Nanomaterial -enabled flexible sensors for continuous biomedical monitoring and closed-loop intervention

Abstract

With the increasing demand for biomedical monitoring transitioning from in vitro intermittent assays to continuous, in situ, and intelligent systems, flexible sensing technology has become a direct bridge between detection and intervention. This review systematically summarizes recent advances in nanomaterial-based flexible sensors for biomedical applications. It first introduces commonly used flexible substrates, including polymers, paper, textiles, and hydrogels, highlighting their mechanical compliance, electrical performance, and strategies for functional modification. Secondly, nanomaterials are classified by dimensionality, zero-dimensional, one-dimensional, and two-dimensional, to discuss their unique advantages in enhancing sensor sensitivity, selectivity, and stability. The review further explores the multi-tiered applications of flexible sensors, spanning surface-worn wearable devices, deep-tissue implantable devices, and intelligent closed-loop systems. At the same time, it addresses persistent challenges such as mechanical fatigue, biofouling, signal drift, and power limitations during long-term operation. Looking forward, the development of high-performance flexible sensors relies on the integration of advanced materials, edge artificial intelligence, and wireless technologies to enable high-fidelity, multimodal, and closed-loop diagnostic and therapeutic systems. This comprehensive overview provides a structured reference for the application of flexible nanomaterials in health monitoring, disease management, and clinical translation, underscoring the promising trajectory of flexible sensing toward intelligent, real-time medical intervention.

Introduction

Biomedical monitoring is transitioning from intermittent, in vitro testing methods toward continuous, in situ, and intelligent systems that integrate sensing with therapeutic intervention [1, 2]. Traditional in vitro assays such as ELISA and PCR, while highly accurate, are inherently limited in their capacity to capture dynamic changes and transient molecular events that underpin physiological or pathological processes [3, 4]. Flexible wearable and implantable devices have emerged to address this gap; however, their broader application remains constrained by signal acquisition accuracy and mechanical rigidity. Within this context, flexible sensors—by virtue of their excellent mechanical compliance, tunable interfaces, and highly sensitive signal transduction—have gradually become a pivotal technological bridge linking surface-based monitoring with in vivo sensing. More importantly, they are reshaping the focus of biomedical monitoring from purely diagnostic purposes toward integrated diagnostic-therapeutic regulation [5–8]. Pioneering work on ultrathin, skin-like electronic systems first established the feasibility of conformal, non-invasive monitoring [9], while subsequent integration of sensing with on-demand drug delivery represented an early milestone toward closed-loop therapeutic platforms [10, 11].

Flexible sensors maintain stable operation under complex deformation conditions, including bending, stretching, compression, and twisting. They offer superior mechanical compatibility with soft tissues (skin, blood vessels, and brain tissue), thereby minimizing signal attenuation and tissue damage caused by interface mismatch [12–14]. Flexible substrates are intrinsically non-conductive and lack electroactive sites. Introducing conductive fillers, such as micron-sized particles, typically demands high loading, which compromises softness and introduces interfacial issues during deformation, ultimately leading to signal hysteresis and drift [15]. In contrast, nanomaterials offer advantages for flexible sensing. Their high aspect ratios and surface areas enable conductive networks at low concentrations (e.g., < 0.1 wt%), thereby preserving substrate flexibility [16]. Their high surface atom ratio provides active sites, and their ready modifiability allows for selective detection [17]. Furthermore, enhancing interfacial bonding between nanomaterials and flexible substrates can reduce hysteresis in strain sensors, as demonstrated by a dopamine-treated CNT/fiber sensor that achieved a low hysteresis of 5% [18]. The utility of nanomaterials in flexible sensing was first established by early demonstrations of carbon nanotube-based stretchable conductors and graphene-based sensitive layers, which collectively confirmed that nanoscale fillers can simultaneously impart electrical conductivity and mechanical robustness to soft polymers [19].

Currently, flexible sensors are widely deployed in wearable devices such as patches and wristbands, where they continuously monitor physiological functions and body fluid indicators for health management and disease screening [20–22]. Nevertheless, these devices continue to face limitations in comfort and long-term stability arising from mechanical fatigue during prolonged wear, sweat interference, signal drift, and power transmission issues [23, 24]. Similarly, implantable flexible sensors can reach target tissues and organs, enabling high spatiotemporal resolution monitoring of physiological signals in organs such as the brain, eyes, and cardiovascular system [25]. However, the high humidity and salinity of the internal environment, combined with dynamic organ movement and immune rejection, frequently induce signal distortion and biofouling [26]. Balancing dynamic tissue adaptation with long-term stability remains a key challenge for clinical translation. More critically, flexible sensors are evolving beyond simple surface or in vivo monitoring toward closed-loop systems that integrate monitoring, analysis, and intervention (Fig. 1). Through biocompatible interfaces, wireless communication, and intelligent algorithms, flexible sensors hold promise for enabling real-time coordination between pathological signal detection and therapeutic interventions in applications such as neuromodulation and chronic disease management [27]. The feasibility of such closed-loop systems was advanced by early integration of flexible sensor arrays with machine learning for automated signal interpretation and decision-making [28]. Existing reviews have largely focused on specific sensing modalities (e.g., pressure/humidity/temperature, nanomaterial-enabled sensors, or wireless technologies), often without systematically addressing the evolutionary continuum from in vitro diagnostics to in vivo implantation and from passive diagnosis to closed-loop therapeutic regulation [29–31]. This review fills this gap by providing a unified framework centered on nanomaterial dimensionality.

Fig. 1.

Schematic diagram of integrated in vivo and in vitro flexible nano sensing, monitoring, analysis, and intervention

In line with this, the present review focuses on flexible nanosensors for biomedical monitoring. It begins by surveying commonly used flexible substrates, including polymer substrates, paper substrates, textiles, and hydrogels. Subsequently, we summarize key nanomaterials by dimensional classification—namely, zero-dimensional, one-dimensional, and two-dimensional nanomaterials. Finally, we discuss recent research progress on flexible sensors based on functional nanomaterials, ranging from surface-attached and implantable devices to closed-loop regulation, and further summarize current challenges and future trends. This review provides a comprehensive overview of functional nanomaterials, applications, challenges, and future directions in flexible sensors, with the aim of guiding the future development of nanomaterials in flexible electronics for everyday life and clinical practice. This review aims to provide a structured and up-to-date overview for researchers and clinicians interested in the translational potential of nanomaterial-based flexible sensors.

Flexible substrates

As a key component of a sensor, the substrate’s primary functions are to provide structural support, carry the active material, and connect and integrate other components. Since flexible substrates must withstand stresses such as bending and tension, materials like polymer films, paper, textiles, and hydrogels are typically used (Fig. 2) [32]. As shown in Table 1, various substrates exhibit distinct performance characteristics due to their inherent structures.

Fig. 2.

Different types of flexible substrates and nanomaterials used in flexible nanosensors

Table 1.

Characteristic comparison of typical substrate materials in flexible sensors

Substrate Material	Typical Material	Ultimate Elongation (%)	Cost	Key Limitations
Elastic polymers	PDMS	5—500	low	High O₂/water permeability
Paper	Cellulose paper	5–15	low	Poor stretchability; insufficient moisture resistance.
Textiles	cotton textile	7–10	low	Decreased conductivity due to moisture absorption
Hydrogels	PSPP	334	low	Durability problems arising from water evaporation

Polymer substrates are the most widely used in flexible sensors. Common types include polymer plastics such as PC, PI and PEN. These materials are widely adopted because they are compatible with state-of-the-art semiconductor processes. For example, Yang et al. used a carbon dioxide laser etcher to develop a microfluidic device made of polyimide (PI) for collecting sweat to detect biomarkers [33]. However, while these materials are flexible, they lack tensile strength. Polymer elastomeric substrates such as PDMS, PU, SEBS, and TPU offer excellent stretchability. Some elastomers can withstand dynamic strains exceeding 100% and remain durable, with their performance remaining unchanged even after thousands of cycles [34]. PDMS is a widely studied elastomeric material. Its combination with nanomaterials such as gold nanotubes is commonly used to fabricate stretchable electrodes, and its excellent stretchability and biocompatibility enable real-time monitoring of cellular physiological activity under mechanical stimuli such as stretching [35]. However, due to their long chains and larger pore sizes, elastic substrates are generally permeable to small molecules such as oxygen and water vapor [36]. For example, PDMS is permeable to small lipophilic molecules, organic solvents and water, which may penetrate the surface of flexible devices and thereby reduce their performance and sensitivity. This drawback is typically addressed by applying a multi-layer coating process, whereby the device is protected by the additional coating [32].

As flexible substrates, paper-based materials offer several advantages, including low cost, wide availability, biocompatibility, and large surface area. More importantly, paper provides capillary action, which allows it to transport aqueous liquids without the need for an external mechanical pump, making it an ideal choice for microfluidic designs [37]. By stacking multiple layers of paper patterned with microfluidic channels and ensuring that the channels between adjacent layers are interconnected, it is possible to build three-dimensional paper-based microfluidic devices with complex functions. Cao et al. used photolithography and screen printing techniques to construct a novel two-layer paper-based microfluidic screen-printed electrode [38]. They immobilized glucose oxidase on an aldehyde-functionalized paper substrate and employed graphene/Prussian blue nanocomposite for hydrogen peroxide sensing, enabling glucose measurement in sweat and blood [38]. However, the rough and porous surface of paper substrates also poses numerous challenges for the fabrication of high-quality two-dimensional thin-film materials and the application of novel two-dimensional materials such as black phosphorus, boron nitride and MXenes; these challenges are expected to be addressed by optimising the surface properties of the paper and exploring new material preparation techniques [39].

Fabrics have recently garnered increasing interest as candidates for flexible substrates due to their high flexibility, stretchability [40], breathability, and comfort [41]. Due to their structural properties, textiles are also highly favored in microfluidics. There are two main design approaches: one leverages the inherent moisture-wicking properties of cotton fibers to transport biological fluids to the sensing area, followed by chemical modification of the fibers; the other involves directly integrating sensing units into the fabric [37]. However, the fiber structure may also absorb moisture, which reduces the sensor’s electrical conductivity. Therefore, developing water-resistant fabric sensors is important. In addition, self-powered textile substrates are a class of novel functional materials attracting significant attention. By combining textile technology with energy harvesting and sensing, researchers have developed textiles that serve as sustainable power supplies [42]. For example, Cheng et al. prepared a multilayer, all-nanofiber Janus textile. After modifying the triboelectric layer with fluorinated polyurethane, the triboelectric output reached 356 V, 2.88 µA, 80.12 nC, and 2.49 Wm^− 2. This output can power small electronic devices via a rectifier circuit and, when combined with deep learning and circuitry, enables material recognition and gesture mimicry.

Hydrogels are gaining prominence in flexible electronics due to their excellent biocompatibility, versatility, and tunable mechanical properties. The performance of conductive nanocomposite hydrogel sensors depends on the nanofillers and the polymer network structure, which combines the characteristics of the nanofillers with the advantages of the polymer network. For example, Wu et al. developed a 3D-printed polyvinyl alcohol (PVA)/ κ-carrageenan (κ-CA)/carbon nanotube (CNT) hydrogel composite (PCC) [43]. The hydrogel matrix confers excellent mechanical properties (tensile strength of 633 kPa and elastic modulus of 243 kPa) while demonstrating outstanding decoding accuracy and information transmission rates, making it highly promising for applications in brain-computer interfaces and wearable devices. The enhancement of hydrogel substrate performance currently faces a number of challenges. On the one hand, to achieve durability, the issue of water evaporation must be addressed, and hydrogel sensors resistant to environmental conditions must be developed; on the other hand, whilst the introduction of nanofillers can effectively enhance the electrical conductivity of hydrogels, excessive addition of these fillers often compromises the mechanical properties of the hydrogels [44]. Consequently, there is an urgent need to develop simple yet effective strategies for fabricating nano-composite conductive hydrogels that combine excellent mechanical properties, electrical conductivity, water retention capacity and temperature resistance.

Nanomaterials

Flexible sensors have emerged as a rapidly expanding research frontier, enabling real-time, surface-based, and in vivo continuous biochemical monitoring for physiological and pathological assessment in healthcare and environmental applications. Their core performance depends on a material system that combines high electrical conductivity, mechanical flexibility, biocompatibility, and interfacial stability. Thanks to their unique structures and physicochemical properties, functional nanomaterials have become indispensable core components in the fabrication of high-performance flexible electrodes and signal-transduction layers. These materials can be broadly classified into three main categories: nanoparticles, one-dimensional nanostructures, and two-dimensional materials (Fig. 2).

Zero-dimensional nanomaterials (nanoparticles)

In flexible sensing systems, metal nanoparticles and their composites play a key role in forming continuous conductive networks and enhancing interfacial catalytic performance.

Gold nanoparticles (AuNPs) possess a high specific surface area and excellent electrical conductivity, enabling efficient integration with flexible substrates such as PDMS, PET, CC, or conductive polymers. Surface functionalization using thiol groups, antibodies, or polymer layers creates tailored bio-interfaces that ensure the stable immobilization of biomolecular probes, thereby enhancing the detection sensitivity and signal-to-noise ratio for biomarkers. Researchers often deposit these materials onto modified flexible substrates such as electron-transfer mediators and probe-anchoring sites to construct flexible sensors. For example, Zhu et al. prepared a tannic acid-3-aminopropyltriethoxysilane (TA-APTES) coating on activated carbon fabric (aCC), then in situ reduced AuNPs to produce an AuNPs/TA-APTES/aCC flexible electrode [45]. The AuNPs were subsequently loaded with glucose oxidase (GOx) to create a flexible skin sensor for monitoring glucose in biological fluids, thereby enhancing the detection sensitivity of the electrode. AuNPs are also frequently combined with other two-dimensional nanomaterials to enhance the electrochemical stability at the interface and form a stable conductive network structure. Wang et al. used chemical vapor deposition to sequentially deposit nitrogen-doped graphene (N-Gr) and AuNPs onto a flexible ITO-PET conductive film [46]. By combining the AuNPs with probes, they constructed a flexible electrochemical Apo-A4 immunosensor. Regardless of whether the sensor undergoes twisting, rolling, or bending, it maintains a stable electrical signal response without delamination or fracture of the N-Gr or gold layers. These strategies have shown encouraging results in enhancing conductivity and mechanical robustness. However, the interfacial adhesion energy and long-term signal drift under dynamic deformation remain to be more thoroughly characterized, and the reproducibility of sensor performance—particularly across different AuNPs loading and surface coverage—merits further investigation.

Silver nanoparticles (AgNPs) are a cost-effective conductive material suitable for the fabrication of printed and stretchable electrodes. Niu et al. employed a polymer-assisted metal deposition strategy to coat ordinary cellulose paper with AgNPs, thereby preparing a highly conductive paper-based electrode with a high conductivity of 4058.1 S/m [47]. The resistance remained stable even after 300 bending cycles, making it suitable as a highly conductive and stable electrode for electronic circuits. The low-temperature sintering properties of AgNPs enable the electrodes to maintain good electrical conductivity even under bending or tensile deformation, while polymer coating or compositing with carbon nanomaterials or MXene flakes can significantly enhance the performance of the composite materials. Li et al. designed a high-performance self-powered piezoionic sensor based on a flexible MXene/AgNPs heterostructure composite electrode [48]. The in-situ growth of AgNPs on MXene nanosheets not only effectively reduces the restacking of MXene nanosheets but also enhances the electrode’s electrochemical active area and structural stability. Lao et al. utilized AgNPs as a bridge to form a dual-conductive network structure using CNTs and AgNPs, thereby enhancing the conductive properties, strain sensitivity, and electrothermal performance of flexible sensors for potential applications in multifunctional wearable electronic devices [49]. AgNPs deliver outstanding conductivity and cost advantages for flexible electrode fabrication, and their hybridization with MXene, CNTs and other functional nanomaterials can further optimize conductive networks and mechanical flexibility, although the inherent oxidation vulnerability of AgNPs may compromise the long-term durability of wearable sensing devices.

Magnetic nanoparticles can serve simultaneously as both catalytic carriers and magnetic control units; they interact with enzymes through electrostatic interactions or covalent bonds, thereby enhancing substrate binding affinity, protecting the enzyme’s redox centers, and facilitating interfacial electron transfer. Li et al. developed a MXene/Fe₃O₄-based micro-glucose sensor patch [50]. With Fe₃O₄ as the catalytic support for GOx, the MXene/Fe₃O₄-based electrode exhibits enhanced enzyme immobilization and catalytic efficiency, achieving a wide glucose detection range of 0.1 mM to 10 mM and a low limit of detection (LOD) of 45.84 µM, and it successfully detects glucose levels in real human sweat and blood samples. Due to their magnetic properties, magnetic nanoparticles are also commonly used in flexible strain sensors. Cao et al. prepared a flexible strain sensor based on a PAm/Fe₃O₄@PPy hydrogel by combining Fe₃O₄@PPy nanocomposites with a PAm network [51]. The sensor exhibits exceptional tensile performance (961.2%), high toughness (1010.5 kJ/m³), and high electrical con ductivity (0.342 S/m) within the hydrogel system. The addition of Fe₃O₄@PPy nanocomposites endows the hydrogel with magnetic responsiveness, enabling it to monitor various human movements. Magnetic nanoparticles integrate catalytic enhancement and magnetic response functions, enabling flexible sensors to achieve excellent biochemical detection capability and mechanical sensing performance, yet their magnetic responsiveness is easily interfered by external magnetic fields, which may affect the sensing accuracy and signal stability of devices.

Quantum dots (QDs) are a class of nanostructures that exhibit quantum confinement effects; common types include graphene quantum dots (GQDs), carbon quantum dots (CQDs), and metal quantum dots (MQDs). Represented by CQDs, they are zero-dimensional carbon-based nanoparticles composed of sp²-conjugated nuclei. They have a high specific surface area, strong electrochemical response, and high electron transfer rate, and are widely used as sensing platforms for multi-parameter electrochemical detection. Wei et al. reported a portable multi-electrode system that utilizes different biosensor elements functionalized on the surface of a CQD-based working electrode [52]. This system facilitates rapid electron transfer at the electrode interface while simultaneously detecting Na⁺, ascorbic acid, and human neuropeptide Y in sweat. In addition, heteroatoms such as S, N, P, and B are widely used to modify the electronic structure of nanocarbon in carbon materials, thereby influencing their fundamental chemical and electrical properties. For example, boron-doped graphene quantum dots (B-GQDs) can provide additional active sites to enhance the electrocatalytic capacity for the oxidation of uric acid. Wang et al. developed a wearable electrochemical device for the detection of uric acid in sweat using a composite structure composed of B-GQDs and CNTs, achieving a sensitivity of 8.92 ± 0.22 µM⁻¹cm^− 2 [53]. The design of such quantum dot-modified electrodes provides new strategies and a material foundation for the continuous, non-invasive monitoring of metabolites in body fluids. With excellent electrochemical activity and tunable electronic properties, quantum dots and heteroatom-doped derivatives enable multi-parameter, high-sensitivity detection for non-invasive wearable sensing. Yet their poor structural stability and easy aggregation during long-term operation impair detection repeatability, which limits their practical application in flexible nanosensing.

One-dimensional nanomaterials

One-dimensional (1D) nanomaterials, including nanotubes, nanowires, and nanorods, serve as conductive backbones and catalytic mass transport channels in flexible sensors [54].

Carbon nanotubes (CNTs) have excellent electrical conductivity, an extremely high aspect ratio, and good mechanical resilience. Networks of carbon nanotubes deposited on polymer or textile substrates can form conductive pathways that maintain structural and performance stability even under cyclic deformation. Wu et al. designed a portable glucose sensor based on a flexible SWNT-based dual-transistor, which enables real-time serum glucose detection within the range of 100 µM to 100 mM [55]. This flexible device requires no additional calibration; its cyclic voltammograms remain highly consistent after approximately 100 repeated torsion cycles at a 90° angle, and the current remains stable over 7 days, demonstrating the potential of CNTs for use in highly stable flexible biosensors. Combining CNTs with metal nanoparticles not only stabilizes the conductive network through anchoring but also synergistically accelerates electron transfer, thereby enabling highly sensitive detection of analytes. For example, a study has reported a stretchable transparent electrochemical sensor based on a AuNPs/CNTs/PDMS composite structure, which can monitor H₂O₂ released by cells in real time under stretched conditions [56]. This composite structure combines excellent electrical conductivity, mechanical stability, and biocompatibility, and maintains a stable current response even after repeated deformation. Although the unique structural and mechanical merits of CNTs make them attractive for constructing robust flexible sensing networks, challenges related to their dispersion uniformity and interfacial adhesion remain important aspects for further optimization—particularly for applications demanding long-term operational stability.

Metal nanowires, nanorods, and nanotubes, such as silver nanowires (AgNWs), gold nanotubes (AuNTs), and platinum nanowires (PtNWs), have an extremely high aspect ratio that enables efficient charge transport and excellent optical transmittance, making them crucial for wearable sensors. Embedding them in elastomers such as PDMS and PU can prevent the conductive network from breaking mechanical deformation. Fan et al. developed a stretchable electrochemical electrode (Au@PtNTs/PDMS) based on a composite of AuNTs and platinum nanoparticles (PtNPs), which maintains stable response under a 50% tensile strain and enables real-time detection of H₂O₂ released by cells [57]. The transparent flexible AgNWs electrodes reported by Qin et al. can be further enhanced in oxidation resistance and fatigue resistance by polymer composites and surface modification, maintaining stable performance even after repeated bending or stretching [58]. This provides a highly conductive, highly transparent, and long-term stable conductive substrate for flexible sensors. Although elastomer embedding and surface modification strategies can effectively improve the mechanical durability and anti-oxidation capability of one-dimensional metal nanostructure electrodes, current studies mostly focus on performance optimization under laboratory conditions while ignoring long-term service degradation and environmental interference, so further testing is needed for this issue.

Common semiconductor nanorods include metal oxides such as ZnO, SnO₂, and TiO₂. These 1D structures have a high specific surface area and excellent optoelectronic properties, and are widely used in the development of gas, electrochemical, and biological flexible sensors [59]. Zhou et al. developed a flexible, non-enzymatic electrochemical glucose sensor based on in situ-grown hexagonal ZnO nanorods [60]. This structure leverages the high electron mobility and abundant surface-active sites of ZnO nanorods to promote valence band electron transitions and generate holes under UV light excitation, thereby enhancing the formation of surface hydroxyl groups. This significantly improves the electrochemical oxidation capacity for glucose and exhibits good linear response and reproducibility. This demonstrates the excellent photocatalytic properties of semiconductor nanorods in flexible, non-enzymatic biosensing, making them suitable for the development of highly sensitive, photo-enhanced flexible biosensors. However, the strong reliance on UV light excitation significantly limits their practical utility in wearable scenarios where UV exposure is minimal, and the structural integrity of these nanorods under repeated mechanical strain—closely tied to sensing stability—has yet to be systematically evaluated.

Two-dimensional nanomaterials

Two-dimensional (2D) nanomaterials combine high electron mobility, excellent surface activity, and strain adaptability, making them highly suited for applications in flexible electrochemical systems.

Graphene and its derivatives possess excellent electrical conductivity and an extremely large specific surface area, enabling them to provide stable electronic transport pathways even under bending stress. For example, Liu et al. integrated rGO-APBA with cellulose nanocrystal (CNC) hydrogels to leverage graphene’s highly efficient conjugated transport pathways, enabling the precise, non-enzymatic capture of glucose concentrations as low as 0.96 µM in sweat [61]. Simultaneously, they utilized the matrix’s high strain capacity (up to 550%) to achieve synchronous monitoring of human joint movements. Combining three-dimensional foam materials with graphene can further enhance the material’s fatigue resistance and effective specific surface area. Zhang et al. co-integrated graphene and MXene into a PU sponge, combining the high carrier mobility of GO with the mechanical adaptability of the sponge [62]. Experimental results show that this composite system maintains a high degree of signal consistency even after 8,000 load cycles and can adapt to a wide range of sensing requirements, from speech recognition to body movement. Graphene-based composites achieve outstanding mechanical flexibility and cyclic stability for multi-functional wearable sensing, yet the unavoidable restacking of graphene nanosheets in long-term operation gradually reduces active surface areas and weakens sustained sensing performance, which may be effectively mitigated by rational interlayer modification and structural spacer design.

MXenes possess metal-like conductivity and a variety of surface functional groups, making them effective signal amplification layers [63]. Composites made from this material combined with hydrogel or elastomer exhibit excellent adhesion and flexibility and are commonly used in the manufacture of flexible sensors. For example, Li et al. utilized the dense conductive network formed by MXene nanosheets on the surface of chlorinated polyurethane (CPU) fibers as a signal amplification layer, and leveraged the synergistic interaction between the physical roughness of SiO₂ nanoparticles and the surface end groups of MXene [64]. As a result, the prepared CPU@MXene@SiO₂ flexible sensor exhibited exceptional superhydrophobic properties (CA > 155°, RA < 3°), a high sensitivity factor (a sensitivity coefficient as high as 4922.6), and a fast response (response time of 94 ms). This not only resolves the issue of signal drift in flexible sensors when exposed to sweat but also enables them to accurately distinguish between large-scale limb movements and weak physiological signals such as heartbeats, thereby expanding the scope of MXene applications in complex environments. Conductive nanocomposites often suffer from hysteresis, and improving the dynamic response accuracy of strain sensors is a key solution to this problem. For example, Zhu et al. developed a sensor based on a novel MGS (MXene/GNP/silicone) hybrid nanocomposite [65]. The MGS features a unique curved, recessed honeycomb structure and is designed to leverage the synergistic effects of silicone and MXene/GNP nanobridges. This sensor utilizes machine learning to accurately classify different knee flexion angles with an accuracy rate as high as 98.55% and provides real-time feedback, highlighting its immense potential for future applications in wearable human-computer interaction and rehabilitation training. The ultra-high sensitivity, rapid response, and anti-interference capability of MXene-based sensors make them highly promising for complex wearable applications. Addressing their environmental stability—particularly in humid and oxidative conditions—remains an active area of research toward improving their long-term reliability and practical deployability.

Metal-organic frameworks (MOFs) are typically three-dimensional crystalline networks; however, a growing class of two-dimensional MOF nanosheets has emerged that retains the tunable pore structures and molecular sieving capabilities of bulk MOFs while offering superior processability for flexible devices. MOFs have tunable pore structures. When combined with conductive substrates, they form highly efficient and selective molecular sieve layers, which enhance detection sensitivity and signal amplification [66]. Through micro- and nano-structural design and multi-component synergistic effects, researchers have addressed issues such as poor interfacial adhesion and insufficient mechanical durability, thereby achieving a seamless integration of functionality and flexibility. Yan et al. combined two-dimensional zinc porphyrin MOF nanosheets with MCNTs to fabricate enzyme-free photoelectrochemical (PEC) sensors that can be read by smartphones, fabricated on flexible non-woven fabric and PC film substrates [67]. The device exhibits high consistency under dynamic deformations such as compression, twisting, and stretching, and can detect vitamin C in sweat at concentrations as low as 3.61 µM, demonstrating exceptional flexibility and durability. Significant progress has also been made in pressure-sensing systems based on conductive MOFs with layered structures. Zhao et al. synthesized a multifunctional 3D carbon nanofiber network (CNFN) hybrid core-shell material incorporating a conductive MOF (copper catechuate, Cu-CAT), endowing the sensor with an ultra-high sensitivity of 30.40 kPa^-1 and a wide detection range of 0.5–60 kPa [68]. The electrodes’ excellent breathability and stability over more than 5000 pressure cycles demonstrate the feasibility of MOF-based composites for use in high-performance electronic skin applications. By optimizing interfacial bonding and morphological control, it is possible to effectively overcome the inherent mechanical limitations of MOF materials, thereby providing a new approach for developing long-lasting, high-precision wearable monitoring platforms. Compared with conventional carbon and metal nanomaterials, 2D MOF materials possess unique structural tunability and molecular screening capabilities for high-precision sensing, whereas most MOF-based flexible composites still suffer from weak structural robustness and poor environmental tolerance compared with conventional carbon and metal nanomaterials, which can be improved via interfacial crosslinking and dual-material synergistic compounding strategies.

Transition metal dichalcogenides (TMDCs), such as MoS₂ and WS₂, possess abundant edge active sites and intrinsic catalytic activity, enabling them to significantly reduce the overpotential of redox reactions and thereby achieve highly sensitive detection of biomarkers at low detection potentials [69]. In biosensing applications, Ge et al. developed a flexible glucose sensor based on copper (Cu)-doped MoS₂ (Cu-MoS₂/CC), which exhibited a sensitivity of 364 µA·mM^-1·cm^-2 over a wide linear range of 1 µM to 20 mM [70]. When combined with one-dimensional nanostructures, TMDCs can overcome their own mechanical limitations in flexible sensors. For example, Li et al. designed a flexible PVA/Ag/Au/MoS₂/CNT electrochemical sensing interface [71]. By leveraging the synergistic effects of monolayer MoS₂ and cross-linked CNTs, they improved the detection limit for dopamine (DA) compared to pure CNT electrodes, while ensuring that the device retained 87.4% of its initial activity after undergoing 50 cycles of 90° bending. TMDCs also demonstrate unique heterojunction advantages in the field of gas sensing. Gao et al. proposed a flexible NO₂ sensor based on a WO₃/MoS₂ heterojunction [72]. Due to the synergistic effect of WO₃ and MoS₂, this sensor exhibits a high response of 364.5% to a target gas concentration of 1 ppm, while maintaining reliable detection accuracy even under extreme service conditions involving repeated bending. Once integrated into a real-time environmental monitoring platform, this technology enables dynamic early warning of environmental safety risks, further expanding the application scope of TMDCs composites in the field of multimodal wearable sensing. TMDCs excel in low-potential catalytic sensing and multi-modal detection via heterojunction synergies, yet their intrinsic low electrical conductivity and structural fragility limit the cyclic stability of pure TMDC-based devices, and constructing multi-dimensional hybrid conductive networks is an effective approach to balance their sensing activity and mechanical stability. The representative performance metrics of the nanomaterials discussed above are summarized in Table 2 for direct comparison across different material categories and sensing platforms.

Table 2.

Representative performance metrics of nanomaterials in flexible sensors

Nanomaterial	Composite / Substrate	Sensing Target	Sensitivity	LOD	Stability / Cycling Performance	Ref.
AuNPs	AuNPs/TA-APTES/aCC + GOx	Glucose (sweat)	72.0 µA·mM− 1·cm− 2	3.3 µM	93.6% after 15 days; 200 bending cycles; response 4.0 s	[45]
AuNPs	N-Gr/AuNPs/ITO-PET (Apo-A4 immunosensor)	Apo-A4	- (EIS-based)	0.010 ng/mL	93.79% (1 week), 87.58% (3 weeks); stable under bending	[46]
AgNPs	Ag/PDA/paper	General electrode	4058.1 S/m (conductivity)	-	300 bending cycles; GF ~ 4.0; stable after ultrasonication	[47]
AgNPs	MXene/Ag heterostructure (piezoionic)	Mechanical strain	11.1 mV at 0.7% strain	-	95% retention over 13,000 s	[48]
AgNPs	PM film (PVA/OP/NCC/PA/AgNPs/CNTs)	Strain / temperature	GF = 1.66; TCR=–0.65% °C⁻¹	-	320 cycles; 30 days air stable	[49]
Fe3O4	MXene/Fe3O4(printed on PET) + GOx	Glucose (sweat, blood)	11.07 / 2.42 µA·mM− 1·cm− 2	45.84 µM	5000 CV cycles; 95% after 1000 bending; >90% over 18 days	[50]
Fe3O4	PAm/Fe3O4@PPy hydrogel	Strain + magnetic field	Conductivity 0.342 S/m	-	961.2% stretch; 1010.5 kJ/m3 toughness; magnetic response	[51]
CQDs	CQD-based SPCE (PET) + Na+-selective membrane / AAOx / NPY-Ab	Na+, AA, NPY (multiplex)	AA: 2.02 nA·µM− 1·L− 1; Na+/NPY: -	Na+: 10.9 µM; AA: 12 nM; NPY: 1.1 fM	8 days stable; pH 5.0–7.4 stable	[52]
BGQDs	BGQDs/CNTs (GCE/flexible)	Uric acid	GCE: 8.92; Flexible: 35.1 µA·µM− 1·cm− 2	0.99 µM	91.28% after 30 days; recovery 98.35%	[53]
SWNTs	SWNT twin transistors (PUA) + GOx/Nafion	Glucose (serum, wound fluid)	- (transistor, 100% per decade)	-(range 0.1–100 mM)	7 days stability; noise reduced 59%; bent/twisted 100 cycles	[55]
CNTs + AuNPs	Nano-Au/CNTs/PDMS	H2O2 (cell-released)	-(amperometric, − 0.6 V)	8 nM	100 bending/stretching cycles	[56]
AgNWs	AgNWs/propolis/PET	General electrode (OLED, pressure sensor)	~ 15 Ω/sq; ~93% transmittance; RMS 6.2 nm	-	10,000 bending cycles; 700 h at 60 °C/85% RH stable; 200 tape tests	[58]
ZnO nanorods	ZnO nanorods / SSWS (stainless steel wire sieve)	Glucose	36.4 µA·mM− 1·cm− 2 (without UV); 91.8 µA·mM− 1·cm− 2 (with UV)	2.1 µM	60% after 100 bending cycles; 90% after 20 uses; 96.1% after 8 uses (with UV); response < 3 s	[60]
rGO-APBA	rGO-APBA / CNCs-Col-PVA hydrogel (functionalized SPE)	Glucose (sweat) + human motion	GF = 1.11 (0–550% strain); glucose: -	0.96 µM (glucose)	550% stretch; self-healing >89%; tensile strength 2.23 MPa	[61]
Graphene + MXene	Graphene/MXene@PU sponge (PDMS adhesive)	Pressure / strain (speech, joint motion, gait)	2.68 kPa− 1 (0–15 kPa); 1.02 kPa− 1 (15–40 kPa); 0.33 kPa− 1 (40–100 kPa)	—	8,000 cycles stable; 92.8% retention; response 61 ms	[62]
MXene (Ti3C2Tx)	CPU@MXene@SiO2	Strain + pressure (swallowing, arm/finger/knee bending, gait, droplet impact)	GF = 4922.6 (0–150% strain)	-	Superhydrophobic; response 94 ms; 5000 cycles stable at 10% strain; stable in acid/alkali/salt/UV/humidity	[64]
MXene + GNP	MGS (MXene/GNP/silicone, auxetic curved re-entrant honeycomb)	Strain (knee bending, finger/palm/wrist/neck/elbow, breathing, cheek blowing)	GF = 2.3 (0–40%); GF = 7.2 (40–60%)	-	Low hysteresis ~ 10%; 98.6% stress retention after 600 cycles; 1000 cycles stable; waterproof; response 120 ms	[65]
MOF	2D-TCPP(Zn)/MCNTs	Vitamin C (sweat)	- (PEC, cathodic photocurrent)	3.61 µM	95.8% retained after 16 days storage; 27 cycles PEC stable; stable under compression/twisting/stretching	[67]
MOF	Cu-CAT/CNFN (carbon nanofiber network) + PU	Pressure (finger touch, pulse, knee motion, sensor array)	30.40 kPa− 1 (0.5–60 kPa)	-	> 5,000 pressure cycles; response <0.24 s, recovery <0.31 s; range 0.5–60 kPa	[68]
MoS2	Cu-MoS₂ / carbon cloth	Glucose	364 µA·mM− 1·cm− 2	-	Linear range 1 µM–20 mM; 5000 bending/stretching cycles with stable performance	[70]
MoS2 + CNT + Au/Ag	PVA/nano-Ag/Au/MoS2/CNT	Dopamine	- (CA, 0.15 V)	0.426 µM	87.4% activity retained after 50 bending cycles (90°); 94% stability over 2 days	[71]
WO3/MoS2heterojunction	WO3/MoS2 / PI	NO2 (gas)	364.5% (ΔR/R0 × 100% at 1 ppm)	-	Response stable under repeated bending (0–120°); 200 bending cycles (180°) with stable performance; 30-day stability	[72]

Design principles and challenges

Rational design of high-performance flexible nanosensors demands a comprehensive balance of electrical conductivity, mechanical flexibility, interfacial stability, catalytic activity, and biocompatibility under complex deformation conditions in wearable and implantable scenarios. Zero-, one-, and two-dimensional nanomaterials can form continuous, deformable conductive networks to enable highly sensitive biochemical detection and physiological signal monitoring. However, current nanomaterial-based flexible sensors still face systematic challenges at the material, structural, and application levels, which induce performance degradation and severely restrict their practical application and clinical translation [73].

From the perspective of intrinsic nanomaterial defects and network failure mechanisms, zero-dimensional nanoparticles inevitably suffer from agglomeration, uneven distribution, and long-term ion leaching under cyclic stretching and humid physiological environments, which gradually breaks the continuity of conductive pathways and causes baseline drift and sensitivity degradation [74]. One-dimensional nanomaterials such as carbon nanotubes and metal nanowires rely on percolation conductive networks; repeated bending and stretching readily trigger abrupt junction resistance changes, network disconnection, and interfacial delamination between nanowire/nanotube skeletons and polymer substrates, resulting in significant mechanical fatigue and unstable dynamic responses [75]. For two-dimensional layered materials, graphene and MXene nanosheets are prone to irreversible restacking and interlayer slippage under sustained strain [76]. MXenes also undergo severe oxidation and corrosion in humid, oxygen-rich environments, further impairing electrode conductivity and sensing repeatability [77]. In addition, the anisotropic charge transport of layered two-dimensional materials leads to direction-dependent signal responses, introducing systematic detection errors under complex multidirectional deformations caused by human motion [78].

In practical applications, long-term skin contact and implantable monitoring bring common biofouling problems, including protein adsorption, cell adhesion, and sweat corrosion. These issues passivate active electrode sites and interfere with interfacial electron transfer, causing chronic signal drift [79]. Most critically, a fundamental performance trade-off widely exists in current flexible sensing systems: improving stretchability and mechanical compliance usually requires increasing polymer matrix content or using loose network structures, which inevitably reduces electrical conductivity and sensing sensitivity; conversely, dense conductive networks enhance detection performance but weaken mechanical deformability and long-term biocompatibility [80]. Beyond material and structural limitations, most reported studies only evaluate short-term performance under ideal laboratory conditions. Systematic research on long-term cyclic stability, anti-interference ability, and durable biocompatibility under real conditions of sweat, skin friction, and variable implant environments remains insufficient, leading to overestimated practical reliability of existing flexible sensors [81].

The long-term biocompatibility and degradation behavior of nanomaterials are critical for implantable and wearable applications. Zero-dimensional nanoparticles carry diverse risks: AuNPs accumulate in organs with uncertain chronic effects [82]; AgNPs continuously release cytotoxic Ag⁺ ions [83]; Fe₃O₄ nanoparticles may induce oxidative stress via Fenton-like reactions [84]; and heavy-metal quantum dots have inherent toxicity from cadmium leaching [85]. One-dimensional nanomaterials pose morphology-dependent hazards: CNTs can trigger asbestos-like frustrated phagocytosis and chronic inflammation [86], metal nanowires combine ion-release toxicity with fibrous geometry, and ZnO nanorods dissolve to release Zn²⁺ [87]. Two-dimensional materials also present distinct challenges: graphene oxide toxicity depends on lateral size and oxidation state [88]; MXenes degrade rapidly, producing acidic intermediates and losing conductivity [77]; TMDCs release molybdate or tungstate ions with unknown chronic effects [88]; and MOFs carry risks of synergistic toxicity from metal nodes and organic linkers [89]. Across all dimensionalities, current biocompatibility assessments mostly rely on short-term static assays, which fail to capture the coupled effects of mechanical degradation, ion leaching, and immune activation under realistic physiological conditions, making general claims of “good biocompatibility” largely unsubstantiated [90].

To address these bottlenecks, future design strategies should focus on resolving the multidimensional performance trade-off. Developing multiscale hybrid nanostructures, interfacial crosslinking engineering, and surface antifouling/antioxidation modifications can effectively suppress material aggregation, network fracture, and signal attenuation caused by biofouling [77, 79, 91]. Furthermore, establishing standardized mechanical deformation evaluation protocols and long-term stability testing systems for wearable and implantable devices is essential to eliminate performance inconsistencies across studies, thus promoting the practical application of flexible nanosensors with high stability, high precision, and reliable biocompatibility [81].

Wearable sensors

In the field of flexible electronics, wearable devices have garnered widespread attention due to their compact size and multifunctional capabilities. Wearable devices typically consist of sensors integrated into wearable forms such as clothing, wristbands, patches, or tattoos, which are attached to the body to continuously monitor a range of physiological parameters [92]. Advanced nanomaterials have attracted widespread attention in wearable sensor technology due to their conductivity, selectivity, sensitivity, accuracy, biocompatibility, and active sites [93]. A wide range of wearable sensors have now been developed to monitor human health parameters such as temperature, pressure, stress, strain, pH levels, and biomolecules, driving a profound shift in healthcare models from passive disease treatment toward proactive, real-time prevention and personalized management [94].

Mechanical monitoring

The human body continuously generates diverse mechanical activities, including heartbeat rhythm, pulse fluctuation, respiratory movement, and limb flexion-extension. The mechanical signals produced by these activities serve as important indicators for assessing health status. To convert these deformations into readable electrical signals, sensors must maintain stable output under repeated stretching, bending, and twisting [30] .

Long-term durability places high demands on the mechanical strength of materials; however, high-strength materials exhibit significant residual strain and stress loss. To resolve the conflict between high strength and low hysteresis, Chen et al. developed a composite hydrogel (AGCT/Ag/Gs) featuring a multiphase structure and particle-polymer interfacial cross-linking. Its network structure facilitates multi-level energy dissipation, thereby ensuring a balance between the hydrogel’s mechanical strength (Tensile strength @ 2.9 MPa, elongation at break @ 660%) and hysteresis (0.15 at 100% strain). Furthermore, the incorporation of the nanomaterial TA@AgNPs into the hydrogel matrix confers antibacterial and free radical scavenging properties. Wearable sensors fabricated from this material enable reliable, real-time monitoring of human movement in humid environments, with an operating range of 0-500%, and can be further integrated with customised integrated circuits and Bluetooth for wireless signal transmission [95]. To enable long-term monitoring, improving sensor stability is also a significant challenge; Cai et al. have developed a flexible sensing layer with high conductivity and stability using MXene-Ag nanotube composites (Fig. 3a) [96]. MXene provides a stable conductive backbone, while Ag nanotubes bridge adjacent MXene sheets to maintain network integrity under deformation. Furthermore, the adoption of a biomimetic microstructure design has optimized the sensor’s mechanical and electrical properties, enhancing the stability of the contact interface and overall sensing performance. Experimental results demonstrate that the sensor exhibits ultra-high sensitivity (with a minimum detectable pressure of 0.1 Pa), a fast response time (42 ms), a rapid recovery time (21 ms), good linear response and excellent cycling stability (> 5000 cycles). This device is capable of accurately detecting pulse signals associated with cardiovascular diseases, offering a low-cost, portable and non-invasive health monitoring solution.

Fig. 3.

Partial wearable sensor strategy diagrams. a Fabrication process flow of the pressure sensor [96]. b Illustration of next-generation nanofiber-based self-powered biosensor to monitor glucose levels in sweat [98]. c Schematic of the flexible patch based on sensing and microfluidic modules [99]. Source: Reprinted with permission from ref [99]. Copyright 2025, with permission from American Chemical Society. d (i) Exudate is pumped from the hydrophobic side to the hydrophilic side with the effect of wettability gradient. (ii) Exploded diagram of the Janus bioelectronic wound care bandage [100]. Source: Reprinted with permission from ref [100]. Copyright 2025, with permission from American Chemical Society.

Nanomaterial-composite hydrogels are widely used in the field of mechanical sensors due to their excellent mechanical properties; however, to ensure their long-term applicability, consideration must also be given to mitigating the effects of water evaporation and the susceptibility of conductive fillers to oxidation on their performance. In addition, flexible wearable force-monitoring devices require further enhancement in areas such as power supply and data transmission; technologies closely linked to force sensing, such as piezoelectricity and triboelectricity, offer potential solutions to these challenges [97].

Humidity monitoring

Abnormalities in respiratory rate and rhythm are common clinical manifestations of respiratory diseases, cardiovascular dysfunction, and psychological stress. Humidity sensors enable continuous monitoring of respiratory parameters by detecting moisture content in exhaled breath [101]. However, the sensing mechanism relies on water molecule adsorption onto the active layer, and the speed of this adsorption–desorption process directly determines the sensor’s response time [102].

Conventional moisture-absorbing materials require time to reach equilibrium with the environment, limiting their ability to capture rapid respiratory changes [103]. Nanomaterials with different chemical compositions offer distinct strategies for achieving rapid humidity response, each leveraging their unique surface properties. The basic structure of an impedance-based humidity sensor involves coating an electrode with a thin film of humidity-sensitive nanomaterial [103]. By utilising the Grotthuss proton conduction mechanism and selecting insulating nanoparticle materials with hydrophilic surfaces, such as silicon dioxide, the issue of rapid conduction can be effectively resolved. For example, Kano et al. designed a head-worn device comprising a sensor chip made of silicon dioxide (SiO₂) nanoparticles, whose abundant surface hydroxyl groups serve as proton hopping sites for rapid moisture equilibration and a microcontroller with Bluetooth connectivity; this device is capable of accurately monitoring a wide range of respiratory rates, from apnoea (0 Hz) to 1.7 Hz (102 bpm) [104]. Furthermore, the electrical conductivity and permittivity of nanomaterial films used in impedance-based humidity sensors can be directly modulated by the adsorption of water molecules. Based on this, semiconductor and metal nanomaterial devices (e.g., graphene oxide, CNTs, MXene nanosheets) or metal-oxide semiconductors can be selected. et al. proposed an MXene composite modified with nickel-iron layered double hydroxide as a humidity-sensing material. MXene possesses hydrophilic surface terminations and metallic conductivity, enabling water adsorption to directly modulate its electrical properties. The LDH modification introduces additional hydrophilic interlayers that further enhance water uptake and proton transfer. Experimental results indicate that the humidity sensor developed in this study exhibits outstanding sensitivity (response/recovery times of 1.10 s and 7.89 s, respectively) and satisfactory response intensity (90,843% at 97% relative humidity), whilst also demonstrating good physicochemical stability and functional durability, enabling the measurement of skin moisture [105]. When selecting materials for humidity sensors, highly hydrophilic surfaces can enhance the response amplitude; however, this results in a high energy barrier for the desorption of water molecules. To address this, the appropriate introduction of hydrophobic chemical modifications can promote spontaneous desorption and shorten the recovery time [104]. Thinner nanomembranes can achieve faster response and recovery times, but the reduced water adsorption capacity leads to a decrease in sensitivity; whilst increasing the thickness of the nanomembrane enhances sensitivity, it slows down the response. These factors should be weighed up during the material optimisation process [106].

Wearable humidity monitoring devices are well-suited for respiratory monitoring due to their configurability and non-invasive nature. Furthermore, they hold potential applications in fields such as speech recognition, contactless user interfaces and skin assessment, although these are currently still at the feasibility stage [103].

Sweat monitoring

Traditional blood testing, while providing accurate biochemical markers, is inherently invasive and unsuitable for continuous monitoring [107]. Sweat, as a non-invasively accessible biofluid, contains various physiologically active substances including ions, metabolites, and hormones [37]. However, the concentration of target analytes in sweat is typically one to two orders of magnitude lower than in blood, and its composition is influenced by diet, exercise intensity, and sweating rate, imposing higher demands on sensor sensitivity, multiplexed detection capability, and anti-interference performance [37].

Different nanomaterials play distinct roles in sweat sensing—some provide conductive backbones, some offer catalytic sites, and some are responsible for molecular recognition. Asaduzzaman et al. investigated an electrochemical epidermal patch based on the MXene@Pt nanocomposite anchoring electro-deposited platinum nanoparticles (PtNPs) functionalized laser-induced graphene (LIG) electrode (LIG/MXene@Pt/PtNPs) for glucose, pH, and Na⁺ monitoring. The metallic conductivity of Ti₃C₂Tx-MXene, the high electrocatalytic performance of PtNPs and PB towards H₂O₂, and the enhanced electrochemical surface area of HNPC-modified electrodes play a crucial role in achieving ultra-high detection sensitivity in glucose and lactate sensor patches. Tests have confirmed that this sensor is capable of monitoring sweat during physical activity [108]. Self-powered designs for sweat sensors have also attracted considerable attention in recent years. Gungordu Er et al. utilized triboelectric nanogenerators (TENGs) combined with functionalized nanofibers to monitor glucose levels in sweat (Fig. 3b) [98]. The electrospun nanofibers integrate multiple functional components: graphene derivatives (GO, PG, GF) form conductive pathways for charge transfer; PPy enhances triboelectric charge density; PCL provides mechanical resilience; and glucose oxidase (GOx) confers biospecificity for glucose recognition. Furthermore, triboelectric testing revealed that the PCL/PPy/GO and polyvinylidene fluoride nanofibers containing glucose oxidase exhibited a peak voltage of 63 V. These findings highlight the potential of PCL/PPy/GO nanofibers to provide highly sensitive glucose detection and non-invasive sweat analysis for diabetes management systems, as well as their robust self-powering capabilities.

Self-powered sweat sensors typically follow two design approaches: TENG and biofuel cells. TENGs commonly utilise electrospun nanofibres as the triboelectric layer; material design must strike a balance between conductivity, triboelectric output and mechanical flexibility: graphene derivatives such as GO and porous graphene provide a conductive scaffold, conductive polymers such as PPy enhance π-conjugation and triboelectric charge density, whilst spinnable polymers such as PCL and PVDF impart mechanical resilience and processability [98]. Biofuel cells, on the other hand, utilise lactic acid or glucose present in sweat as fuel, generating electrical energy through the action of redox enzymes or microbial catalysts; however, the immobilisation of enzymes or catalysts within nanofibre or hydrogel networks may lead to leakage or inactivation following prolonged immersion in sweat or over time [109]. Current self-powered designs still face challenges regarding energy conversion efficiency and output stability.

The application of sweat sensors in flexible wearable devices still requires further research and development, such as reducing the impact of noise caused by strain or movement at the interface between soft sensor components and more rigid electronic devices, and determining how to induce sweat production in sedentary environments for continuous monitoring [92].

Wound monitoring

Traditional wound care relies on subjective clinical assessment and periodic laboratory testing—a “passive” management model that often leads to delayed intervention [110]. Biomarkers provide real-time information on the wound healing process and the development of complications. Monitoring changes in wound biomarkers offers valuable guidance for clinical prognosis. Flexible bioelectronics not only offers advantages such as low cost, rapid implementation and ease of use, but also exhibits excellent mechanical compliance and biocompatibility. It holds significant advantages when integrated into wearable devices and offers immense potential for personalized wound care.

Multi-channel wound sensors must detect multiple biomarkers in a complex and fluctuating environment. In particular, the protein-rich and cell-laden nature of wound exudate makes sensor surfaces highly susceptible to biofouling—a challenge less pronounced in sweat or blood monitoring. Wang et al. developed a Janus dressing with a sensor array incorporating AuNPs and AgNPs (Fig. 3d) [100]. Their high specific surface area and conductivity enhance electrochemical signal transduction for real-time monitoring of glucose, urea, and pH in wound exudate. Deng et al. used a flexible patch with carbon dot-decorated CNTs (CDs-CNTs) as the sensing material (Fig. 3c) [99]. CNTs form a high-aspect-ratio conductive network, while carbon dots provide electrochemical activity for oxygen and uric acid detection. To address pH interference, Kaewpradub et al. designed a polyaniline/CNT-based sensor that detects pyocyanin while correcting dynamic pH fluctuations [111]. CNTs ensure high conductivity for sensitive detection, while polyaniline’s pH-responsive properties contribute to real-time signal correction.

Sensitivity is a key criterion for selecting wound sensor materials, as it enables the detection of low-concentration biomarkers to identify early-stage infections or subtle changes in wound condition. Another important consideration is that sources of interference—such as sensor drift or bodily fluids—can easily compromise this precision, necessitating improved specificity toward non-target molecules [112]. These challenges can be addressed through the application and optimization of nanomaterials. Carbon-based materials (e.g. CNTs) are widely used due to their excellent conductivity, biocompatibility, and large specific surface area; however, their intrinsic selectivity is limited. They typically require functionalization with biomolecules such as enzymes or antibodies to specifically recognize glucose, lactate, or cytokines, thereby mitigating interference from off-target molecules [113]. Precious metals (e.g. AuNPs) possess faster electron transfer kinetics, enabling the detection limits of inflammatory markers such as IL-6 to be reduced to sub-micromolar levels; however, their durability in long-term exposure to moist exudate and cost issues still need to be balanced through robust encapsulation or low-power designs [114]. In summary, material systems must be optimized through strategies such as nanostructuring, surface functionalization, and anti-fouling coatings to ensure both stable monitoring and precise selectivity in the complex environment of exudate.

Wearable wound monitoring devices are currently being integrated with artificial intelligence and machine learning. The quantification of biomarker indicators by sensors, combined with algorithmic model analysis, provides valuable support for predicting treatment trajectories. However, the transition to clinical application still faces a range of challenges, including regulatory issues, cost-effectiveness, energy-related concerns and patient compliance [110]. Overcoming these obstacles is crucial for translating these promising technologies into routine clinical practice.

In this section, we provide an up-to-date overview of the applications of nanomaterials in emerging flexible wearable devices. As shown in Table 3, wearable flexible sensors have achieved significant progress, enabling continuous, real-time monitoring of human physiological health status across multiple dimensions, ranging from motion and humidity to various bodily fluids such as sweat and wound exudate. For wearable sensors, the development of nanomaterials and effective structural designs have become crucial directions for further improvement. These advancements not only enhance the fundamental properties of flexible sensors (e.g., sensitivity, linearity, and durability), but also introduce specific characteristics via functional modification (e.g., biocompatibility and adhesion), which facilitates long-term device usage and functional customization. This offers broad prospects for the commercialization and clinical implementation of flexible wearable devices.

Table 3.

Comparison of representative flexible nanotechnology-enabled wearable sensing platforms

Study		Substrate Material	Nanomaterial	Sensing Target	Sensing Mode	Sample Matrix / Validation Model	Key Performance and Advantages
[95]	AGCT hydrogel		TA@AgNPs	Human motion	Strain sensing	Skin surface (humid environment)	Tensile strength 2.9 MPa, elongation at break 660%, working range 0–500%; hysteresis 0.15 (at 100% strain); Bluetooth wireless transmission
[96]		—	MXene-Ag nanotube composite	Pressure/pulse	Piezoresistive	Skin surface	Detection limit 0.1 Pa, response time 42 ms, recovery time 21 ms; >5000 cycles
[104]		PI	SiO₂NPs	Humidity/respiration	Humidity sensing	Exhaled breath	Response/recover time:0.7 s /1.7s
[105]		Paper substrate	NiFe layered double hydroxide modified MXene composite material	Humidity/ respiration	Humidity sensing	Exhaled breath	Response/recoverytime:1.10s/7.89s response intensity: 90,843% at 97% relative humidity
[108]		LIG electrode	MXene@Pt/PtNPs	Glucose, pH, Na⁺	Electrochemical	Sweat (during exercise)	Glucose sensitivity 86.45 µA·mM⁻¹; Lactate 0–100 mM; pH sensitivity 50.99 mV/pH; Na⁺ 56.26 mV/decade
[98]		Electrospun nanofibers (PCL/PPy/GO, PVDF)	GO, PG, GF, PPy, GOx	Glucose	TENG	Sweat	Peak voltage 63 V; self-powered
[100]		Janus dressing (hydrophobic TPU + hydrophilic gauze)	AuNPs, AgNPs, LEG	Glucose, urea, pH	Electrochemical sensor array	Wound exudate	Glucose range 0–14 mM, LOD 0.15 mM; UA range 100–600 µM, LOD 6.85 µM; pH sensitivity 60.76 mV/decade over pH 4–8
[99]		Flexible PET patch (biomimetic conical microstructure)	CDs-CNTs, AgNDs/CDs-CNTs	pH, lactate, UA, O2,temp.	Microfluidic + sensor array	Wound exudate	O₂ detection range 5–100 mmHg; UA linear range 50–900 µM, LOD 2.2 µM; Lactate range 1–30 mM
[111]		Textiles	polyaniline/CNT	Pyocyanin, pH	Square wave voltammetry + potentiometry	Wound exudate	Partial least squares regression correcting dynamic pH fluctuations

Implantable sensors

Compared with traditional in vitro testing and external wearable devices, implantable sensors are placed directly within the target tissue or the microenvironment of bodily fluids, enabling them to sensitively detect the faint and fluctuating biological signals characteristic of the early stages of disease [115]. Traditional rigid microelectrodes are prone to causing significant tissue damage and glial scarring following long-term implantation [116, 117]. Flexible sensors minimize trauma to healthy tissue during both implantation and explantation. Furthermore, the incorporation of nanomaterials enhances the sensitivity and biocompatibility of the sensors. Currently, flexible implantable sensors based on nanomaterials are commonly used in the eye, brain and cardiovascular systems, amongst others. However, in the path toward clinical application, implantable sensors still have limitations: the high humidity, high salinity, and continuous organ movement in vivo restrict the interfacial impedance stability, antifouling capability, and electrochemical durability of the electrodes [117].

Eye

The onset and progression of eye diseases are typically accompanied by persistent abnormalities in local physiological parameters, particularly intraocular pressure, corneal deformation and changes in the composition of ocular surface and intraocular fluids [117]. Flexible implantable sensors based on nanomaterials effectively adapt to intraocular tissues through their ultra-thin structure, transparent conductive networks and deformable interface design. They can be directly implanted into ocular tissues to monitor various biomarkers, including glucose, lactate, electrolytes, and inflammation-related factors, in intraocular pressure, aqueous humor, and tears [118].

Compared to traditional single-point intraocular pressure (IOP) measurements, implantable flexible sensors are crucial for continuously monitoring IOP to track changes in the disease. In recent studies, implantable microfluidic IOP sensors have improved measurement accuracy and precision by coming into direct contact with the aqueous humor. Li et al. reported a microfluidic IOP sensor fabricated using microfluidic technology with PDMS as the biocompatible material. Unlike methods that rely on optical technology to record fluid displacement caused by intraocular pressure, this device employs a piezoresistive sensor principle, converting pressure changes into changes in electrical resistance. Implanted into the vitreous humor of rabbits, it achieves precise intraocular pressure measurement with high linearity (R² = 0.999) and sensitivity as high as 0.00253 Ω/mbar [119]. While the piezoresistive approach simplifies readout compared to optical methods that require external cameras or smartphones, its reliance on resistive changes makes it inherently susceptible to temperature drift and long-term baseline instability in the thermally variable intraocular environment — a limitation rarely discussed in proof-of-concept studies. Wireless contact lenses also offer irreplaceable advantages in IOP measurement. For example, traditional microcrack-based IOP sensors face a trade-off between “high sensitivity” and a “narrow monitoring range”. Wu et al. adopted a biomimetic “disconnected yet connected” design to develop a flexible contact lens for IOP measurement based on Ti₃C₂Tx/MXene@C-MWCNTs nanomaterials [120]. They utilized MXene nanosheets as “lotus roots” that fracture under minute strains, providing high sensitivity (33.21 mV/mmHg), while introducing carbon nanotubes as “threads” that bridge the cracks, maintaining the conductive pathway under greater strains, thereby significantly expanding the sensor’s monitoring range to 60 mmHg. This dual-network strategy effectively mitigates—though does not eliminate—the sensitivity–range trade-off; notably, the long-term stability of the intentionally fractured MXene network under cyclic ocular deformation and tear film exposure remains unaddressed, raising questions about whether the initial sensitivity can be sustained over clinically relevant durations. T. Y. Kim et al. reported a gold hollow nanowire (AuHNW)-based piezoresistive sensor for long-term monitoring of intraocular pressure, which enabled continuous measurement of intraocular pressure levels in rabbits. AuHNWs offer reasonable stretchability and high transparency, allowing the sensor to be made transparent for use in contact lenses [121]. Compared to MXene/CNT hybrid strategies, AuHNWs offer intrinsically superior chemical stability and long-term biocompatibility due to the bio-inert nature of gold. However, this chemical inertness also means fewer active sites for signal amplification, illustrating a recurring dilemma in ocular sensor design: materials that are the most biochemically stable are often the least responsive as sensing elements, and vice versa. Implantable sensors are also used to analyze important biomarkers in the eye, such as tears and aqueous humor. In conditions like diabetes, elevated blood glucose levels are accompanied by elevated tear glucose levels; therefore, tear glucose monitoring plays a crucial role, and the response speed of the electrodes is critical in minimizing the lag time between tear glucose and blood glucose levels. Based on this, Park et al. developed a wireless smart contact lens using an electrochemical enzyme sensor (Fig. 4a) [122]. This device enables high-frequency (every 20 s) TG monitoring, providing unprecedented continuous tear glucose data collection at sub-minute intervals, suggesting the potential of TG monitoring as a non-invasive alternative to BG monitoring in human subjects, although clinical validation remains pending. The team led by S. K. Kim utilized bimetallic nano catalysts (A-Au@Pt BiNCs) embedded in a nanoporous hydrogel attached to a smart contact lens for long-term, robust continuous glucose monitoring (CGM) in diabetic rabbits; however, the long-term robustness demonstrated in rabbits (hours to days) remains far from the months-long stability required for clinical CGM [123]. Metal nano catalysts promote the rapid decomposition of hydrogen peroxide (a product of glucose oxidase) and facilitate nanoparticle-mediated charge transfer. Furthermore, rapid expansion of the nanoporous hydrogel significantly improves diffusion, markedly shortening the response time. Yet the core challenge is material-based: the irreversible deactivation of the glucose oxidase enzyme itself, which the incorporated nanocatalysts do not mitigate, undermining the claimed long-term robustness.

Fig. 4.

Partial implantation sensor strategy diagrams. a Schematic diagram of a smart contact lens for tear glucose, including a schematic of human eye plasma leakage and a schematic of a glucose biosensor [122]. b Schematic diagram showing the working principle of a subcutaneous, wireless, battery-free device for long-term monitoring of catecholamine dynamics in the brain [125]. Source: Reprinted with permission from ref [125]. Copyright 2023, with permission from American Chemical Society. c Overview of real-time hemodynamic monitoring and vascular medical SVG systems, including a schematic of the SVG system showing that the SVG is composed of a bottom PDMS substrate layer, a middle patterned graphene layer, and a top PDMS encapsulation layer [126]. Source: Reprinted with permission from ref [126]. Copyright 2025, with permission from American Chemical Society.

A single signal always has its limitations; therefore, the development of multifunctional smart eye sensors is a key direction for future research. For example, Xie et al. used inorganic γ-Fe₂O₃@NiO magnetic oxide nanosheets (MNS) as the sensing material to develop a multifunctional contact lens (MCL) capable of simultaneously monitoring biochemical glucose levels in tears, biophysical eye movements, and intraocular pressure [124]. In ocular implantable sensors, the use of nanomaterials can further modify the sensor surface, enhancing biocompatibility and sensing performance. Although multifunctional integration represents a strategic merging in this field, the risk of signal crosstalk—like how eye movements might interfere with IOP or glucose readings-hasn’t been systematically studied yet; the reported synchronized sensing capabilities are often demonstrated in isolated, controlled conditions rather than in the complex, overlapping stimulus environments of real eye movements.

The current literature on ocular implantable sensors reveals a pronounced discrepancy between the sophistication of nanomaterial engineering and the rigor of long-term in vivo validation. Strategies for enhancing conductivity and sensitivity have converged on metal-carbon hybrid networks, yet conflicting claims persist regarding the optimal balance between sensitivity and mechanical range. More critically, the biosafety of implanted nanomaterials-including detachment-induced inflammation, metal ion leaching, and the unknown chronic effects of nanoscale debris in the confined ocular space—remains a decisive yet insufficiently addressed barrier to clinical translation.

Brain

Brain-related diseases such as epilepsy, Parkinson’s disease, stroke, brain injury, and brain tumors are often accompanied by abnormal brain electrical activity, neurotransmitter imbalances, and dysfunction in the metabolic microenvironment. Flexible brain-implantable sensors, utilizing composite nanomaterials and further integrated with functionalized recognition layers or electrocatalytic interfaces, can not only detect neural electrical signals but also enable real-time monitoring of chemical indicators such as dopamine, glutamate, lactate, glucose, oxygen concentration, and pH [127].

Achieving long-term, stable integration between implantable sensors and soft, dynamic brain tissue is a critical prerequisite for brain sensor implantation. Conventional neurotransmitter sensing instruments predominantly rely on silica-encapsulated carbon fiber electrodes. These electrodes are hard and brittle, and the tunability of their sensing functionality is extremely limited. The introduction of stretchable conductive nanomaterials represents an important solution to this challenge. Li et al. introduced a soft and stretchable biosensing neural interface based on a graphene-elastomer matrix, named “NeuroString” [128]. Graphene, with its atomic-scale thickness, exhibits high mechanical compliance under bending, stretching, and twisting; however, it fractures at strains above approximately 5%. To address this issue, the researchers embedded a network of graphene nanofibers, laser-modified with Fe₃O₄ nanoparticles, into a SEBS elastomer matrix, achieving high levels of flexibility and stretchability while preserving graphene’s unique electrochemical properties. When implanted into the mouse brain, NeuroString can detect various monoamine neurotransmitters, including DA, NP, 5-HT, and EP, enabling long-term, stable, and multi-neurochemical sensing within the central nervous system. However, the gradual dissolution of Fe₃O₄ nanoparticles in the acidic peri-implant microenvironment may degrade electrocatalytic activity over time, a material-level stability concern that remains unaddressed. To address the buckling deformation that flexible probes are prone to during implantation, temporary stiffening coatings—such as pullulan or polyethylene glycol—have been employed to facilitate insertion, with the coating dissolving after implantation to restore flexibility [128, 129]. While this strategy effectively bridges the rigidity–flexibility conflict, the transient dissolution process itself may introduce chemical perturbations that are rarely characterized. Beyond the implantation stage, once the probe is in place, the nanomaterial-based sensing interface must maintain stable electrochemical performance amid brain pulsations; highly stretchable microelectrode arrays have been developed to conform to such continuous tissue motion. For example, Fang et al. developed a kirigami-based spiral microelectrode array, in which the polyimide substrate enables a three-dimensional kirigami structure and facilitates large-scale, high-throughput implantation [130]. The device exhibits up to 100% stretchability, allowing it to conform to the brain’s continuous movement and deformation, and successfully achieved stable single-neuron recordings for up to 36 weeks, though this stability has only been demonstrated in rodent models and may not directly translate to larger, more mobile brains. However, the long-term hydrolytic stability of polyimide at kirigami-cut edges—where mechanical stress concentrates—has not been systematically evaluated, posing a risk of insulation failure and signal crosstalk. Beyond flexibility, the concept of fully bioresorbable neural implants has been demonstrated using phosphorus-doped silicon nanomembranes as a transient conductive layer, offering a pathway to eliminate long-term foreign-body responses [131]. However, such bioresorbable materials face fundamental challenges in achieving predictable in vivo degradation kinetics due to pH and enzymatic variability, making their sensing reliability difficult to guarantee over a defined window.

Owing to their high specific surface area, nanomaterials enable enhanced sensing performance while reducing electrode dimensions. The sensor device designed by Stuart et al. represents the smallest and lightest (49 mg) wireless, battery-free, fully implantable neurochemical sensing platform to date (Fig. 4b) [125]. They constructed a working electrode using a composite material of single-walled carbon nanotubes and carbon fibers. Its enormous specific surface area endows it with sensing performance that ranks among the highest currently available (1264.1 nA·µM⁻¹·cm⁻²), enabling real-time recording of drug-induced changes in catecholamine levels in freely moving mice. The high specific surface area that enables this sensitivity, however, equally amplifies biofouling-induced signal drift-an intrinsic material trade-off that surface modifications have yet to fully resolve.

Future flexible brain implants are envisioned to evolve toward multimodal integration, intelligence, and adaptability - combining nanomaterials with flexible electronics to simultaneously monitor neurotransmitters, electrical signals, and metabolic markers, while leveraging artificial intelligence for data analysis and disease prediction. However, realizing this vision will require moving beyond the current trend of prioritizing acute implantation success and initial sensing performance. The long-term material - level degradation, progressive biofouling, and unresolved neuroinflammatory risks of nanomaterials within brain tissue remain critically underexplored, and these fundamental challenges - not insufficient sensitivity or channel count - now constitute the primary barrier to clinical translation. Meaningful progress will therefore depend on the development of biodegradable and self-powered materials with predictable degradation kinetics and well - characterized immunogenicity, enabling devices that are not only multifunctional and intelligent, but also genuinely stable and safe over clinically relevant timescales.

Cardiovascular

Compared with intermittent surface-based monitoring, implantable monitoring can directly capture key information, such as local pressure, blood flow, vascular strain, and myocardial electrical activity, near the site of cardiac and vascular lesions, offering significant value for risk prediction, postoperative follow-up, and device performance evaluation [132, 133]. Flexible sensors based on nanomaterials can accommodate the substantial deformation of cardiovascular tissues through stretchable conductive networks and ultra-thin film designs, while maintaining stable electrical output [134].

The blood environment is the primary factor limiting the long-term use of implanted sensors. When traditional electrochemical sensors meet blood, their stability rapidly deteriorates due to protein adsorption and cell deposition, leading to a decline in performance. Bionic nano-interface design is an effective method for improving sensor stability in blood. Drawing inspiration from the layered structure of the intestinal mucosa in living organisms, Chen et al. constructed a three-dimensional, dual-continuous, nanoporous gold interface and then coated it with a hyperbranched polyethylene glycol layer. This created a spatial separation effect similar to that of a mucin-glycan layer, effectively separating the target small molecules from larger interfering components while delaying the deposition of biomacromolecules and the degradation of the probe [135]. This device remains stable for one month in undiluted serum and can continuously monitor the concentration of the antibiotic kanamycin in the femoral vein of rats for four days without the need for calibration. Nevertheless, the long-term integrity of the hyperbranched PEG coating under continuous blood flow shear stress and enzymatic degradation remains unverified, and the gradual oxidation or restructuring of the nanoporous gold interface could progressively compromise its size-sieving function - a material-level degradation pathway that one-month serum stability data cannot capture. The integration of nanomaterials into implantable sensors and interventional devices can further expand the applications of implantable sensing in the cardiovascular system. Kwon et al. developed a wireless, battery-free vascular implant system by integrating silicon nanomembrane piezoresistive sensors into millimeter-scale probes. In this design, the atomically thin silicon nanomembrane undergoes deformation under vascular pressure and flow, with the resulting strain-induced band structure changes modulating its electrical resistance. The system simultaneously monitors pulmonary artery pressure, blood flow, aortic pressure, and left ventricular pressure in real time, then transmits the data remotely to a mobile device via Bluetooth Low Energy (BLE) technology. Its measurement accuracy is comparable to that of the clinical gold standard (Swan-Ganz catheter). By overcoming the limitations of traditional catheters that require bedside monitoring, this technology represents a promising wireless platform toward post-operative home monitoring, although its performance has only been validated in large animal models, not in human patients with heart failure, pulmonary arterial hypertension, and other conditions [132]. However, the long-term mechanical fatigue of silicon nanomembranes under the cyclic pulsatile stress of arterial blood flow and the resultant drift in piezoresistive response has not been systematically characterized, raising questions about whether the initial accuracy can be maintained over clinically meaningful implantation durations. Health monitoring of vascular grafts is a critical application for managing post-transplant complications. Ma et al. seamlessly integrated a laser-induced graphene flow sensor into an artificial blood vessel to create a smart vascular graft (Fig. 4c) [126]. This device can detect hemodynamic changes in real time and quantitatively diagnose the location and degree of occlusion (0-100%) of a thrombus. In a rabbit carotid artery model, it successfully monitored changes in the state of vascular occlusion. Yet the structural stability of laser-induced graphene under the continuous pulsatile deformation of a vascular graft—specifically, the potential microcrack propagation and delamination at the graphene–polymer interface—remains unexplored; such degradation could generate false flow signals that mimic thrombotic events. The key features of the representative implantable sensing platforms discussed above are summarized in Table 4 for direct comparison across different target organs and sensing strategies.

Table 4.

Comparison of representative flexible nanotechnology-enabled implantable sensing platforms.

Study	Substrate Material	Nanomaterial	Sensing Target	Sensing Mode	Sample Matrix / Validation Model	Key Performance and Advantages
[119]	PDMS	Liquid metal (EGaIn)	IOP	Piezoresistive (microfluidic, four-wire)	Rabbit vitreous humor	Sensitivity0.00253 Ω/mbar, Good cyclic repeatability (100 cycles)
[120]	Flexible contact lens (PDMS/WPU)	Ti₃C₂Tx MXene@carboxylated MWCNTs	IOP (for ICP)	Piezoresistive (Wheatstone bridge, microcrack)	Ocular surface (rabbit)	Sensitivity 33.21 mV/mmHg; 0–60 mmHg range; 96.6% transmittance; wireless Bluetooth;
[121]	Smart contact lens (silicone elastomer)	Au hollow nanowires + D-PEDOT	IOP	Piezoresistive (strain)	Rabbit eye (glaucoma)	Sensitivity > 75% at 200 mmHg; >30% stretchability; >84% transmittance; wireless, battery-free with ASIC; closed-loop drug delivery
[122]	Smart contact lens (silicone elastomer)	GOx / Prussian blue electrode	Tear glucose	Electrochemical enzymatic (amperometric, NFC)	Human tear (healthy/diabetic)	LOD 0.02 mM; 20 s interval; wireless NFC
[123]	Smart contact lens (nanoporous hydrogel)	HA-Au@Pt bimetallic nanocatalysts	Tear glucose	Electrochemical (GOx, nanoporous hydrogel)	Diabetic rabbit tear	Sensitivity 180 µA·cm⁻²·mM⁻¹; response 3.6 s; LOD 0.01 mg/dL; wireless
[124]	Multifunctional contact lens (Ecoflex)	γ-Fe₂O₃@NiO magnetic oxide nanosheets	Glucose, eye movement, IOP	Enzyme-free electrochem. + magnetic + inductive	Ocular surface (rabbit)	IOP sensitivity 0.17 MHz/mmHg; glucose LOD 0.43 µM; eye movement accuracy 95.27%; wireless IOP
[128]	SEBS elastomer	Laser-induced graphene nanofibers + Fe₃O₄ or NiO nanoparticles	DA, 5-HT, NP, EP	FSCV	Mouse brain (NAc, BLA, striatum) and mouse colon	LOD: DA 5.6 nM, 5-HT 3.5 nM, NP 7.2 nM, EP 6.6 nM; >1700% stretchability; 16-week stable recording; brain-gut dual-site sensing
[129]	Polyimide	Pt-black (electroplated on Au)	LFP and neural spike	Electrophysiological recording (MEMS)	Mouse brain (M1 cortex)	N/A (electrophysiological); bending stiffness 2.6 × 10⁻¹⁴ Nm²; 30-day recording; reduced glial scarring
[130]	Polyimide or parylene C	— (Au electrodes)	Single-neuron action potential	Kirigami-based microelectrode array (spiral threads)	Rodent brain and macaque monkey brain (V1, PMd, M1, S1)	N/A (electrophysiological); >700 neurons simultaneously; 36-week (rat) and 22-week (monkey) stability; parallel high-throughput implantation
[131]	PLGA (waveguide)	Phosphorus-doped Si nanomembrane + Mo bilayer (Mo/Si)	ECoG/LFP	Electrophysiological recording + optogenetic stimulation (460 nm)	Thy-1:ChR2 transgenic mouse brain	N/A (electrophysiological); fully bioresorbable (8 weeks); wireless power; 2-week operation
[125]	Carbon fiber + polyimide	SWCNT + CF/chitosan composite	Dopamine (catecholamines)	Electrochemical (chronoamperometry) + optogenetic stimulation (µ-LED)	Freely moving mouse brain (nucleus accumbens)	Sensitivity 1264.1 nA·µM⁻¹·cm⁻² ; 49 mg; wireless, battery-free; subdermal implant; simultaneous optogenetics + electrochemical recording
[135]	3D bicontinuous nanoporous gold (npAu)	Hyperbranched PEG (8-arm, MW 20k) coating + aptamer switches (methylene blue redox tag)	Kanamycin (small-molecule antibiotic)	Electrochemical (SWV, aptamer-based)	Rat femoral vein (blood)	stable in undiluted serum for 1 month; 7-day intravenous stability in freely moving rats; 4-day continuous pharmacokinetic monitoring
[132]	Silicon nanomembrane (on PI substrate)	— (Si-NM strain gauges, Au interconnects)	Pulmonary artery pressure, blood flow (bi-directional), aortic pressure, left ventricular pressure, temperature	Piezoresistive (Si-NM strain gauges)	Porcine (PA) and ovine (aorta, LV) models	GF = 38 (pressure), GF = 90 (flow); wireless, battery-free BLE; accuracy comparable to Swan-Ganz catheter; 200 Hz continuous
[126]	PDMS (encapsulation)	Laser-induced graphene (LIG, porous)	Blood flow / thrombus (occlusion detection, 0–100%)	Piezoresistive (strain sensing)	Rabbit carotid artery (in vivo)	Detection limit 0.0034% GF = 71.6 (0–50% strain), 805 (50–60% strain); >32,400 cycles stability; WiFi

The future development of cardiovascular implantable sensors will no longer rely on the simple addition of functions, but rather on the synergistic integration of biofouling resistance, long-term encapsulation performance, and dynamic tissue adaptability. However, the key material-level challenges that enable these capabilities—namely, the durability of anti-fouling coatings under flow conditions, the fatigue characteristics of materials under cyclic strain, and the long-term electromechanical stability in the blood environment - remain insufficiently investigated. Closing this gap between short-term performance and long-term material reliability is a prerequisite for clinical translation. Moreover, the vast majority of current studies remain at the proof-of-concept stage, validated in rodent or rabbit models over hours to weeks; large-animal studies and human clinical trials are virtually absent, and performance claims should be interpreted within this limited validation context.

Closed-loop sensing and therapeutic intervention

The integration of flexible nanotechnology with intelligent regulation is reshaping modern medicine by enabling a transition from static monitoring toward dynamic closed-loop diagnosis and therapy [136, 137]. By combining the biocompatible interfaces and functional properties of nanomaterials with artificial intelligence (AI)-assisted analysis, these systems facilitate real-time physiological monitoring, adaptive decision-making, and personalized therapeutic intervention [138, 139]. Applications in neurological disorders [140, 141], wound management [142, 143], and chronic disease treatment demonstrate the potential of closed-loop platforms to improve precision healthcare through continuous feedback and autonomous regulation.

In this review, closed-loop intervention systems are defined as integrated platforms consisting of three essential components: real-time physiological sensing, intelligent data-driven decision-making, and autonomous therapeutic response. As illustrated in Fig. 5, flexible nanotechnology-enabled closed-loop systems integrate multimodal physiological sensing, intelligent signal processing, and adaptive therapeutic intervention through a continuous feedback pathway.

Fig. 5.

Overview of flexible nanotechnology-enabled closed-loop healthcare systems, highlighting the integration of physiological sensing, AI-driven signal processing, and adaptive therapeutic intervention

Fully closed-loop platforms differ from conventional monitoring systems by establishing a feedback-regulated pathway that dynamically adjusts therapeutic strategies according to continuously acquired physiological signals. Systems lacking this autonomous intervention capability are therefore only supportive monitoring platforms, not complete closed-loop therapeutic systems.

Neurological disorders

The combination of flexible nanosensing technologies and neural closed-loop regulation provides a promising strategy for real-time monitoring and adaptive intervention in neurological disorders.

Zhou et al. utilized spindle-shaped carbon nanotube fibers (SSCNTFs) as a flexible substrate to construct a core-shell structured nano-molecularly imprinted polymer (nanoMIP) sensor (Fig. 6a). The system enables real-time monitoring of L-Dopa concentrations in interstitial fluid (ISF) and initiates adaptive drug delivery when drug levels deviate from the therapeutic range. By dynamically regulating infusion rates, the platform reduced the coefficient of variation (CV) and diffusion coefficient (DF) of ISF L-Dopa concentrations by 55.9% and 68.0%, respectively [144]. However, these promising results have only been demonstrated in rat models. Key challenges for clinical translation include the long-term stability of the nanoMIP layer in cerebrospinal fluid environments and the robustness of closed-loop decision-making algorithms against individual variations in pharmacokinetics. Mao et al. developed an AuNP-modified screen-printed carbon electrode (SPCE) for detecting amyloid-β oligomers (AβOs), a key biomarker associated with Alzheimer’s disease (AD). The AuNPs provide anchoring sites for thiolated aptamers through Au–S bonding, enabling specific recognition of AβOs on the electrode surface. Electrochemical detection was performed using differential pulse voltammetry (DPV). A miniature handheld analyzer transmitted quantitative detection results to a smartphone application via Bluetooth Low Energy. In mouse models, sensor responses showed strong correlation with injected AβOs concentrations (R² = 0.995, p < 0.001) [145]. However, this platform currently functions primarily as a diagnostic tool rather than a complete closed-loop therapeutic system, and its long-term antifouling capability under in vivo conditions remains to be validated. For movement disorder intervention, Qu et al. developed a wearable closed-loop system based on triboelectric sensors integrated into eyeglass frames. The system combines a butanol-modified melamine-formaldehyde (BMF)/CaCu₃Ti₄O₁₂ (CCTO) composite with microscale hemispherical structures. The incorporation of CCTO into the BMF matrix, together with the surface microstructures, achieves a 2.3-fold enhancement in triboelectric output performance. By continuously monitoring facial muscle activity, the system automatically activates electrical stimulation targeting the corresponding facial nerve region when abnormal signals are detected. Clinical trials demonstrated 98% accuracy in identifying hemifacial spasm (HFS), with improved symptom control and facial symmetry [146]. Zhong et al. developed a dual-layer flexible sensor platform for post-stroke aphasia rehabilitation. The system integrates VGr/PDMS strain sensors to detect low-frequency muscle deformation and Ag-MnO₂/ITO/PET pressure sensors to capture high-frequency vocal cord vibrations. This dual-material design—VGr for strain sensing and Ag-MnO₂ for pressure sensing—enables complementary detection of both low- and high-frequency mechanical signals associated with speech production. Collected signals are wirelessly transmitted to a terminal device, where a generative adversarial network (GAN)-based model analyzes signal features, predicts intended speech, and activates speech reconstruction algorithms (Fig. 6b) [147]. For epilepsy management, Li et al. developed a dual-modal closed-loop neuromodulation implant system (BCNIS) integrating ultra-flexible neural probes with subcutaneous drug capsules. The system employs MEMS-fabricated polyimide probes with Pt-Ir coating to achieve low-impedance neural recording, while a subcutaneous drug capsule enables on-demand diazepam release when seizure patterns are detected by the LSTM algorithm [148]. However, increased system complexity also introduces challenges related to long-term biocompatibility and implantation stability. Wei et al. developed SonoCap, a topological fiber-structured acoustic capacitor assembled from piezoelectric-dielectric composite nanospheres (UCapT) and two-dimensional cellulose. Under external ultrasound stimulation, the device converts acoustic energy into electrical signals through a piezoelectric-triboelectric hybrid mechanism. The resulting EStimX closed-loop system integrates deep learning algorithms to analyze cortical ECoG signals and automatically trigger ultrasonic stimulation upon seizure detection. Experiments in rats and pigs demonstrated effective suppression of temporal lobe seizures. After 14 days of implantation, no obvious inflammation or tissue damage was observed, indicating favorable short-term biocompatibility [149]. Nevertheless, the trade-off between ultrasound penetration depth and spatial resolution limits its potential for deep-brain regulation, and the 14-day observation period remains insufficient for evaluating long-term safety.

Fig. 6.

Representative closed-loop sensing and therapeutic intervention architectures. a Schematic diagram of the sensing mechanism of nano MIP shell and CNT [144]. b (i) Schematic diagram of the dual-sensor structure. (ii) Diagram showing the sensor detecting resistance changes caused by vocal cord vibrations and laryngeal muscle stretching, and wirelessly transmitting the data to a mobile terminal [147]. Source: Reprinted with permission from ref [147]. Copyright 2025, with permission from John Wiley and Sons. c (i) Construction and assembly of multi-scale component groups. (ii) Microstructure of fiber membranes. (iii) Mechanism of RMF-induced generation of wireless microcurrents [153]. Source: Reprinted with permission from ref [153]. Copyright 2025, with permission from John Wiley and Sons. d The structure of smart contact lenses used for treating glaucoma [121].

Overall, flexible nanotechnology-enabled closed-loop neural systems have demonstrated considerable potential for neurological intervention [150, 151]. However, most existing platforms remain at the preclinical validation stage, with limited evidence from long-term implantation studies [152]. Chronic neurological applications require biointerfaces capable of maintaining stable signal acquisition and stimulation performance under complex physiological conditions. Furthermore, the lack of validation in large-scale animal models, such as non-human primates, remains a significant barrier to clinical translation. Future development should therefore focus not only on improving sensing performance but also on long-term biointerface engineering, adaptive edge-AI algorithms, and clinically relevant validation strategies.

Smart trauma management and targeted drug delivery

The integration of flexible nanomaterials with closed-loop regulation has provided new strategies for intelligent wound management and targeted drug delivery. Unlike conventional passive dressings, these systems combine physiological monitoring with therapeutic functions, enabling more adaptive responses to dynamic wound environments.

Multifunctional conductive hydrogels represent an important platform for integrated wound sensing and treatment. Si et al. developed a C-CST/PMC conductive organohydrogel using polyvinyl alcohol (PVA) as the matrix, incorporating silk nanofibers (SNFs), catechol-modified chitosan (C-CS), tannic acid (TA), and Mg²⁺/Ca²⁺-mediated conductive networks. This multi-component design integrates multiple functions into a single platform: the hydrogel enables real-time monitoring of mechanical deformation and wound microenvironmental changes while simultaneously providing antimicrobial and antioxidant effects to regulate inflammation, promote angiogenesis, and enhance collagen deposition. In a diabetic rat model, the system achieved a wound healing rate of 97.54% after 14 days, demonstrating the potential of integrated “monitor-and-treat” wound platforms [154]. However, interactions between different functional components and possible interference between sensing and therapeutic modules require further investigation. Liu et al. developed a smart electronic skin patch (TENG-gel) based on a triboelectric nanogenerator (TENG). The device consists of a three-layer flexible structure incorporating PDMS/PTFE films, an E-GaIn conductive layer, and HQPS@MoS₂. The MoS₂ nanosheets contribute to antibacterial activity through three mechanisms: physical membrane disruption, peroxidase-like catalytic effects that generate reactive oxygen species, and photothermal enhancement. The system achieved more than 95% inhibition against both E. coli and S. aureus, while triboelectric-generated microcurrents promoted tissue regeneration. In a full-thickness skin defect mouse model, the wound healing rate reached 96.8% ± 1.4% after 14 days [155]. Although self-powered operation represents a major advantage, triboelectric output remains sensitive to environmental factors, including humidity and mechanical deformation, which may affect long-term reliability in practical applications. Magnetic field-mediated regulation provides another approach for externally controlled wound therapy. Zhou et al. developed a PMB fiber membrane consisting of PCL and Ti₃C₂T_x MXene composites, which enables multimodal monitoring of wound-related signals, including strain, temperature, and exudate changes. A portable rotating magnetic field (RMF) generator induces wireless microcurrents within the membrane, promoting pro-healing cellular pathways while suppressing inflammatory responses. In diabetic rat models, the system improved wound healing by 36.3% and enhanced nerve regeneration by 283.8% (Fig. 6c) [153]. However, the energy consumption and miniaturization of the RMF generator remain unresolved challenges. Moreover, the reported enhancement in nerve regeneration was obtained under specific experimental conditions, and its reproducibility and applicability to human wound healing require further validation. Li et al. developed an anisotropic hydrogel (Gel-ODex-MPG, GOH-MPG) containing magnetoelectric nanoplates composed of Fe₃O₄-PDA-rGO. The integration of these three functional components—magnetic Fe₃O₄, photothermal PDA, and conductive rGO—enables the hydrogel to respond to multiple external stimuli. Magnetic-field-induced alignment improved both mechanical anisotropy and electrical conductivity, while near-infrared (NIR)-triggered photothermal effects enabled on-demand antibacterial treatment. The conductive network also provided strain-sensing capability for real-time wound monitoring. In animal experiments, the GOH-MPG plus NIR treatment achieved a wound healing rate of 98.6% within 14 days [156]. Despite its multifunctionality, further studies are required to evaluate long-term safety and adaptive regulation under clinically relevant wound conditions. Microneedle-based platforms provide another promising strategy for closed-loop drug delivery through minimally invasive sensing and intervention. Fu et al. developed a 3D-printed barbed microneedle system integrating Ag/AgCl and gold nanoparticles (AuNPs). The AuNPs are deposited on the microneedle electrode surface to improve the electrochemical sensing interface for impedance-based wound monitoring. The platform consists of solid barbed microneedles (3D-BMN) for real-time wound impedance monitoring and hollow barbed microneedles (3D-BHMN) equipped with an ultrasonic nebulizer for controlled drug delivery. When abnormal impedance changes are detected, the system enables condition-responsive drug release. Gel wound model experiments demonstrated an impedance monitoring error below 0.22% and improved delivery efficiency [157]. Nevertheless, potential risks associated with microneedle fracture, skin irritation, and long-term implantation safety remain important considerations for clinical translation.

Taken together, flexible nanotechnology is transforming wound management from passive treatment toward more active and responsive therapeutic systems [158]. Current studies have achieved significant progress in multifunctional integration, real-time sensing [159, 160], and localized treatment. However, most smart wound platforms remain primarily validated in simplified diabetic animal models, which do not fully reproduce complex clinical wound environments involving mixed infections, variable exudate composition, and patient-specific healing conditions. Furthermore, many systems combine sensing and therapeutic functions without achieving fully autonomous closed-loop regulation, as intervention decisions are often based on predefined stimulation conditions rather than adaptive responses to continuously monitored biological signals. Future development should therefore focus on environmentally stable and anti-biofouling materials, clinically relevant validation models, and intelligent control algorithms capable of dynamically optimizing therapeutic responses.

Chronic disease management and personalized treatment

Chronic diseases require continuous physiological monitoring and personalized therapeutic strategies due to their complex progression. Flexible nanomaterial-based platforms provide opportunities for long-term biomarker detection, physiological assessment, and adaptive intervention. However, achieving fully autonomous closed-loop healthcare remains challenging due to limitations in long-term stability, energy supply, biosafety, and clinical validation.

Nie et al. developed a flexible sweat sensing platform based on laser-induced graphene (LIG) electrodes and deep eutectic solvent (DES)-based ionogels for personalized diabetes management. The LIG electrodes serve as a closed bipolar electrode array with high conductivity and large specific surface area, while the DES-based ionogels function as stable electrochemiluminescence (ECL) emitters that enhance signal stability and device flexibility. The system integrates multiple sensing modules to continuously monitor glucose-related physiological information and other biochemical parameters in sweat, providing dynamic metabolic information through a wearable interface [161]. By combining flexible sensing technology with intelligent analysis, the platform demonstrates potential for personalized health monitoring and disease management. However, this system primarily functions as a physiological monitoring and prediction platform rather than a complete closed-loop therapeutic system. Further integration with validated therapeutic modules and adaptive control strategies will be required to transform such monitoring platforms into autonomous closed-loop systems. Fang et al. developed a machine learning‑assisted textile triboelectric sensor. It uses fabric as the substrate and converts arterial pulsations into electrical signals through the contact‑separation effect between CNTs and FEP. The sensor has a high signal-to-noise ratio, responds quickly, and offers high sensitivity. It can stably capture pulse waves during exercise, at rest, and even in humid environments. When paired with machine learning algorithms to analyze parameters like heart rate (HR) and vascular stiffness, the system can perform non‑invasive blood pressure measurement. Data is sent wirelessly to a custom mobile app and then uploaded to the cloud for analysis, providing an integrated solution for continuous data acquisition, physiological assessment, and personalized health management [162]. However, this platform mainly provides physiological assessment rather than autonomous therapeutic regulation. In addition, the dependence on machine-learning models requires further validation under diverse populations and long-term real-world conditions to ensure robustness and reliability. Li et al. developed a smart monitoring system based on AgNP/rGO/PEDOT: PSS/TPU nanofiber electrodes. The composite electrode integrates a conductive network of AgNP/rGO with PEDOT: PSS as a self-healing material, supported by a TPU substrate. The highly sensitive electrodes collect ECG and EMG signals in real time and maintain signal clarity even during physical activity. The data is then transmitted wirelessly to a terminal for analysis. The material’s self-healing capability achieves a 98.3% damage recovery rate under 200% strain, and its resistance retention rate stays at 95.5% in a sweat-exposed environment. This makes it well suited for long-term, continuous monitoring [163]. The self-healing property significantly extends device lifespan, but the reported 98.3% resistance recovery rate was measured under ideal conditions; the decay pattern of healing efficiency after repeated cycles in actual wear remains unclear. Kim et al. developed a smart contact lens system for wireless intraocular pressure (IOP) monitoring. The system integrates hollow gold nanowire (AuHNW) sensors, a flexible drug delivery system (DDS), wireless circuitry, and an ASIC chip. Monitoring data is sent through the wireless circuitry to the ASIC chip for real‑time analysis, enabling continuous IOP monitoring. When the IOP exceeds a safe threshold, the system automatically triggers the flexible DDS, which then releases the pressure-lowering drug timolol in a controlled manner. This release is achieved by electrochemical dissolution of the gold film channels and SU-8 photoresist (Fig. 6d) [121]. However, validation of this system remains limited to animal studies, and no clinical evidence has yet demonstrated its long-term safety or therapeutic effectiveness in humans.Challenges including the lag between tear-fluid measurements and actual IOP changes, the balance between optical transparency and electronic integration, and long-term ocular biocompatibility remain important barriers for clinical translation.

In summary, flexible nanotechnology-enabled platforms have expanded the scope of chronic disease management by enabling continuous physiological monitoring, personalized health assessment, and targeted therapeutic intervention. Compared with conventional clinical measurements, these systems offer improved adaptability through wearable interfaces, wireless communication, and intelligent data analysis. However, most current platforms remain primarily focused on monitoring and prediction, while fully autonomous therapeutic feedback systems are still limited. Further advances in long-term stability, energy management, and clinically validated therapeutic integration will be essential for translating these technologies toward practical personalized healthcare applications.

Overall, flexible nanotechnology-enabled platforms demonstrate significant potential for advancing personalized healthcare by integrating real-time sensing, intelligent analysis, and therapeutic intervention. As summarized in Table 5, representative systems have achieved diverse functionalities, including physiological monitoring, wireless data transmission, AI-assisted analysis, and targeted therapeutic regulation. Nevertheless, most reported platforms remain at the stages of biomarker detection, physiological prediction, or externally assisted regulation rather than fully autonomous closed-loop therapy. Current validation is largely limited to in vitro experiments, animal models, or preliminary small-scale human evaluations, while regulatory-approved clinical applications remain rare. Future advances will require the integration of stable nanomaterials, reliable power management, robust AI-driven decision algorithms, and clinically validated therapeutic modules to bridge the gap between experimental demonstrations and practical healthcare applications.

Table 5.

Comparison of representative flexible nanotechnology-enabled closed-loop sensing and therapeutic platforms

Study	Substrate Material	Nanomaterial	Sensing Target	Sensing Mode	Sample Matrix / Validation Model	Key Performance and Advantages
[144]	SSCNTs	Core-shell nanoMIP	L-Dopa	Electrochemical	Interstitial fluid (ISF) / Rat model	CV and DF reduced by 55.9% and 68.0%, respectively; adaptive drug delivery based on detected L-Dopa levels
[145]	SPCEs	AuNPs	AβOs	DPV	Mouse model	R² = 0.995, p < 0.001; Bluetooth Low Energy (BLE) wireless transmission and smartphone-based analysis
[146]	Eyeglass frame	BMF/CaCu3Ti4O12(CCTO) composite	Facial muscle activity	Triboelectric sensing	Facial skin / clinical trial	2.3-fold output enhancement; 98% accuracy; integrated electrical stimulation module for therapeutic intervention
[147]	PDMS, ITO/PET	VGr, Ag-MnO₂	Laryngeal muscle stretching and vocal cord vibration	Strain + pressure sensing	Human laryngeal monitoring	GAN-based speech reconstruction; Bluetooth transmission and AI-assisted signal analysis
[148]	Polyimide	Pt-Ir alloy coating	Epileptic seizure (ECoG)	ECoG sensing + LSTM analysis	KA-induced epilepsy mouse model	Detection accuracy of 94%; seizure suppression of 75.16%; integrated electrical stimulation and drug capsule delivery
[149]	Topological fiber + 2D cellulose	Piezoelectric-dielectric composite nanospheres (UCapT)	Epileptic seizure (ECoG)	Ultrasound-activated capacitive discharge	Epidural rat/pig model	No inflammation or tissue damage after 14 days; deep-learning-assisted detection; ultrasound-powered stimulation
[154]	PVA hydrogel (EG/H₂O)	SNFs, C-CS, TA, Mg²⁺/Ca²⁺	Wound strain and inflammatory factors	Strain sensing + antibacterial/antioxidant function	Diabetic rat wound model	14-day healing rate of 97.54%; integrated wound monitoring and therapeutic regulation
[155]	PDMS/PTFE, E-GaIn conductive layer	HQPS@MoS2	Wound status	TENG + photothermal/nanoknife antibacterial therapy	Mouse full-thickness skin defect model	400% strain tolerance; adhesion ≈ 5 kPa; antibacterial efficiency > 95%; healing rate of 96.8 ± 1.4%; self-powered microcurrent generation
[153]	PCL composite fiber membrane	Ti3C2Tx MXene	Strain, temperature, and exudate	Resistive sensing + RMF-induced wireless microcurrent	Diabetic rat model	Healing improvement of 36.3%; nerve regeneration enhancement of 283.8%; wireless microcurrent stimulation
[156]	GOH hydrogel	Fe3O4-PDA-rGO magnetoelectric nanoplates	Wound strain	Strain sensing + NIR photothermal	Animal wound model	14-day healing rate of 98.6% with NIR treatment; NIR-triggered on-demand antibacterial therapy
[157]	3D-printed barbed microneedles	Ag/AgCl, AuNPs	Wound impedance	Closed-loop impedance feedback	Gel wound model	Impedance monitoring error < 0.22%; feedback-controlled ultrasonic nebulization drug delivery
[161]	LIG flexible electrode array	Deep eutectic solvent (DES) ionic gel ECL matrix	Sweat glucose	ECL c-BPE	Human volunteer sweat test	Results consistent with commercial test kits; microfluidic platform with smartphone-assisted readout
[162]	Textile substrate	CNTs/FEP	Pulse wave	Triboelectric sensing + machine learning	Human wearable test (exercise/rest/humidity conditions)	Wireless physiological monitoring and cloud-based AI analysis for personalized health assessment
[163]	TPU substrate	AgNP/rGO/PEDOT: PSS conductive network	ECG/EMG signals	Electrophysiological sensing	Human skin wearable test	Self-healing recovery of 98.3% under 200% strain; sweat resistance retention of 95.5%; wireless ECG/EMG data transmission

Summary and outlook

In summary, this review highlights a significant transition in biomedical monitoring: from intermittent, lab-based tests to flexible, nanomaterial-enabled systems that can continuously sense, analyze, and even intervene in real time. This paper systematically reviews flexible sensing technologies based on nanomaterials for biomedical monitoring, covering innovations in underlying architectures ranging from flexible substrates (polymers, paper-based materials, hydrogels, etc.) to multidimensional nanomaterials (0D, 1D, 2D). It also delves into their multi-level application trajectories, spanning from surface-worn devices and deep-tissue implants to intelligent closed-loop regulation. Leveraging the exceptional mechanical conformability, highly customizable biointerfaces, and superior signal transduction capabilities of nanomaterials, flexible sensors have successfully overcome the physical limitations of traditional rigid medical devices, laying a solid technological foundation for modern medicine’s transition from mere “disease diagnosis” to “integrated diagnosis and treatment”.

However, as this field transitions from the laboratory to clinical translation and commercial application, it still faces numerous significant challenges. Nanomaterials in flexible substrates have poor long-term dispersion stability. Their anti-agglomeration ability is also weak. Agglomeration causes signal drift. This signal drift directly reduces sensing reliability. Batch-to-batch synthesis reproducibility is limited. Scalable production capacity is also limited. Therefore, device-to-device performance consistency is poor. This hinders the transition from laboratory research to mass production. The long-term biocompatibility of nanomaterials in implantable scenarios is not fully understood. Key aspects include degradation behavior, metabolic pathways, and potential nanotoxicity. This lack of understanding is a critical obstacle to long-term safe application.

In the context of wearable devices on the body’s surface, long-term use leads to mechanical fatigue, interference from sweat, signal drift, and limitations in power supply and transmission, all of which continue to compromise the reliability of long-term monitoring. In complex internal environments such as the brain, eyes, and cardiovascular system, implantable devices must not only withstand the harsh physiological microenvironments characterized by high humidity and high salinity but also overcome the mismatch between soft and hard interfaces caused by the dynamic movement of organs. Biofouling and tissue encapsulation can easily lead to signal distortion or even device failure. Furthermore, as systems evolve toward higher-level closed-loop regulation, ensuring long-term biocompatibility, achieving high-precision closed-loop signal feedback, and overcoming limitations in sustainable energy supply represent the core challenges that urgently need to be addressed.

Looking ahead, the ultimate form of flexible nanosensors will undoubtedly be intelligent closed-loop control systems that achieve a high degree of integration between monitoring and diagnosis/treatment. To achieve this goal, future research should not be limited to improving individual sensing capabilities but should focus on optimizing the deep synergy among materials, artificial intelligence, and power supply systems. First and foremost, establishing a long-term stable interface that dynamically adapts to tissue and resists biofouling is a prerequisite for closed-loop systems. Future research must be dedicated to developing long-term encapsulation technologies and smart responsive materials with adaptive and anti-biofouling properties, ensuring that implanted devices can achieve long-term, high-fidelity signal acquisition and non-invasive, on-demand drug delivery or therapeutic stimulation in complex bodily fluid environments. Second, we must promote the deep integration of flexible nanotechnology with embedded Edge AI. In the intervention of neurodegenerative diseases, intelligent trauma management, and chronic disease management, future flexible closed-loop sensors must possess robust local computing power to decode multimodal physiological/pathological signals in real time via AI algorithms and automatically generate precise intervention strategies, such as adaptive adjustment of drug delivery rates, to achieve seamless, dynamic diagnostic and therapeutic feedback independent of large external devices. Finally, breakthroughs in self-powered and wireless integration technologies are essential. Integrating miniaturized self-powered strategies with low-power wireless transmission technologies to address long-term life battery challenges is a critical pathway for closed-loop systems to advance toward clinical application.

Abbreviations

ELISA

Enzyme-Linked Immunosorbent Assay

PCR

Polymerase Chain Reaction

PI

Polyimide

PC

Polycarbonate

PEN

Polyethylene Naphthalate

PDMS

Polydimethylsiloxane

PU

Polyurethane

SEBS

Styrene-Ethylene-Butylene-Styrene

TPU

Thermoplastic Polyurethane

PVA

Polyvinyl alcohol

κ-CA

κ-carrageenan

GNP

Graphene Nanoplatelets

PCC

Hydrogel composite

AuNPs

Gold nanoparticles

PET

Polyethylene terephthalate

CC

Carbon fabric

TA-APTES

Tannic acid-3-aminopropyltriethoxysilane

N-Gr

Nitrogen-doped graphene

ITO

Indium tin oxide

AgNPs

Silver nanoparticles

GOx

Glucose oxidase

LOD

Limit of detection

QDs

Quantum dots

GQDs

Graphene quantum dots

CQDs

Carbon quantum dots

MQDs

Metal quantum dots

AgNWs

Silver nanowires

AuNTs

Gold nanotubes

PtNWs

Platinum nanowires

PtNPs

Platinum nanoparticles

APBA

3-aminophenylboronic acid

CNC

Cellulose nanocrystal

MOFs

Metal-organic frameworks

PEC

Photoelectrochemical

CNFN

Carbon nanofiber network

TMDCs

Transition metal dichalcogenides

MWCNTs

Multi-walled carbon nanotubes

LIG

Laser-induced graphene

HNPC

Hierarchical Nitrogen-doped Porous Carbons

TENGs

Triboelectric nanogenerators

GO

Graphene oxide

PG

Porous graphene

GF

Foamed graphene

PPy

Polypyrrole

PCL

Polycaprolactone

IOP

Intraocular pressure

AuHNW

Gold hollow nanowire

BiNCs

Bimetallic nanocatalysts

CGM

Continuous glucose monitoring

MCL

Multifunctional contact lens

DA

Dopamine

5-HT

5-Hydroxytryptamine

NP

Neuropeptide

FEP

Fluorinated Ethylene Propylene

PEDOT:PSS

poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate)

Author contributions

YZ, SZH and CYW contributed equally to this work. YZ, SZH and CYW conceived the work. YZ, SZH and CYW organized the content and prepared the draft manuscript. HYZ, GCZ and YL supervised the work. All authors contributed to the writing and revision of the manuscript.

Funding

This work was supported by the project supported by National Training Program of Innovation and Entrepreneurship for Undergraduates (202510631039, 202510631031), the Project of Excellent Medical undergraduate Tutoring System in Lab Teaching & Management Center in Chongqing Medical University (LTMCEMTS202626, LTMCEMTSDC202617).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval

Not applicable.

Consent to participate

Not applicable.

Consent to publish

Not applicable.

Competing interests

The authors declare no competing interests.