyridinedicarboxylic acid; IoT: internet of things; MOF: metal-organic framework.

In addition to ions, fluorescence is particularly effective for monitoring metabolites via enzymatic pathways. Cho et al. employed smartphone-linked enzymatic fluorometric assays (for total amino acid, alanine, and lysine) incorporated into a skin-interfaced, flexible microfluidic patch, to investigate amino acid loss in eccrine sweat during exercise [Figure 6C]^[103]. The device featured grouped microreservoirs filled via CBVs for sequential sampling. The assays, based on coupled enzymatic reactions producing a fluorescent signal (excitation/emission: 535/587 nm), exhibited high sensitivity, low LODs, and linear responses across physiological ranges. This system provided a cost-effective, non-invasive method for real-time monitoring of amino acid turnover, emphasizing sweat’s role in metabolite loss and the importance of dietary management during prolonged exercise. In a different optical design focused on a single metabolite, Hu et al. developed a wearable hydrogel patch embedding an UCNPs functionalized with p-DMAC for urea sensing based on the inner inner filter effect (IFE) [Figure 6D]^[54]. The reaction between urea and p-DMAC under acidic conditions generated a product that absorbed the green emission of UCNPs, causing a ratiometric color shift from yellow to red under 980 nm near infrared (NIR) excitation. This NIR excitation strategy minimized background autofluorescence from skin and ambient light. A portable 3D-printed reader integrating a smartphone, a miniaturized laser, and an optical filter enabled in situ quantitative analysis, demonstrating a robust approach for anti-interference monitoring in complex environments.

The pursuit of higher accuracy has driven the development of advanced ratiometric fluorescence systems. For example, Cui et al. fabricated a wearable skin pad using luminescent porous silicon (PSi) particles loaded with carbon quantum dots (CQDs) and decorated with bimetallic (Au/Ag) nanoparticles (BiM), co-immobilized with glucose oxidase (GOX) in a chitosan film on a PU membrane, for the noninvasive and real-time visual detection of sweat glucose [Figure 6E]^[104]. In this ratiometric fluorescent nanohybrid (BiM-CQDs@PSi), the PSi acts as a red-emissive, oxidizable carrier, CQDs serve as the blue-emissive donor, and BiM enhances both the initial fluorescence of PSi and catalyzes its H₂O₂-triggered oxidation. The working principle relies on an initial FRET from CQDs to PSi, quenching blue emission and showing red fluorescence. Upon contact with sweat glucose, GOX-generated H₂O₂ oxidizes PSi, weakening FRET, decaying red fluorescence, and recovering blue emission, resulting in a visible red-to-blue color shift. The transition kinetics are proportional to glucose concentration, enabling semiquantitative analysis via smartphone imaging and RGB analysis, which was successfully demonstrated in clinical tests to distinguish diabetic hyperglycemia from normal levels. Similarly, for pH monitoring, Chen et al. developed a skin-attachable ratiometric sensor using a dual-emissive Tb-based metal-organic framework (Tb-DPA MOF, where DPA = 2,6-pyridinedicarboxylic acid)^[105]. The H⁺-induced collapse of the MOF structure differentially affected the emissions of the ligand (DPA at 380 nm) and the Tb³⁺ center (544 nm), providing a stable and reversible fluorescence response across the physiologically relevant sweat pH range (3.50-6.70). The sensor demonstrated excellent long-term stability and was successfully used for real-time sweat pH monitoring during exercise.

Fluorescence sensing occupies a distinct niche in the wearable optical sensing landscape. However, translating fluorescence into robust, widely deployable wearables faces challenges, including the inability for real-time continuous monitoring in some endpoint-based designs^[79], potential temperature sensitivity affecting enzyme activity or assay stability^[79,103], and limited detection sensitivity for certain analytes such as zinc^[102]. Signal integrity can be compromised by photobleaching risks requiring light-shielding measures^[102], as well as interference from autofluorescence and background signals, especially for short-wavelength-excited probes^[100]. Furthermore, the irreversible and non-reusable nature of most assays necessitates single-use formats^[103]. Fluorescence is particularly well-suited for applications demanding sensitive, quantitative, and multiplexed tracking of key metabolites and electrolytes, especially when integrated with advanced strategies such as NIR excitation to minimize background noise^[54]. Future advancements focusing on improved stability, interference resistance, and continuous monitoring capability will be key to developing more robust fluorescence-based wearable systems.

ECL sensors

ECL is emerging as a promising but nascent optical sensing technique that uniquely combines electrochemical and luminescent processes for highly sensitive bio-sensing applications. Unlike other optical methods, ECL operates without an external light source and offers intrinsic low background, high sensitivity, and precise spatiotemporal control of light emission^[106]. The basic principle of ECL involves the generation of excited-state luminophores by exergonic electron transfer reactions at the surface of the electrode, which then relax and emit photons [Figure 7A]^[107]. This process can be initiated through various mechanisms. One commonly used system involves coreactant pathways, such as the classic tris(2,2′-bipyridine)ruthenium(II) ([Ru(bpy)₃]²⁺)/trialkylamine pair, in which amine-containing analytes participate in a redox cycle to regenerate the excited state of [Ru(bpy)₃]²⁺. Another common approach is the enzyme-coupled ECL, where enzymatic reactions generate coreactants such as nicotinamide adenine dinucleotide (NADH) or H₂O₂ that interact with ECL luminophores (e.g., [Ru(bpy)₃]²⁺ or luminol) to produce light^[108,109]. From a performance point of view, ECL presents several compelling advantages including high sensitivity (e.g., LODs down to pM levels for lactate) [Table 1], wide dynamic range, and minimal interference from ambient light due to its self-emissive nature^{[106,107,110]}.

Figure 7. (A) Schematic diagrams of a stainless steel electrode for ECL detection using the luminol/H₂O₂ system and the mechanism of photoinduced ECL at silicon electrodes protected by nanoscale SiO_X and Ni-stabilizing thin films. Reproduced with permission^[107]. Copyright 2021, American Chemical Society; (B) Schematic illustration of the Ru-PEI@SiO₂ synthesis and the assembly route of the flexible MIP-ECL skin sensor: electrochemical conversion, surface immobilization, MIP layer imprinting, sensor elution, and subsequent detection of epidermal analytes. Reproduced with permission^[106]. Copyright 2019, The Royal Society of Chemistry; (C) Schematic diagram of neurotransmitter monitoring using a stretchable photonic-crystal-hydrogel-enhanced ECL platform, together with machine learning-assisted detection and classification. Reproduced with permission^[108]. Copyright 2025, American Chemical Society; (D) Schematic depiction of rGO@hydrogel used for monitoring motion behavior and evaluating multiple physiological indicators in sweat, including urea, lactate, and chloride. Reproduced with permission^[111]. Copyright 2023, Elsevier. ECL: Electrochemiluminescence; TEOS: tetraethyl orthosilicate; PAM: polyacrylamide; PVA: polyvinyl alcohol; O-β-CD: oxidized β-cyclodextrin; c-BPE: closed bipolar electrode.

Integrating ECL into wearable devices demands flexible and robust architectures that can maintain electrical and optical performance under deformation. Pioneering designs have focused on merging ECL-active components with soft, stretchable materials such as flexible conductive substrates (e.g., Au nanotube networks, carbon textiles) functionalized with ECL luminophores^[106]. Hydrogels are particularly attractive as they can serve dual roles as sweat-absorbing reservoirs and ECL electrode matrices^[111]. ECL presents several compelling advantages for wearable sweat analysis. It combines the high sensitivity, wide dynamic range, and ease of multiplexed analysis inherent to chemiluminescence with the good stability and tunability of electrochemical methods^[107]. Crucially, its operation without an external light source eliminates ambient light interference and simplifies device architecture. Compared to traditional electrochemical sensors, ECL offers superior sensitivity for non-electroactive analytes and minimizes interference from electroactive species in complex biofluids such as sweat^[106]. These attributes, coupled with the potential for simple, low-cost instrumentation, make ECL a promising candidate for next-generation wearable diagnostics^[110,112].

A pioneering example is the flexible ECL platform developed by Chen et al. for wearable sweat analysis [Figure 7B]^[106]. They immobilized highly luminescent ruthenium-polyethyleneimine (Ru-PEI@SiO₂) nanospheres on stretchable gold nanotube networks, followed by a coating of an elastic molecularly imprinted polymer (MIP) for effective detection. This design enabled in situ sampling and quantification of urea and lactate in human sweat with high selectivity and excellent mechanical stability, achieving LODs of 15.0 μM and 16.7 μM, respectively. This study translated the ECL mechanism into a practical wearable format, establishing a novel paradigm for non-invasive metabolic sensing. Advancing the integration of ECL with flexible optics, other researchers have developed composite hydrogel electrodes for enhanced performance. Chen et al. designed an ECL electrode using a stretchable photonic-crystal hydrogel (PCH), which synergistically combined a polyacrylamide (PAM-PVA)-(O-β-CD)-Gel conductive hydrogel with a poly latex-sphere photonic crystal nanomatrix for enhanced neurotransmitter detection in sweat [Figure 7C]^[108]. The mechanochromic PCH allowed dynamic tuning of its photonic bandgap under strain to match the emission wavelength of the [Ru(bpy)₃]²⁺ luminophore, resulting in a signal amplification of up to 9-fold. This platform achieved simultaneous qualitative and quantitative analysis of 11 different neurotransmitters with 100% classification accuracy in artificial sweat and serum, leveraging linear discriminant analysis of the ECL responses. This work represents a significant innovation by merging stretchable optics with ECL sensing, creating a versatile and powerful platform for multiplexed biomarker detection in wearable healthcare applications.

Beyond metabolite sensing, ECL platforms are being engineered to combine biochemical analysis with physical sensing for holistic health monitoring. Hu et al. utilized rGO- and luminol (Lum)-decorated self-healing hydrogels to construct a sweat sensor incorporating a closed bipolar electrode (c-BPE) sweat detection unit and a strain sensing unit for simultaneous motion monitoring and multiplexed sweat biomarker analysis such as chloride ions, lactic acid, and urea [Figure 7D]^[111]. In this platform, the rGO@hydrogel serves as a strain sensor with excellent stretchability (up to 400%), adhesiveness to skin, and stable electrical responses to body movements, while its porous 3D structure enables in-situ sweat collection and storage; subsequently, it spontaneously self-heals with Lum@hydrogel to form a c-BPE system, where the rGO@hydrogel (cathode) containing sweat analytes and the Lum@hydrogel (anode) enable ECL detection through anode signal fluctuations correlated to cathode reactions, achieving accurate quantification without direct sample-reagent contact. The platform demonstrated rapid response to physical activities, high stability, prolonged luminescence duration, and reliable on-body detection of sweat components, validated against commercial kits. Complementing these integrated systems, enzymatic ECL biosensors highlight the technique’s potential for extreme sensitivity. Cai et al. developed such a biosensor by co-immobilizing pyruvate oxidase (PYOD) and dehydrogenase (LDH) with their cofactor nicotinamide adenine dinucleotide (NAD⁺) on a carbon nanotube-modified electrode, using luminol as the emitter^[110]. The enzymatic oxidation of lactate produced hydrogen peroxide, which strongly enhanced the luminol ECL signal. This coupling yielded an ultra-sensitive lactate sensor with a remarkably wide linear range and a LOD down to 8.9 × 10^-12 mol/L, successfully applied to track lactate dynamics in athlete sweat during exercise.

ECL faces several challenges that need to be addressed to achieve widespread adoption of the wearable. The main obstacles are the instability of luminescence on flexible substrates and the incompatibility of traditional rigid electrodes with conforming skin contact^[108]. High operating voltages in some systems raise concerns regarding power consumption and on-body safety^[113]. The integration of multiple functional units (e.g., sensing, fluidics, and electronics) on a single flexible substrate often increases system complexity and can degrade overall performance^[111]. In addition, ECL signals can be susceptible to interference from co-existing species in real sweat samples^[110].

Other optical sensors

In addition to colorimetric, SERS, fluorescence, and ECL methods, optical fiber-based sensing represents a distinct and promising approach for wearable sweat analysis, operating on fundamentally different principles. The core mechanism relies on light guidance via total internal reflection within the optical fiber, where a portion of the electromagnetic field extends as an evanescent wave at the core-cladding interface. By modifying the fiber through tapering, side-polishing, or cladding removal, this evanescent field can be exposed to interact with the surrounding sweat. Changes in the local refractive index due to biomarkers modulate the guided light’s characteristics (intensity, wavelength, phase), enabling sensitive detection^[114,115]. Diverse sensing architectures are built on this foundation, including Mach-Zehnder interferometers, surface plasmon resonance structures, and Fiber Bragg Gratings (FBG), each transducing specific physical or chemical changes into optical signals^[115-117]. Hybrid designs, such as combining a semi-distributed interferometer (SDI) with an FBG, allow simultaneous, decoupled monitoring of sweat biomarkers and skin temperature, effectively eliminating cross-sensitivity^[118].

Optical fiber sensors offer unique advantages for wearability, including complete immunity to electromagnetic interference, excellent biocompatibility, and minimal invasiveness due to their small diameter (~ 125 µm). Their flexibility allows seamless integration into textiles, garments, or flexible substrates, enabling conformal skin contact without compromising comfort^[119]. Performance-wise, they achieve high sensitivity, fast response, and low cross-sensitivity. For instance, a fiber-optic pH sensor demonstrated a sensitivity of -27.06 nm/pH with a response time of 15-20 s and negligible temperature cross-sensitivity^[117]. A surface plasmon resonance (SPR)-based cholesterol sensor achieved a LOD as low as 0.0051 µM^[116]. Integration strategies include embedding fibers in soft polymers such as PDMS, weaving them into fabrics, or incorporating them into functional hydrogels. Readout systems are increasingly miniaturized, employing LEDs, photodiodes, and microcontrollers, with data transmitted wirelessly via Bluetooth. For simpler outputs, smartphone-based colorimetric or intensity analysis combined with machine learning enables practical on-body use^[119,120].

Advancing towards multifunctional integration, Zhang et al. fabricated a wearable hydrogel optical fiber sensor via photopolymerization molding for simultaneous human posture and sweat pH monitoring [Figure 8A]^[120]. The sensor features a polyethylene glycol diacrylate (PEGDA) hydrogel core and a calcium alginate hydrogel cladding. Posture sensing is achieved through bending-induced optical path changes, while sweat pH sensing relies on the fluorescence intensity variation of Rhodamine B. The sensor exhibits excellent biocompatibility, flexibility (withstanding 700 bending cycles), and low optical loss (0.6 dB·cm^-1), demonstrating the potential of multifunctional wearable optical fibers for integrated physiological monitoring. Building on this need for enhanced chemical detection, Lei et al. developed a tapered fiber Mach-Zehnder interferometer sensor functionalized with a chitosan/polyacrylic acid polyelectrolyte film for sweat pH detection^[117]. The film’s swelling degree varies with pH, altering its refractive index and causing a wavelength shift in the interference spectrum. The sensor achieved a high sensitivity of -27.06 nm/pH within the physiologically relevant sweat pH range of 5.0-7.5, with low temperature cross-sensitivity (0.012 nm/°C), showing a high-performance design for wearable metabolite detection.

Figure 8. (A) Schematic of a dual-function hydrogel optical fiber (HOF) sensor. The fiber features a step-index core-cladding structure doped with phenol red, enabling simultaneous monitoring of sweat pH and joint motion. Reproduced with permission^[120]. Copyright 2024, American Chemical Society; (B) Diagram of a reflective tapered optical fiber sensor for SERS-based lactate analysis. The fiber tip is chemically etched and decorated with Ag NPs to create high-density hotspots for rapid and sensitive Raman detection. Reproduced with permission^[121]. Copyright 2025, Elsevier; (C) Illustration of a fiber optic SPR biosensor for cholesterol detection. The sensing probe is fabricated via electroless silver plating on the fiber core and functionalized with cholesterol oxidase for specific enzymatic recognition. Reproduced with permission^[116]. Copyright 2025, Elsevier. APTES: (3-aminopropyl)triethoxysilane; β-CD: β-cyclodextrin; CHOL: cholesterol; SPR: surface plasmon resonance; EDC: 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide; NHS: N-hydroxysuccinimide; SERS: surface-enhanced Raman scattering.

Further advancing sensitivity and specificity for trace metabolite detection, Wang et al. fabricated a reflective tapered optical fiber SERS sensor using Ag NPs via chemical modification techniques for lactic acid detection in sweat [Figure 8B]^[121]. The tapered geometry enhances the evanescent field, while the nanoparticles provide strong electromagnetic enhancement. The sensor achieved a LOD of 2.67 × 10^-8 M for lactic acid and successfully detected lactic acid Raman peaks in real sweat samples, highlighting its potential for highly sensitive wearable Raman-based monitoring. Despite high sensitivity, real-world sweat monitoring often involves complex environments where temperature fluctuations and multiple interfering factors must be decoupled to ensure accurate metabolite quantification. To address such multi-parameter challenges, Karipbayeva et al. developed a SDI assisted by a fiber Bragg grating (FBG) for dual-parameter sensing^[118]. The SDI is sensitive to external refractive index (biomarker concentration), while the FBG provides precise temperature measurement. This combination enables low-cross-talk, simultaneous monitoring of sweat biomarkers (e.g., azelaic acid) and temperature, offering a robust platform for applications in complex physiological environments. Expanding the detection to low-concentration lipid metabolites in sweat, Zhang et al. fabricated a fiber-optic SPR cholesterol biosensor using a polydopamine and β-cyclodextrin functional film [Figure 8C]^[116]. The sensor operates via SPR wavelength shifts induced by cholesterol binding. It achieved a high sensitivity of 133.83 nm/µM and a LOD of 0.0051 µM, presenting a significant innovation for the non-invasive, label-free monitoring of trace cholesterol in sweat and sebum.

Despite the distinct advantages of fiber-optic sensing in terms of mechanical stability and operational simplicity, translating this technology into robust, wearable platforms presents several significant challenges. Key obstacles include the miniaturization of readout electronics, as commercial interrogators for FBGs or spectrometers remain bulky and power-intensive^[115]. The integration of brittle silica fibers with soft, flexible substrates poses risks of delamination and fracture under mechanical deformation, while the scalable fabrication of sensitive elements such as fiber tapers and gratings contributes to increased production costs^[116]. For implantable applications, ensuring long-term biocompatibility and resistance to biofouling necessitates advanced material engineering strategies^[119]. Future development will focus on novel materials such as hydrogel-based fibers, multifunctional sensor networks, and integration with AI and the internet of things (IoT) for intelligent health monitoring^[119]. Through continued innovation, fiber-optic sensing is poised to complement established optical methods, offering quantitative, real-time capability distinct from semi-quantitative colorimetry, yet with simpler instrumentation than SERS and greater stability than fluorescence or ECL systems.