Seamless human electronics interfacing through advanced skin attachable and implantable sensor technologies

Abstract

The transition from episodic clinical assessment to continuous physiological monitoring represents a transformative shift in biomedical engineering. This review comprehensively examines recent advancements in human-interfaced sensing systems, categorized into skin-attachable and implantable platforms. We analyze innovations in interface architectures designed to resolve the mechanical mismatch between rigid electronics and soft biological tissues. Key strategies include the utilization of elastomeric substrates, functional hydrogels, and high-performance nanomaterials to ensure mechanical compliance, biocompatibility, and stable electrical contact. Furthermore, this paper details diverse signal measurement modalities, ranging from electrochemical biosensors for metabolic biomarkers (e.g., glucose, lactate) to electrophysiological recordings (ECG, EEG, neural probes) and physical sensing (strain, pressure). Representative applications including smart contact lenses and wireless mouthguards highlight the potential of these technologies in real-world diagnostics. Despite significant progress in wireless telemetry and miniaturization, challenges regarding long-term biostability, enzymatic degradation, and the foreign body response persist. We conclude by discussing the future trajectory toward autonomous, closed-loop theranostic systems integrated with artificial intelligence and biodegradable materials, paving the way for ubiquitous, personalized health management.

Introduction

The boundary between humans and electronics is rapidly being reduced as personalized medicine, artificial intelligence (AI), and human–machine interfaces continue to expand to daily life [1–4]. Continuous health monitoring, neural prosthetics, and emotion-aware systems are transforming the way biological information is sensed, processed, and utilized [5–9]. The combination of biomedical sensing and intelligent computing has led to the development of human-interfaced electronics that can collect physiological signals directly from the skin or from internal tissues for real-time diagnosis and feedback [10–15]. As these technologies expand from medical use to daily healthcare, rehabilitation, and cognitive support, the need for close and reliable connection between the human body and electronic devices is expected to grow [16].

However, achieving such close integration is still very difficult. The human body is a soft, moist, and ion-rich environment, but conventional electronic systems are rigid, dry, and rely on interface and electron conduction [17, 18]. This difference in mechanical and electrical properties often causes strain, delamination, and unstable electrical contact [19–21]. The ionic and reactive nature of tissues can lead to corrosion, impedance drift, and poor responses around the interface [22–25]. These issues reduce sensing accuracy and limit the operational lifetime of the device, especially when it is attached to or implanted in moving, sweating, or deforming tissue. To overcome these challenges, materials and device structures must be designed to match the softness, bio-material stability, and biological compatibility of human tissues [26–30]. Recently, many studies have reported using biocompatible organic materials-based electronic devices with various structures [21, 31–43].

Based on the above technology requirements, developing human-interfaced sensors requires balancing sensitivity, selectivity, and biocompatibility [44–47]. These three factors are closely related and often conflict with one another. Materials with high sensitivity may be unstable, while improving selectivity can reduce flexibility or speed [48, 49], Making devices more biocompatible may lower their electrical performance [50, 51], Owing to these trade-offs, the key is not to maximize one property but to design systems that combine all three in harmony.

Considering the mentioned circumstances, this review provides an integrated view of how this human-interfaced sensors can be achieved (Scheme 1). We discuss materials and interface structures that reduce mechanical and chemical mismatch, with a focus of different types of sensing signals such as chemical, electrical, mechanical, and optical, and highlights system-level integration including wireless operation, self-healing materials, and AI-based signal analysis. Section  2 reviews skin-attachable human-interfaced sensors, emphasizing nanostructured materials and soft device architectures that enable conformal contact with the epidermis for reliable signal acquisition. Section  3 then focuses on implantable sensing platforms, highlighting advanced interface architectures, encapsulation strategies, and in vivo demonstrations for long-term biological integration. Finally, we discuss emerging challenges and future opportunities toward fully integrated human-interfaced bioelectronic systems.

Scheme 1.

Schematic illustration of applications for human-interfaced attachable and implantable sensors. (a) Biosensors.; Reproduced with permission [161]. Copyright 2023, MDPI. (b) Mechanical sensors.; Reproduced with permission [163]. Copyright 2023, Wiley-VCH GmbH. (c) Photo sensors.; Reproduced with permission [145]. Copyright 2025, Springer Nature. (d) Electrical sensors.; Reproduced with permission [162]. Copyright 2022, Springer Nature. (e) Electrophysiological sensors.; Reproduced with permission [243]. Copyright 2023, Nature Publishing Group. (f) Smart contact lens.; Reproduced with permission [214]. Copyright 2022, Nature Publishing Group. (g) Mouthguard biosensors.; Reproduced with permission [245]. Copyright 2016, Elsevier

Human-interfaced attachable sensors

Interface architecture: materials, design, and integration

Skin-attachable biosensors require interfaces that achieve mechanical compliance with skin’s dynamic deformation (15–30% strain), maintain stable electrical contact, and ensure biocompatibility during prolonged wear [52–54]. Elastomeric polymers serve as primary substrates, with polyurethane (PU) offering exceptional elasticity (500–800% elongation) and biocompatibility, while polydimethylsiloxane (PDMS) provides gas permeability and optical transparency (> 90%) despite a higher elastic modulus (0.5-3 MPa) than epidermis (~ 140–600 kPa) [10, 55, 56]. Functional hydrogels bridge rigid electronics and soft tissues with tissue-like mechanical properties, with Young’s modulus of 1-100 kPa and ionic conductivity of 0.1–10 S m^–1, where conductive hydrogels incorporating PEDOT: PSS enable simultaneous compliance and electrical functionality, while adhesive hydrogels utilizing catechol chemistry achieve strong bonding to sweaty skin [53, 55, 57, 58].

Nanomaterial-based conductive composites enable stretchable electrodes maintaining functionality under deformation. Carbon nanotubes (CNTs) provide high conductivity (> 10⁴ S cm^–1) with mechanical strength, silver nanowires (AgNWs) achieve sheet resistance below 10 Ω sq^–1 at > 90% transmittance, and MXenes combine metallic conductivity (~ 15,000 S cm^–1) with electromagnetic interference shielding [59–61]. In addition to conductive performance, sustainable biomaterials, including silk fibroin, cellulose nanofibrils, and chitosan, offer biodegradability, addressing environmental concerns [53, 62]. Ultrathin architectures (100–300 nm) reduce bending stiffness, enabling conformability, while serpentine and fractal geometries achieve > 100% stretchability through buckling mechanisms [54, 56, 63]. Manufacturing technologies, including laser ablation, screen/inkjet printing, and roll-to-roll processing, enable high-throughput production, while island-bridge architectures isolate strain-sensitive components on rigid islands connected by stretchable bridges [54, 64].

The functional integration of these materials is governed by their ability to bridge the mechanical and physiological gap between rigid electronics and soft biological tissues. Elastomeric polymers such as PDMS and PU are selected not only for their flexibility but for their entropic elasticity, which allows them to undergo large, reversible deformations (up to 30% strain) that match the dynamic movements of human skin [65]. This mechanical compliance minimizes interfacial shear stress, thereby preventing delamination and ensuring stable sensor-to-skin contact. Furthermore, the inherent gas permeability of these elastomers facilitates a vapor-venting mechanism, allowing for sweat evaporation and oxygen transport, which is critical for preventing skin maceration during long-term monitoring [66]. In contrast, natural polymers like silk fibroin provide a mechanism for structural programmability [67]. By modulating the content of crystalline β-sheet domains, the mechanical toughness and biodegradation rates can be precisely tuned to match the target tissue’s remodeling cycle [68, 69]. Additionally, hydrogels serve as an aqueous bridge that mimics the extracellular matrix environment [70]. Their primary mechanism involves facilitating high-fidelity ion-to-electron signal transduction at the biotic-abiotic interface, which reduces contact impedance and enhances the signal-to-noise ratio for electrophysiological sensing [71].

Wireless power and data transmission through near-field communication (NFC) or Bluetooth Low Energy eliminates batteries and wires, with stretchable coil antennas maintaining functionality during deformation [72–74]. Self-healing capability through reversible dynamic bonds achieves > 90% healing efficiency, extending device lifetime [56, 60, 75] while nanoscale encapsulation (parylene-C, atomic layer deposited oxides) provides moisture barriers, maintaining flexibility [54, 60] Beyond mechanical and structural considerations, interface performance critically depends on electrical and electrochemical characteristics. Electrochemical stability employs surface modifications, differential measurements, and machine learning algorithms, minimizing drift and interference [54] Impedance minimization at skin-electrode interfaces is critical for biopotential measurements, where wet electrodes achieve 5–50 kΩ at 10 Hz but suffer gel drying, while dry electrodes with nanostructured surfaces reduce impedance 10–100 fold, and conductive hydrogel electrodes offer intermediate performance with self-adhesion [52, 55].

Signal measurement modalities

Chemical biosignals

Chemical biosensors detect molecular species in sweat, interstitial fluid (ISF), or blood through electrochemical transduction. Sweat provides an accessible sampling containing glucose (10–200 µM), lactate (5–25 mM) and electrolytes, though composition varies with flow rate [6, 57, 76, 77] ISF accessed via microneedles (300–800 μm) mirrors blood plasma, with glucose tracking blood levels with 5–15 min delay [77, 78] Electrochemical sensing employs three primary methods: amperometry, measuring oxidation/reduction currents, achieving micromolar detection limits, potentiometry measuring Nernstian potential changes (59 mV decade^–1), providing wide dynamic ranges, and impedimetric sensing detecting binding-induced impedance changes achieving femtomolar sensitivity [57, 59, 76, 77] Sweat pH (4.5-7.0) is measured using polyaniline electrodes with near-Nernstian response (50–60 mV pH^–1), metal oxide semiconductors (IrO₂ and RuO₂) providing robust sensing with minimal drift, and ion-selective field-effect transistors (ISFETs) integrating sensing with amplification for spatially resolved pH mapping [57, 59]. Sweat lactate (5–25 mM during exercise) monitored through lactate oxidase-based sensors indicates metabolic state [79–82], with Prussian blue-modified electrodes achieving 20–50 µA mM^–1 sensitivity [83].

Continuous glucose monitoring (CGM) via subcutaneous microneedle sensors using glucose oxidase represents the most successful commercial wearable biosensor, achieving a mean absolute relative difference below 10% with 7–14 day operational lifetimes limited by enzyme degradation and biofouling [77, 84]. Non-invasive sweat glucose faces challenges from 100-fold lower concentrations and inconsistent blood correlation [85–87], where machine learning models incorporating multiple parameters improve estimation accuracy [88–91]. Advanced electrode materials including CNT-modified electrodes enable operation at lower potentials (0.2–0.4 V) reducing interference; [92, 93] gold nanoparticles facilitate direct electron transfer eliminating oxygen dependence [94–97], and Prussian blue minimizes interferences at 0 V potential [98–100]. Glucose dehydrogenase enzymes offer superior stability and oxygen independence versus glucose oxidase [101, 102], while non-enzymatic sensors eliminate protein instability but suffer from fouling requiring and frequent calibration [103]. Beyond single-analyte detection, integrated sensor systems enable comprehensive metabolic monitoring. Spatial multiplexing integrates 3–6 sensors targeting multiple analytes within 1–5 cm² areas, where screen-printing enables sequential deposition and microfluidic channels ensure temporal synchronization [57, 76, 104]. Correction algorithms address cross-sensitivities through temperature compensation using Arrhenius correction (Q₁₀ values of 1.5–2.5) [105, 106] and simultaneous pH measurement for enzymatic sensor correction [89, 107, 108].

Regardless of target analyte or sensing mechanism, all electrochemical sensors face common challenges in maintaining long-term stability. Anti-fouling strategies prevent protein adsorption using poly(ethylene glycol) (PEG) brushes reducing adsorption by more than 90% and zwitterionic polymers demonstrating superior performance [53, 59]. Membrane barriers including Nafion selectively transport target molecules while excluding interferents [109–112]. Electrically regenerable sensors apply potential pulses restoring more than 80% sensitivity [113–116]. Enzyme stabilization through crosslinking [117], chemical modification with PEG (PEGylation) [118, 119], and co-immobilization with stabilizers [120] extends lifetimes, though enzyme-based sensors typically exhibit 1–14 day operational lifetimes, while non-enzymatic sensors offer extended lifetimes but lower selectivity [110–112, 121].

Electrical biosignals: electrocardiography (ECG), electromyography (EMG), and electrooculography (EOG)

ECG measures cardiac electrical activity through skin-surface potentials, with traditional wet Ag/AgCl electrodes achieving excellent signal quality (SNR > 40 dB) but limited to hours of monitoring [52] Modern dry electrodes using conductive polymers (PEDOT: PSS) or nanostructured surfaces (CNT forests, gold nanopillars) achieve clinical-grade ECG despite higher impedance [55, 122] Ultrathin electrodes (< 5 μm) establish van der Waals adhesion without adhesives, achieving impedance comparable to wet electrodes while maintaining stability during activity, with long-term studies demonstrating stable recording > 7 days [52] Emerging materials, including graphene, laser-induced graphene, and MXenes, provide ultralow impedance with EMI shielding, reducing power-line interference, while CNT-textile hybrid electrodes enable garment-integrated ECG monitoring [54, 59, 73, 123].

EMG detection has similarly benefited from flexible dry electrode technologies that prioritize mechanical conformability and signal stability during muscle contraction [124]. Stretchable high-density EMG arrays fabricated on flexible substrates enable gel-free, multichannel muscle activity mapping [125–128], while MXene/CNT composite electrodes with three-dimensional network architectures achieve low impedance and biocompatibility suitable for 64-channel non-invasive EMG acquisition [129–131]. For prolonged wear, hydrogel-conductive textile hybrid electrodes, such as “smart sock” configurations, offer skin conformability and breathability for extended surface EMG monitoring [127], whereas aerosol jet-printed and liquid metal-based electrodes reduce motion artifacts and contact noise, proving advantageous for gesture recognition and human-machine interface applications [132–135]. Recent studies employing MXene fiber electrodes for neuromuscular stimulation and EMG recording have demonstrated SNR values (~ 15 dB) comparable to conventional metal electrodes, underscoring the viability of advanced nanomaterials in practical EMG systems [125, 136–138].

EOG monitoring, which captures ocular electrical potentials during eye movements and blinks, demands ultra-thin, low-impedance electrodes with minimal skin irritation for extended periorbital placement. MXene and graphene electrodes leverage high conductivity and mechanical flexibility to detect eye blink and saccade signals with low noise, while exhibiting reduced skin irritation during chronic wear [139–141]. Integration of conductive polymers (PEDOT: PSS) and textile composites into eyeglass frames or mask patches facilitates dry EOG acquisition without conductive gels [141], and AI-driven signal processing combined with multi-sensor fusion (EMG, EOG, ECG) is expanding EOG applications into gaze-based user interfaces and sleep stage monitoring [139]. Collectively, these advances in EMG and EOG electrodes mirror the trajectory of ECG technology, converging on shared material platforms-graphene, MXenes, CNTs, PEDOT: PSS-that deliver mechanical compliance, electrochemical performance, and seamless integration into wearable form factors [141].

Mechanical signals: strain and pressure

Mechanical sensors transduce pressure, strain, and deformation into electrical signals, enabling monitoring of respiratory rate, pulse waveform, and joint angles. Piezoresistive sensing achieves gauge factors (GF) of 10–10,000 through percolation network reconfiguration, with graphene foam sensors exceeding GF > 100 through crack formation mechanisms [56, 58, 60]. Capacitive sensing offers wide operating ranges (> 50% strain), high linearity, and minimal temperature sensitivity, with structured dielectrics achieving pressure sensitivities exceeding 1 kPa⁻¹. Piezoelectric materials such as PVDF and triboelectric nanogenerators enable self-powered sensors generating 1–10 V and > 100 V MPa^–1, respectively [56, 142].

Pulse waveform analysis employs pressure sensors over arteries recording pulsations (20–50 mmHg), with flexible sensors (sensitivity 0.1-1 kPa^–1) capturing pulse wave velocity, indicating arterial stiffness [56, 58, 72]. Respiratory monitoring uses thoracic strain sensors detecting < 0.5% strain for tidal breathing while tolerating > 30% for deep breaths, with dual-site monitoring discriminating thoracic versus diaphragmatic patterns.

Optical signals: photoplethysmography (PPG)

PPG technology monitors cardiovascular dynamics through optical detection of blood volume changes rather than electrode-based potential measurements, offering reduced motion sensitivity and simplified skin interfacing [143]. Recent advances focus on material innovation, flexible architectures, optical efficiency enhancement, and multi-wavelength signal processing to enable conformal, wearable implementations [144]. Organic semiconductor-based PPG sensors represent a paradigm shift toward ultra-thin, stretchable photonic systems [145]. All-organic configurations integrating organic light-emitting diode (OLED) emitters with organic photodiode (OPD) receivers achieve mechanical compliance and optical durability suitable for wrist-worn and earlobe-mounted peripheral capillary oxygen saturation (SpO₂) monitoring, with trilayer solution-processed architectures (OPD-OLED-encapsulation) demonstrating scalable fabrication and stable on-skin SpO₂ detection [146, 147]. Advanced composite materials, particularly MXene and graphene integrated with flexible substrates provide photothermal-enhanced responses [148] that mitigate signal distortion from temperature fluctuations and mechanical deformation [149]. Perovskite-polymer-hydrogel composites [150] deliver optical transparency and skin conformability with minimal signal degradation at a bending radius below 5 mm, thereby facilitating ergonomic integration into curved body surfaces [151].

Multi-wavelength and hyperspectral PPG systems extend capabilities beyond pulse rate to comprehensive hyperspectral profiling, capturing ten or more wavelengths simultaneously to extract blood pressure, oxygen saturation, and glucose-related chromophore signatures [143, 152]. AI-driven algorithms compensate for contact pressure variability and motion noise [153–155], while camera-based remote PPG enables non-contact surveillance [156]. Collectively, PPG complements electrical biosignal modalities with artifact-resistant, electrode-free cardiovascular assessment, with material strategies, including organic semiconductors, MXene composites and flexible photonic substrates, mirroring ECG, EMG, and EOG development toward soft, skin-integrated wearable diagnostics [144, 149, 157, 158].

Representative studies of various attachable sensors

Skin-attachable biosensors

Dervisevic et al. [159] developed a silicon micropillar array (MPA)-based wearable glucose sensor to address mechanical damage to the biological recognition layer caused by friction between the skin and sensor surface during wear. The sensor featured a three-electrode configuration with approximately 72,000 pillars per cm², where the working electrode was modified with Prussian Blue (PB), chitosan-gold nanoparticle composite (Ch-AuNP), and glucose oxidase (GOx) (Fig. 1a). This hierarchical micropillar structure protected the immobilized enzyme by positioning it at the bottom and sidewalls of the pillars, shielding it from direct mechanical stress, as demonstrated by comparative friction studies where planar electrodes lost 95% response after 4 h while MPA-based electrodes maintained stable performance (Fig. 1b). The sensor demonstrated a linear detection range from 50 µM to 1.4 mM with a sensitivity of 4.7 ± 0.8 µA mM^− 1 and a limit of detection of 26 ± 5 µM in artificial sweat (Fig. 1c). On-body testing successfully tracked sweat glucose concentration changes from 22 µM during fasting to 83 µM at 30 min post-meal with subsequent gradual decrease over 2 h.

Fig. 1.

Skin attachable Biosensors. Hierarchical micropillar array-based wearable glucose sensing platform (from 1a to 1c). (a) MPA-based patch toward the skin interface (upper) and modification layers of the W electrode, along with the glucose sensing mechanism (bottom). (b) Comparison of MPA with flat gold W electrode positioned on the skin surface. (c) Electrochemical current response of the MPA-based detection patch.; Reproduced with permission [159]. Copyright 2022, American Chemical Society. Architectural design of skin-mounted perspiration-gathering patch (from 1d to 1f). (d) Schematic representation of perspiration-gathering patch. (e) Schematic depiction of superhydrophobic and superhydrophilic triangular-shaped wettability-designed surface (upper), OM image and water contact angle measurements (lower) of superhydrophobic (left) and superhydrophilic surfaces (right). (f) Schematic representation of integrated perspiration-gathering patch combined with perspiration sensor (left) and current measurements from the lactate and glucose sensors; Reproduced with permission [160]. Copyright 2021, Wiley-VCH GmbH. Biosensor devices with reusability utilizing GNFET (from 1 g to 1i). (g) Anatomical structure of the human sudoriferous gland illustrating the flexible LSPR biosensor interface within the dermal cross-sectional view. (h) UV-Vis absorption spectra demonstrating spectral peak displacement after aptamer and cortisol binding with gold nanoparticles. (i) Cortisol measurement utilizing the perspiration biosensor.; Reproduced with permission [161]. Copyright 2023, MDPI

Son et al. [160] developed a biomimetic sweat-collecting patch inspired by cactus spines to address irregular and low sweat secretion rates, utilizing wedge-shaped wettability-patterned channels on a hierarchical microstructured/nanostructured PDMS surface combining superhydrophobic boundaries with superhydrophilic wedge patterns (Fig. 1d). The superhydrophilic regions were fabricated by spray-coating a poly(vinyl alcohol) (PVA)/silica nanoparticle onto O₂ plasma-treated PDMS, while superhydrophobic surfaces were formed through octadecyltrichlorosilane (ODTS) treatment (Fig. 1e). The patch structure featured radially arranged wedge-shaped channels (3.4°) directing sweat from the perimeter toward a central 5 mm diameter reservoir on a 30 mm diameter circular patch. The biomimetic design generated unidirectional Laplace pressure enabling spontaneous droplet transport at approximately 70 mm s^− 1 regardless of substrate orientation, demonstrating gravity-independent operation (Fig. 1f). During on-body testing with subjects exercising on a stationary bicycle, the lactate sensor responded within 5 min of exercise onset, while glucose monitoring successfully tracked blood glucose changes following a high-sugar beverage consumption with a 15 min lag time between blood and sweat glucose elevation.

Nan et al. [161] developed a flexible localized surface plasmon resonance (LSPR)-based biosensor for non-invasive cortisol detection in human sweat by immobilizing gold nanoparticles (AuNPs, 80 nm diameter) onto 3-aminopropyltriethoxysilane (APTES)-functionalized PDMS substrates, followed by thiolated cortisol-specific aptamer modification. The fabricated substrate exhibited uniform AuNP distribution with approximately 16.7 particles µm^− 2 and an average diameter of 86.268 ± 7.127 nm, generating consistent LSPR signals (Fig. 1g). The sensing mechanism relied on LSPR wavelength shifts upon cortisol-aptamer binding, showing a characteristic absorption peak at 545 nm for bare AuNPs, 544 nm after aptamer immobilization, and 547 nm after cortisol capture (Fig. 1h). Quantitative analysis demonstrated excellent sensitivity across 0.1–1000 nM with a cortisol concentration range with a detection limit of 0.1 nM and high linearity (Fig. 1i). On-body validation showed morning cortisol levels 2–3 times higher than afternoon levels consistent with circadian rhythm, and exercise-induced stress elevated cortisol from 1 nM at rest to 100 nM after 50 min of moderate-to-high intensity exercise.

Skin-attachable electrical, mechanical, and photo sensors

Shin et al. [162] developed a wireless earbud-like electroencephalography device (e-EEGd) to address motion artifacts, user discomfort, and wireless communication limitations of conventional systems. The e-EEGd comprised tattoo-like electrodes (900 nm), connectors (2.8 μm), and a wireless earbud (8.24 g) with electrodes positioned at the mastoid, the temple, and the forehead for frontal lobe measurements (Fig. 2a). Quantitative comparison demonstrated the superior performance of the e-EEGd system over commercial EEG devices (c-EEGd) with wet electrodes and dangling connectors (Fig. 2b). During walking, the e-EEGd exhibited significantly reduced signal fluctuations compared to the c-EEGd (Fig. 2c). While both devices showed similar average root mean square (RMS) values (11 µV) during sitting, the e-EEGd maintained 18.57 µV during walking compared to 45.49 µV for c-EEGd (Fig. 2d). Overall, the e-EEGd generated approximately 8.5 times less artifactual noise than the commercial device, enabling high-quality EEG measurements during outdoor activities.

Fig. 2.

Skin attachable electrical, mechanical, and photo sensors. (a) Image showing the e-EEGd system (left), the placement of electrodes positioned on the mastoid, temple, and forehead areas, respectively (right). (b) Photograph wireless EEG systems of e-EEGd and c-EEGd. (c) Sixty superimposed EEG traces during ambulatory movement, from the e-EEGd (upper panel) and c-EEGd (lower panel). (d) Root mean square measurements of EEG signals obtained from both systems during stationary conditions (left panel) and ambulatory conditions (right panel).; Reproduced with permission [162]. Copyright 2022, Springer Nature. (e) Manufacturing process of the biomimetic flexible electronic sensor with antimicrobial properties. (f) Detection capabilities of the flexible electronic sensors when exposed to various signals; Reproduced with permission [163]. Copyright 2023, Wiley-VCH GmbH. (g) Schematic of ring-shaped pulse oximetry sensor with vertically stacked tandem OLED and dual-region OPD. (h) OPD output voltage comparison showing 2.7 times higher signal for ring-shaped design. (i) Pictures of finger-worn sensor in black thimble (top) and PPG signals with systolic/diastolic peaks (bottom).; Reproduced with permission [145]. Copyright 2025, Springer Nature

Liu et al. [163] developed a bioinspired multifunctional sensor inspired by human skin microstructure, featuring healable, recyclable, and antibacterial polyurethane elastomer (PUPDU-Cu) with MXene-coated microdome arrays. The PUPDU-Cu incorporating triple dynamic bonds (reversible hydrogen bonds, oxime carbamate bonds, and copper (II) ion coordination bonds with nitrogen atoms of adjacent oxime groups in DMG), achieved 94.5% healing efficiency after 6 h at 60 °C, 3.11 MPa tensile strength, 386% elongation, and antibacterial activity over 90% (Fig. 2e). The sensor demonstrated versatile healthcare monitoring capabilities detecting 0.98 Pa, throat vibrations, radial artery pulse, jugular venous pulse, finger bending, gait abnormalities, and bicycle tire pressure (Fig. 2f).

Choi et al. [145] developed an all-organic pulse oximetry sensor featuring vertically stacked ring-shaped architecture for ultralow-power continuous health monitoring. The device employed a three-terminal tandem OLED configuration where red (625 nm) and green (530 nm) emitting units were vertically stacked on flexible polyethylene terephthalate substrates with a 25-nm-thick silver intermediate electrode, enabling independent alternating operation at 200 Hz with 25% duty cycle while maintaining sub-1-µW power consumption at luminance levels of 25–27 cd m^–2 (Fig. 2g). The ring-shaped OLED geometry combined with dual-concentric OPD configuration, featuring circular detection regions both inside and outside the OLED ring, demonstrated 2.7 times higher photocurrent output compared to conventional circular OLED designs (Fig. 2h). This enhancement was attributed to improved photon utilization efficiency, as most OPD active areas remained within the characteristic scattering length from the light source. The finger-worn sensor integrated into an 8-mm-radius cylindrical black thimble successfully acquired high-fidelity PPG signals with clearly distinguishable systolic and diastolic peaks for both red and green channels (Fig. 2i). The device demonstrated robust mechanical durability, maintaining stable performance after 10,000 bending cycles at 8 mm bending radius.

Multimodal skin-attachable sensing systems

While individual sensing modalities provide valuable physiological information, real-world health monitoring demands the simultaneous acquisition of multiple biosignals to capture the complexity of human physiology. Multimodal skin-attachable sensing systems integrate two or more signal modalities (chemical, electrical, mechanical, and optical) within a single wearable platform, enabling cross-modal data fusion and comprehensive physiological profiling that would be impossible with any single modality alone.

A representative example of chemical-plus-physical multimodal integration is the multiplexed sweat sensing platform reported by Gao et al. [6] which simultaneously measured glucose, lactate, sodium, potassium, and skin temperature within a single flexible patch. The inclusion of a resistive temperature sensor alongside electrochemical biosensors was not merely additive: temperature data were used to correct enzyme kinetics in real time, demonstrating that cross-modal calibration is essential for accurate metabolic monitoring during exercise. This work established a paradigm in which physical signals serve as calibration channels for chemical readouts, a principle now widely adopted in multimodal wearable design.

Mechanical and optical modalities have also been productively combined in skin-attachable formats. The all-organic pulse oximetry sensor developed by Choi et al. (Sect.  2.3.2) acquires both photoplethysmographic waveforms and blood oxygen saturation simultaneously, illustrating how multi-wavelength optical detection enables concurrent measurement of distinct physiological parameters from the same sensing site [145]. When such optical sensors are co-integrated with strain or pressure transducers, mechanical artifacts caused by motion can be identified and subtracted from the optical signal, improving measurement fidelity during ambulatory monitoring. The bioinspired multifunctional sensor by Liu et al. (Sect.  2.3.2) further demonstrates this principle by combining pressure, strain, and vibration sensing in a single healable platform capable of monitoring pulse waveform, throat vibration, and limb motion at the same time [163].

Chemical and electrical multimodal sensing represents another powerful combination. By co-locating electrochemical biosensors with electrophysiological electrodes on the same flexible substrate, it becomes possible to correlate metabolic biomarkers such as lactate and glucose with electromyographic or electrocardiographic signals during physical exertion or cardiac events. This co-registration in time and space enables system-level insights, for example, linking elevated sweat lactate to abnormal cardiac rhythm during exercise that are inaccessible when signals are measured sequentially or at different body locations. The integration of wireless power and data modules shared across all sensing channels is a critical system-level design consideration, as a common communication backend reduces device footprint, power consumption, and signal latency in multimodal platforms.

Despite these advances, multimodal skin-attachable systems face several integration challenges. Electrical crosstalk between sensing channels sharing a common substrate must be managed through careful electrode layout and ground plane design. Sweat sampling geometry must accommodate sensors with different fluidic requirements, and optical emitters must be spectrally isolated from electrochemical measurement zones to prevent photocurrent interference. Achieving all of this while maintaining skin-level compliance, low-power operation, and long-term signal stability remains an active area of engineering. Future multimodal platforms are expected to incorporate on-board signal processing capable of fusing heterogeneous sensor streams in real time, enabling adaptive health monitoring that responds dynamically to evolving physiological states.

Human-interfaced implantable sensors

Interface architecture: biocompatible materials and encapsulation strategies

Implantable sensors demand intimate and stable integration with dynamic biological environments while preserving long-term electrical and chemical reliability. The primary challenges at the bioelectronic interface include mechanical mismatch between rigid electronic materials and soft tissues, moisture and ion penetration, and chronic immune responses that can degrade signal quality [26, 164–170]. Therefore, the architectural design of implantable interfaces must incorporate biocompatible materials with mechanical moduli comparable to living tissue (10–100 kPa), ensuring conformal contact and minimal inflammatory encapsulation. Soft polymers such as parylene-C, polyimide, and silicone elastomers have been widely adopted as flexible substrates and insulating layers owing to their high dielectric strength, biostability, and low permeability [171–175]. Recent advances extend to hydrogel-based encapsulants that replicate the viscoelasticity and hydration of native tissues, establishing mechanically imperceptible and biointegrated sensor platforms [28, 176, 177].

Biocompatible interface materials also play a key role in electrical coupling and long-term hermeticity [178–180]. Conductive coatings such as PEDOT: PSS, Au–Pt nanocomposites, and MXene films lower interfacial impedance and enhance charge transfer at the electrode–tissue boundary [138, 181, 182]. In addition, flexible parylene-C threads and stretchable TiO₂–Au nanowire grids illustrate how soft metallic and polymeric hybrids minimize micromotion-induced trauma during chronic implantation [26, 183, 184]. Moreover, dynamic and self-healing encapsulation chemistries based on urethane, catechol, or siloxane bonds provide adaptive resilience under physiological strain, maintaining electrical integrity for months to years in vivo [185–187].

Signal measurement modalities

Electrophysiological signals: electroencephalography (EEG), Electrocorticography (ECoG), Local Field Potential (LFP), and electroneurogram (ENG)

EEG records voltage fluctuations generated by synchronized neuronal activity in the cerebral cortex, typically measured from the scalp using non-invasive electrodes. EEG provides macroscopic insight into brain states such as sleep, cognition, and pathological oscillations in epilepsy or neurodegenerative disease. While traditional EEG is surface-based, recent developments in minimally invasive and subdermal electrodes have enabled high-fidelity acquisition with improved spatial resolution and reduced motion artifacts [188, 189]. Implantable EEG systems using ultrathin, flexible electrode arrays often composed of parylene-C or PEDOT: PSS coatings allow chronic monitoring of cortical rhythms and closed-loop neuroprosthetic control without inducing significant tissue irritation [181, 190].

ECoG measures field potentials directly from the cortical surface, bridging the scale between non-invasive EEG and intracortical microelectrode recordings. Because ECoG electrodes are placed subdurally or epidurally, the signals exhibit higher amplitude and spatial specificity than EEG, typically in the 1–200 Hz frequency range [191, 192]. ECoG is widely employed in clinical mapping of epileptic foci and brain–computer interface (BCI) research for motor decoding [193–195]. Soft, conformal ECoG grids made of gold, platinum, or MXene interconnects embedded in elastomeric substrates have demonstrated long-term stability on gyrified cortical surfaces, maintaining low impedance and biocompatibility during chronic implantation in rodents and primates [184, 196, 197].

LFP represent the summed synaptic currents within a small neuronal population (∼100 μm scale) recorded using intracortical microelectrodes or penetrating neural probes [198–200]. LFPs capture mesoscale neural dynamics including oscillations, synchronization, and population coding essential for understanding sensory and cognitive processes [201, 202]. Implantable probes for LFP recording, such as polymer-based shanks, carbon fiber arrays, and ultra-flexible mesh electronics, enable stable in vivo recording in deep brain regions with minimal glial response [203–205]. Integration of low-noise amplifiers and multiplexing circuits directly at the recording site has further enhanced signal fidelity for chronic animal studies and neural prosthetic applications.

ENG measure the compound electrical activity of peripheral nerves, reflecting both afferent and efferent signal transmission between the central nervous system and target organs. ENG signals are vital for prosthetic limb control, vagus nerve stimulation, and closed-loop neuromodulation therapies [206, 207]. Recording requires microcuff or intrafascicular electrodes that conform to delicate nerve bundles while minimizing mechanical strain and fibrotic encapsulation [208–210]. Advances in stretchable and self-healing electrode materials, such as TiO₂–Au nanowire meshes, PEDOT: PSS coatings, and hydrogel-based interfaces, have significantly improved chronic stability and biocompatibility, enabling long-term recording and stimulation in freely moving animal models [28, 184].

Chemical and metabolic signals: glucose, lactate, and cortisol

Glucose is a key metabolic biomarker that reflects the body’s energy balance and is directly associated with diabetes and other metabolic disorders. Continuous glucose monitoring (CGM) systems detect fluctuations in interstitial or tear glucose levels to provide real-time glycemic control. Implantable glucose sensors typically employ enzyme-mediated electrochemical detection using glucose oxidase (GOx) or redox-active nanocatalysts (Au–Pt, Ni–Cu), which catalyze glucose oxidation to generate proportional current signals [211, 212]. Recent developments in soft and transparent smart contact lenses and subcutaneous hydrogel-encapsulated sensors have demonstrated stable, biocompatible glucose monitoring over weeks, showing strong correlation with blood glucose levels in animal and human trials [212–215].

Lactate serves as a critical indicator of anaerobic metabolism and tissue oxygenation, reflecting muscle fatigue, hypoxia, and ischemic conditions. Implantable lactate sensors such as mouthguard typically rely on lactate oxidase (LOx) based enzymatic electrochemical reactions, converting lactate to pyruvate with simultaneous hydrogen peroxide generation. Integration of these sensors with conductive hydrogels and polymeric membranes enables continuous, real-time lactate tracking during strenuous physical activity or organ perfusion studies [216–218]. In vivo implementations in rodents and human tissues have validated their use in metabolic monitoring, sports medicine, and wound healing applications, where soft polymer encapsulants and porous electrode architectures minimize biofouling and enhance mass transport [219, 220].

Cortisol is a steroid hormone secreted by the adrenal glands in response to stress and circadian rhythm, serving as a key biomarker for endocrine and psychological states. Conventional immunoassay-based detection has evolved into implantable electrochemical or field-effect transistor (FET) sensors functionalized with cortisol-binding aptamers or molecularly imprinted polymers [221–224]. These sensors enable continuous stress monitoring via sweat, interstitial fluid, or even tear analysis, offering high specificity and rapid temporal response. Integration with flexible electronics and wireless telemetry allows real-time correlation between physiological stress levels and neural or cardiovascular responses, supporting personalized health monitoring and psychiatric research [225, 226].

Mechanical and physical signals: intraocular pressure (IOP), temperature, and oxygen tension

Intraocular pressure is a critical physiological parameter reflecting the balance between aqueous humor production and drainage in the eye, and its dysregulation is a primary cause of glaucoma. Continuous IOP monitoring has been achieved through implantable or smart contact-lens sensors that convert corneal deformation or fluidic pressure changes into electrical or capacitive signals [227–229]. Flexible microelectromechanical systems (MEMS) and piezoresistive or capacitive transducers integrated into transparent elastomer substrates such as PDMS, parylene-C, or silicone hydrogels enable minimally invasive IOP sensing with high mechanical compliance and optical clarity [230–232].

Temperature is a fundamental biophysical signal that reflects metabolic rate, inflammation, and homeostatic regulation in living tissues. Implantable temperature sensors leverage resistive, thermoelectric, or optical mechanisms to provide high-resolution spatiotemporal mapping of local thermal dynamics, crucial in applications such as hyperthermia therapy, wound healing, and neural thermoregulation studies [233–235]. Flexible and biocompatible temperature sensors based on platinum nanofilms, PEDOT: PSS, or liquid-metal microheaters have been integrated into polymeric or hydrogel matrices for chronic monitoring without thermal drift [236–238].

Oxygen tension, or partial pressure of oxygen (pO₂), serves as a direct indicator of cellular respiration, ischemia, and wound healing progression. Implantable pO₂ sensors employ electrochemical Clark-type electrodes, phosphorescent optical quenching, or luminescent nanomaterial-based probes to quantify oxygen concentration in tissues [239, 240]. The integration of these transducers with hydrogel encapsulants and oxygen-permeable membranes enhances diffusion while preventing fouling, enabling long-term functionality in dynamic biological environments [241, 242].

Representative studies of implantable sensors

Brain and heart implantable sensors

Musk et al. [183] reported a flexible, high-density brain-machine interface featuring up to 3,072 electrodes per array, individually insertable threads, and a compact implant capable of full-bandwidth streaming (Fig. 3a). A custom Neuralink ASIC was integrated by 256 programmable amplifiers, on-chip ADCs, and control circuitry to record thousands of neural signals with low noise and minimal power consumption. Multiple ASICs are assembled on a flip-chip PCB to create modular 1,536- and 3,072-channel brain-machine interface systems supporting full-bandwidth USB-C data transfer. The entire platform is encapsulated in a titanium case coated with parylene-C, providing hermetic protection and long-term stability for chronic neural recording. Ultrathin parylene-C thread probes (50 μm wide, ~ 20 mm long) incorporating multiple recording sites and on-probe reference electrodes enable high-density neural interfacing (Fig. 3b). Broadband neural recordings from 1,280 implanted electrodes (90% insertion success) using System A demonstrated simultaneous acquisition from 1,020 channels with a spiking yield of up to 70% (Fig. 3c).

Fig. 3.

Brain and heart implantable sensors. (a) Packaged sensor device. A: Individual neural processing application-specific integrated circuit capable of processing 256 channels of data (12 chips for 3072 channels), B: Polymer threads on parylene-c substrate, C: Titanium enclosure (lid removed), D: Digital USB-C connector for power and data. (b) Linear edge polymer probes with 32 electrode contacts spaced by 50 μm. (c) The broadband signals recorded from a representative thread. (d) Microscopy image and impedance of the electrode at 0% and 30% strain. Reproduced with permission [183]. Copyright 2019, JMIR. (e) Implantation and in vivo neural recording of soft SEGs. Reproduced with permission [184]. Copyright 2018, Wiley-VCH. (f) Images of the implanted all-hydrogel bioelectronic interfaces on rat heart. (g) Printed all-hydrogel bioelectronic interfaces for heart. (h) Epicardial recordings by different channels in the all-hydrogel bioelectronic interface on day 0. Reproduced with permission [243]. Copyright 2023, Nature Publishing Group

Another brain neural recording was demonstrated by Tybrandt et al. [184] using a soft, high-density, stretchable electrode grid composed of gold-coated TiO₂ nanowires embedded in a silicone matrix, achieving stable in-vivo recording of cortical signals in freely moving rats over a three-month implantation period. The electrode impedance slightly decreased under 30% strain, consistent with the increased effective electrode area (Fig. 3d). Soft and stretchable electrode grids (SEGs) were implanted on the rat somatosensory cortex to record somatosensory evoked potentials both intraoperatively and in freely moving conditions (Fig. 3e). The flexible grids could be folded and inserted through a 30% smaller craniotomy, then self-expanded to conform to the cortical surface for stable long-term contact. The recordings showed clear spatially patterned neural responses with over 85% of electrodes functional, demonstrating high reliability and uniform impedance across the array.

Other neural recording was demonstrated by Zhou et al. [243] who developed a 3-D-printable, high-performance conducting-polymer hydrogel interface that enabled stable long-term electrophysiological recording and stimulation of cardiac tissue in rat models (Fig. 3f). By combining bi-continuous conducting polymer hydrogel with printable bioadhesive and insulating hydrogel inks, all-hydrogel bioelectronic interfaces with tissue-like softness and high-water content were 3D-printed (Fig. 3g). Using the all-hydrogel bioelectronic interfaces, they achieved long-term in vivo electrophysiological recording of rat hearts and effective stimulation of sciatic nerves and spinal cords, demonstrating versatile multi-organ compatibility enabled by multi-material 3D printing (Fig. 3h).

Smart contact lens-based implantable sensors

Park et al. [213] reported soft, smart contact lenses integrating wireless circuits, glucose sensors, and micro-LED displays that enable real-time, noninvasive monitoring of glucose levels in tear fluid through wireless power transfer. (Fig. 4a). The system employs a stress-tunable hybrid substrate with reinforced islands and elastic joints to maintain mechanical stability while incorporating stretchable antennas, rectifiers, and functional components within the soft lens structure. In vivo tests on a live rabbit demonstrated that the soft, smart contact lens could wirelessly detect elevated glucose levels in tears, indicated by the LED turning off above 0.9 mM glucose, while maintaining stable operation during blinking without behavioral abnormalities(Fig. 4b). Thermal and safety analyses confirmed that the device temperature remained around 37 °C and the maximum specific absorption rate (1.399 W/kg) was well below regulatory limits, ensuring safe wireless operation.

Fig. 4.

Smart Contact lens-based implantable sensors. (a) Stretchable and transparent smart contact lens system for detecting glucose. (b) In vivo test on a live rabbit. Reproduced with permission [213]. Copyright 2018, American Association for the Advancement of Science. (c and d) Theranostic smart contact lens with a fully integrated AuHNW-based IOP senor, a drug DDS, and wireless circuits for wireless glaucoma treatments. (e) The conventional IOP monitoring and the IOP control by IOP monitoring and on-demand drug delivery for the treatment of glaucoma. Reproduced with permission [214]. Copyright 2022, Nature Publishing Group. (f) Smart contact lens for diabetes monitoring. (g) Glucose monitoring in blood and tears in diabetic rabbits. (i) stable change, (ii) increase, and (iii) decrease of blood glucose levels. (h) Clark error grid analysis for clinical validation. (i) Glucose monitoring with a smartphone application (i) Measurement system and (ii) continuous glucose monitoring with glucomter, commercial CGM and proposed smart contact lens. Reproduced with permission [212]. Copyright 2022, Wiley-VCH

Wireless theranostic smart contact lens for monitoring and control of intraocular pressure in glaucoma by Kim et al. [214] demonstrated a fully integrated smart contact lens that wirelessly monitors intraocular pressure and delivers on-demand medication for glaucoma treatment in real time in rabbit models (Fig. 4c). A highly integrated theranostic smart contact lens was developed for simultaneous monitoring and control of intraocular pressure in glaucoma. The system incorporates a gold hollow nanowire–based pressure sensor, a flexible drug delivery module, wireless circuitry, and an ultra–low-power ASIC on a parylene-C substrate (Fig. 4d). It achieves high sensitivity and stability for long-term IOP monitoring and enables electrochemically triggered, on-demand drug release for effective glaucoma treatment in animal models (Fig. 4e).

Kim et al. [212] reported the use of bimetallic nanocatalysts immobilized within nanoporous hydrogels in a smart contact lens to achieve long-term, robust continuous glucose monitoring in diabetic rabbits, demonstrating high sensitivity, minimal hysteresis, and real-time tear-to-blood glucose correlation (Fig. 4f). The smart contact lenses exhibited a strong tear-to-blood glucose correlation (ρ > 0.70, up to 0.90) with a ~ 10 min lag, attributed to the enhanced diffusion and sensitivity provided by hyaluronate-modified gold–platinum bimetallic nanocatalysts in nanoporous hydrogels (Fig. 4g). The smart contact lens showed high clinical accuracy for continuous glucose monitoring with a strong correlation (ρ = 0.82) and 94.9% of data within the clinically acceptable Clarke error grid zones, with minor inaccuracies likely due to reduced tear circulation under anesthesia (Fig. 4h). The clinical feasibility of the smart contact lens was validated by converting tear glucose measurements to blood glucose levels using the established correlation equation and comparing the results with those from a glucometer and a commercial CGM (Fig. 4i).

Mouth implantable sensors

Kim et al. [244] developed an instrumented wireless mouthguard biosensor for continuous non-invasive monitoring of salivary uric acid levels. The device integrated an enzyme-modified screen-printed electrode system with Bluetooth low energy communication, addressing the limitations of invasive blood sampling and bulky NFC systems in wearable chemical sensors. The mouthguard platform comprised a Prussian Blue carbon working electrode with immobilized uricase enzyme, an Ag/AgCl reference electrode, and a miniaturized wireless electronics (1.8 cm × 1.9 cm) (Fig. 5a). The working electrode incorporated multiple functional layers (Fig. 5b), including a Prussian Blue transducer for selective hydrogen peroxide detection, uricase cross-linked with bovine serum albumin (BSA) and glutaraldehyde, and an electropolymerized o-phenylenediamine membrane preventing biofouling and rejecting interfering electroactive species during prolonged measurements in undiluted saliva. Chronoamperometric measurements at 0.3 V in artificial saliva demonstrated excellent linearity (slope: 2.32 µA mM^–1, R² = 0.998) across 0–1 mM uric acid with minimal interference from ascorbic acid, glucose, lactate, and acetaminophen, covering both normal (100–250 µM) and hyperuricemia ranges (Fig. 5c and d).

Fig. 5.

Mouth implantable sensors. (a) photos showing the mouthguard biosensor with a wireless electrochemical measurement circuit module. (b) Biochemical coating layer on the chemically functionalized printed carbon electrode modified with Prussian Blue and containing uricase enzyme for serum uric acid detection (upper), and photos of the wireless electrochemical circuit module. (c) Current-time profiles measured for uric acid. (d) Interference assessment of sensor response to (i) UA with (ii) co-present ascorbic acid, (iii) glucose, (iv) lactic acid, and (v) acetaminophen.; Reproduced with permission [244]. Copyright 2015, Elsevier. (e) Diagram illustrating the glucose detection device fabricated on polyethylene terephthalate glycol mouthguard substrate. (f) Design schematic of the personalized mouthguard biosensing platform. (g) The oral biosensor device affixed to the model mandible within an open-circuit flow system.; Reproduced with permission [245]. Copyright 2016, Elsevier

Arakawa et al. [245] developed a “Cavitas sensor” mouthguard biosensor with integrated telemetry for continuous non-invasive salivary glucose monitoring. The device overcame the discomfort, infection risk, and measurement discontinuity associated with invasive self-monitoring of blood glucose (SMBG). The microelectromechanical systems (MEMS)-fabricated device featured a 0.2 mm² platinum (Pt) working electrode and a 4.0 mm² silver/silver chloride (Ag/AgCl) reference/counter electrode on a 0.5 mm polyethylene terephthalate glycol (PETG) substrate, with glucose oxidase (GOD) immobilized via poly(MPC-co-EHMA) (PMEH) polymer coating that catalyzed glucose oxidation to produce electrochemically detectable hydrogen peroxide (H₂O₂) (Fig. 5e). The complete mouthguard glucose sensor was fabricated through a multi-step process involving dental impression, cast formation, vacuum forming of the PETG mouthguard support over the dental cast, electrode fabrication, and wireless transmitter integration (Fig. 5f). The optimized biosensor exhibited exponential glucose response (ΔOutput Current (µA) = 0.077[Glucose (µmol L^–1)]^0.90 across 5-1000 µmol L^–1, encompassing physiological ranges, with high selectivity (negligible interference from fructose, mannitol, sorbitol, and xylitol) and demonstrated stable wireless monitoring exceeding 5 h in a 38 °C phantom jaw system with continuous artificial saliva flow, achieving response times within 60 s for increasing concentrations and 180 s for decreasing concentrations (Fig. 5g).

To facilitate a clearer comparison of different human-interfaced sensing platforms, Table 1 summarizes representative skin-attachable and implantable sensors discussed in this review, highlighting their interface sites, operational stability, key advantages, and typical limitations.

Table 1.

Comparative performance and operating stability in representative implantable sensors

Platform	Representative study	Implant site / modality	Operational stability	Key advantage	Typical limitation
Skin-attachable	Dervisevic et al. [159]	Epidermis /sweat glucose	few days	Non-invasive monitoring	Sweat variability
	Shin et al. [162]	Temple /e-EEGd	> 7 days	Motion-artifact reduction	Adhesion / drift
	Choi et al. [145]	Finger /PPG	10,000 cycles	Low-power monitoring	Motion interference
Implantable	Neuralink [183]	Brain /neural recording	chronic	High channel density	Feedthrough/packaging, insertion trauma
	Tybrandt et al. [184]	Cortex /electrophysiology	~ 3 months	Soft stretchable interface	Micromotion
	Zhou et al. [243]	Heart, nerve, spinal /electrophysiology	> 2 months	Tissue-like hydrogel interface	Hydrogel dehydration / encapsulation breach

Multimodal implantable sensing systems

Implantable sensor systems increasingly pursue multimodal architectures to overcome the fundamental limitations of single-modality monitoring within dynamic in vivo environments. Because biological processes are inherently coupled—neural activity modulates cardiovascular function, metabolic state influences electrophysiology, and mechanical forces alter chemical microenvironments—sensors capable of simultaneously capturing multiple signal dimensions provide a more complete and mechanistically interpretable picture of tissue physiology than single-channel devices.

The all-hydrogel bioelectronic interface reported by Zhou et al. (Sect.  3.3.1) exemplifies multimodal implantable design: by 3D-printing conducting polymer hydrogel alongside bioadhesive and insulating inks, the platform simultaneously supported electrophysiological recording from cardiac tissue and electrical stimulation of sciatic nerves and spinal cords in rat models [243]. The ability to both record and stimulate within the same implanted construct represents a closed-loop multimodal system in which sensory feedback directly informs actuation, a design principle central to next-generation neuroprosthetics and electroceuticals. Similarly, the smart contact lens platforms described in Sect.  3.3.2 integrate chemical sensing, mechanical sensing of intraocular pressure, and optical output on a single transparent elastomeric substrate, demonstrating that multimodal integration is achievable even in highly space-constrained implantable formats [203, 204].

Cross-modal data fusion in implantable systems unlocks diagnostic capabilities that are unavailable to unimodal devices. For instance, simultaneous recording of local field potentials and interstitial glucose in deep brain structures allows researchers to correlate oscillatory neural dynamics with metabolic fluctuations, providing mechanistic insight into conditions such as epilepsy and Parkinson’s disease where both neurological and metabolic dysregulation occur. Likewise, combining electrophysiological ECoG recordings with local temperature or oxygen tension measurements enables real-time detection of cortical spreading depression and ischemic penumbra, which are accompanied by characteristic thermal and hypoxic signatures that are invisible to electrical sensing alone. These synergistic cross-modal measurements motivate the design of implantable probes with heterogeneous sensing modalities co-localized at the micron scale.

System-level integration poses unique challenges for multimodal implantable devices beyond those encountered in surface-worn sensors. Hermeticity requirements are more stringent, as any moisture ingress can corrupt electrochemical measurements or short electrical channels. The mechanical compliance of multi-sensor arrays must be maintained across all constituent materials, conducting polymer electrodes, enzyme-functionalized sensing zones, and wireless coils—each with distinct moduli, and chronic micromotion at the implant-tissue interface must not degrade signal fidelity across any channel. Power budgets are extremely constrained in implantable formats, so shared wireless power harvesting and multiplexed readout electronics are essential to avoid prohibitive device bulk. The mouthguard biosensors reviewed in Sect.  3.3.3 illustrate how multiplexed wireless electronics can simultaneously service multiple electrochemical channels from a single compact transceiver, a scalable architectural principle directly applicable to fully implanted multimodal probes [244, 245].

Looking forward, the convergence of soft bioelectronics, in-sensor neuromorphic computing, and biocompatible encapsulation chemistry will enable implantable multimodal systems capable of chronic, drift-free operation over years rather than weeks. On-implant signal processing that fuses electrophysiological, chemical, and mechanical data streams before wireless transmission will reduce bandwidth requirements, while adaptive closed-loop architectures will allow implanted multimodal devices to autonomously modulate therapy in response to real-time physiological feedback. Realizing this vision will require not only advances in individual sensing modalities but a deliberate systems engineering approach that treats cross-modal integration and signal fusion as first-order design criteria from the earliest stages of device development.

Conclusions

The advancement of human-interfaced sensing systems represents a fundamental paradigm shift in biomedical engineering, transitioning from rigid, episodic medical instrumentation to soft, continuous, and seamless physiological monitoring. As explored in this review, the convergence of materials science, mechanical engineering, and bioelectronics has bridged the gap between abiotic electronic devices and biotic soft tissues. The evolution of interface architectures, ranging from elastomeric skin patches to hydrogel-encapsulated neural probes, demonstrates that achieving mechanical compliance and maintaining stable electrical contact are prerequisites for high-fidelity signal acquisition. The current state-of-the-art technologies tackle a diverse array of functional materials, including carbon nanotubes, graphene, MXenes, and conductive polymers like PEDOT: PSS, to create sensors that are not only highly sensitive but also mechanically resilient against the dynamic deformations of the human body. Whether through non-invasive skin-attachable platforms monitoring sweat metabolites and electrophysiological signals (ECG, EMG), or through invasive implantable systems recording neural activity (ECoG, LFP) and ocular dynamics, these devices have demonstrated the capability to provide comprehensive, real-time insights into human health. Notable progress in wireless power transfer and data communication has further unencumbered these systems, moving them closer to practical, unobtrusive daily use. Despite these significant strides, several critical challenges remain barriers to widespread clinical adoption. Long-term stability continues to be the critical bottleneck, particularly for chemical sensors where enzymatic degradation and biofouling limit operational lifetimes to days or weeks. The long-term reliability of human-interfaced sensors is primarily hindered by several interfacial challenges that degrade signal fidelity through distinct mechanisms. Mechanical mismatch between rigid electronic components and curvilinear biological tissues leads to localized stress concentration [246]. Under dynamic physiological conditions, this mismatch induces interfacial delamination and micro-slippage, which manifests as significant motion artifacts and a loss of conformal contact, ultimately compromising signal-to-noise ratios [247, 248]. Furthermore, impedance drift often arises from the dynamic fluctuations of the ionic environment and the hydration state of the interface [249]. This shift in the baseline impedance complicates the interpretation of chronic data, as it necessitates continuous recalibration to maintain sensing accuracy [250]. For implantable devices, biofouling presents a critical barrier; the non-specific adsorption of proteins triggers a foreign body response, leading to the formation of a dense fibrous capsule [251, 252]. This capsule acts as a physical diffusion barrier for target analytes, significantly increasing response times and reducing sensitivity. Simultaneously, the saline-rich, oxidative environment of physiological fluids can catalyze the electrochemical corrosion of metallic electrodes [252]. This process leads to the structural degradation of the active sensing area and the leaching of potentially toxic ions, which not only shortens the device’s operational lifespan but also raises biocompatibility concerns. For implantable devices, mitigating the foreign body response and ensuring chronic biocompatibility without fibrous encapsulation requires further innovation in surface chemistry and stealth materials. Additionally, achieving robust power management and signal reliability in uncontrolled, motion-rich environments remains a complex engineering hurdle that necessitates advancements in energy harvesting and artifact-cancellation strategies. Beyond these technical challenges, clinical translation of human-interfaced sensors requires addressing practical issues including calibration, reproducibility, safety, regulatory approval, and scalable manufacturing. Reliable calibration is needed to compensate for long-term drift and inter-individual variability, while standardized validation and reproducible fabrication will be essential for real-world healthcare deployment. Looking forward, the future of human-interfaced sensors lies in the realization of fully autonomous, closed-loop theranostic systems. Advances in AI-assisted signal processing enable more reliable interpretation of complex physiological data, while closed-loop theranostic systems may integrate sensing and therapeutic actuation within a single platform. Furthermore, the development of sustainable, biodegradable electronics addresses the growing environmental concerns associated with disposable healthcare devices. Ultimately, as these technologies mature, they hold the potential to redefine the healthcare landscape, moving away from reactive hospital-based treatment toward proactive, personalized, and ubiquitous health management.

Author contributions

W.K. and Y.K. wrote the main manuscript text and prepared the figures. H.Y. and E.K.L. supervised the project and revised the manuscript. All authors reviewed the manuscript.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent to publish

Not applicable.

Competing interests

The authors declare no competing interests.