Motion artefact management for soft bioelectronics

Abstract

Soft bioelectronic devices can be interfaced with anatomically curved organs, such as the heart, brain and skin, to provide continuous analysis of physiological information. However, body movements and physiological activities may induce motion artefacts, which can adversely affect signal accuracy and stability. Importantly, motion artefact management is key to promoting the clinical translation of soft bioelectronics to ensure that soft bioelectronic devices can selectively detect target biological signals with high accuracy. In this Review, we discuss how body activities can affect the soft bioelectronic–tissue interface and result in motion artefact signals, including interface impedance instability motion artefacts, biopotential motion artefacts and mechanical motion artefacts. We then investigate different motion artefact management strategies, including materials engineering, device and circuit design, and algorithmic intervention, to reduce the contribution of motion artefacts to signal acquisition, processing and interpretation.

Introduction

Soft bioelectronics can continuously produce clinical-grade data for the personalized and remote monitoring of health and disease^1–3. In contrast to rigid silicon microelectronics, which fail to establish long-lived and robust contacts with soft biological surfaces, soft bioelectronics are designed for seamless and conformal integration with the human body⁴, particularly for the management of diseases currently limited by delayed or inefficient medical services such as infectious diseases and care for the ageing population^5–8. Soft bioelectronics-based body area networks with monitoring, therapeutic and energy functions^9,10 can consistently collect and analyse physiological information, providing continuous and real-time health-care data to identify dynamic physiological changes and potential health issues at an early stage^8,11,12. Therefore, soft bioelectronics may enable personalized, predictive and participatory health care for disease prevention and health promotion.

A soft bioelectronic system usually encompasses sensing units and electrode components for the acquisition of diverse physiological information, spanning electrophysiological, biochemical and biomechanical signals as well as circuits for signal processing and transmission (Fig. 1a). By establishing a seamless and stable bioelectronic–tissue interface, soft bioelectronics can provide conformability and monitoring accuracy^13,14. However, the stability of the bioelectronic–tissue interface is susceptible to human physiological activities, which can cause motion artefacts that may disruptively modify target signals, thereby challenging the accuracy and reliability of soft bioelectronic measurements. Depending on the causes of formation, motion artefacts can be classified into three categories: interface impedance instability motion artefacts, biopotential motion artefacts and mechanical motion artefacts.

Fig. 1 |. Soft bioelectronic systems and bioelectronic–tissue interfaces.

a, A typical soft bioelectronic system includes electrodes and sensors for signal acquisition as well as circuits for signal processing and transmission. b, Soft bioelectronics interact with the human body to extract physiological information, including continuous electrophysiological, biochemical and biomechanical signal monitoring. ECG, electrocardiogram.

Interface impedance instability motion artefacts originate from impedance instability at the bioelectronic–tissue interface, attributed to soft bioelectronic detachment and displacement-induced changes in the electrode contact state¹⁵. Interface impedance instability motion artefacts can induce signal baseline drift and severe signal distortion¹⁶.

Biopotential motion artefacts stem from biopotential signals, produced by the electrochemical activities of excitable cells in nervous and muscular tissues. These biopotential signals may be captured and can interfere with the acquisition and interpretation of the target signals, which may limit physiological monitoring, in particular, of weak or closely spaced signals^17–19.

Mechanical motion artefacts can be caused by biomechanical activities such as body movements and physiological functions. These activities lead to mechanical disturbances that propagate to soft bioelectronics, thereby introducing irrelevant biomechanical signal contributions from the device itself to the target signals^20,21.

Such motion artefacts place challenges to the clinical translation of soft bioelectronics and thus need to be addressed. In this Review, we introduce the configuration of a typical soft bioelectronic system and bioelectronic–tissue interfaces for electrophysiological, biochemical and biomechanical signal measurements. We then analyse the three types of motion artefact and examine methods for their elimination and management. Specifically, to mitigate motion artefacts from interface impedance instability, we delve into the construction of soft, stretchable and adhesive soft bioelectronics, proffering multifaceted solutions, including material–tissue mechanical matching, device structure engineering and bioelectronic–tissue adhesion. For biopotential and mechanical motion artefact management, we underscore the pivotal role of algorithmic intervention from the software perspective, alongside hardware-based strategies, collectively encompassing both sensor and circuit designs. By exploring these solutions to motion artefact management strategies, the clinical translation of soft bioelectronics for personalized health care may be greatly accelerated.

Soft bioelectronic–tissue interfaces

Soft bioelectronics, owing to their mechanical conformability, can seamlessly integrate with curved tissues. The bioelectronic–tissue interface between soft bioelectronics and living tissues enables stable mechanical coupling and the efficient capture of physiological signals^13,22–29. Depending on the nature and source of physiological signals, this interface facilitates the monitoring of electrophysiological, biochemical and biomechanical signals (Fig. 1b). For electrophysiological monitoring, soft electrodes (with or without an adhesive layer) establish an electrical interface with living tissues to accurately and imperceptibly capture weak electrophysiological activities through various biopotential recordings, including electrical activities in the brain (electroencephalogram (EEG)), heart (electrocardiogram (ECG)), eye (electrooculogram (EOG)) and muscle (electromyogram (EMG)). Soft microelectrodes can further be implanted to establish brain–machine interfaces^30,31.

Biochemical monitoring is typically employed for the analysis of biomarkers in easily accessible body fluids such as sweat, saliva and blood. For example, biochemical sensors can initiate chemical reactions with body fluids on the skin, such as electrolytes (for example, Na⁺, K⁺ and Ca²⁺) and organic molecules (for example, glucose, lactate, hormones), generating electrical signals for real-time biochemical parameter measurement³.

For biomechanical monitoring, soft bioelectronics employ mechanical sensors to capture biophysical parameters, such as stress, pressure and deformation, enabling the detection of biomechanical processes within the body, including wrist pulse, respiration movements and body-bending signals, through biomechanical-to-electrical conversion³².

However, the coupling state of bioelectronic–tissue interfaces can be influenced by various factors. Over time, changes may occur in the structure and properties of soft bioelectronics, leading to alterations in their sensing performance. For example, structural instability in bioelectronics may deteriorate sensing performance, including dislodgement of the conductive layer on the electrode^33,34 and corrosion within the ionic biofluid environment^35,36. Similarly, ion-selective, gel-based sensors may exhibit property fluctuations in response to varying environmental ion concentrations. The impact of such device changes at the bioelectronic–tissue interface may be reduced by material design and sensor replacement. Nonetheless, motion artefacts, the most prevalent factor affecting the bioelectronic–tissue interface, can result in distortions or unreliability in monitoring data and remain difficult to address.

Mechanisms of motion artefacts

Interface impedance instability motion artefacts

The electrical characteristics of the bioelectronic–tissue interface play a crucial role in electrophysiological monitoring^35,37. To capture high-quality electrophysiological signals from living tissues using flexible electrodes, a stable interface has to be maintained. A simplified equivalent circuit model consisting of resistance and capacitance can be applied to represent the interface impedance characteristics at different frequencies^38,39 (Fig. 2a). Depending on the contact state of the interface, the electrode can establish direct contact (as a dry electrode) or engage through a conductive gel layer (as a wet gel electrode) with the tissue. The interface impedance of these contact modes can be quantitatively expressed as:

$$Z=R_{\text{t}}+\frac{R_{\text{e}}}{1+j\omega C_{\text{e}}{R}_{\text{e}}}$$ (1)

$$Z\prime =R_{\text{t}}+\frac{R_{\text{e}}}{1+j\omega C_{\text{e}}{R}_{\text{e}}}+\frac{R_{\text{d}}}{1+j\omega C_{\text{d}}{R}_{\text{d}}}$$ (2)

$$c=\frac{{\epsilon }_0{\epsilon }_{r}A}{d}$$ (3)

where R_t, R_e and R_d are the resistance of tissue, electrode and gel, respectively; C_e and C_d are the capacitance of the electrode–epidermis-or-gel interface and gel–epidermis interface, respectively; j is the imaginary component; ω is the angular frequency; and Z and Z′ are the impedance of the bioelectronic–tissue interface in dry electrode and wet gel electrode contact modes, respectively. For capacitance, ε₀ and ε_r are the relative permittivity of free space and material, respectively; A is the contact area of the plate; and d is the distance between two plates. Resistance primarily influences the transmission of low-frequency signals by attenuating them as they traverse the circuit, whereas capacitance predominantly facilitates the transmission of high-frequency signals by storing and releasing electrical energy. Compared to the dry electrode, the gel layer of the wet gel electrode acts as an electrolyte between the electrode and epidermis, providing adhesion and reducing the interface impedance from 200 Ω of the dry electrode to 10 Ω of the wet gel electrode. This configuration offers stable tissue integration and accurate bioelectrical signal capture^40–42.

Fig. 2 |. Motion artefacts in soft bioelectronic systems.

a, Interface impedance instability at electrode–tissue interfaces. b, Biopotential signals are generated by excitable cells through ion transmembrane transport. Overlapping biopotential signals can disturb the acquisition of target electrophysiological signals. c, Signal artefact disruptions to target signals can result from biomechanical activities, including body movements and physiological functions. R_e, resistance of electrode; R_t, resistance of tissue; R_d, resistance of gel; C_e, capacitance of the electrode–epidermis interface; C_d, capacitance of the electrode–gel interface; U, transmembrane potential.

Soft bioelectronics continuously slide and are displaced owing to body activities such as skin deformation or sweat accumulation at the interface. The resulting contact-state changes of the interface lead to contact capacitance variations within the circuit model, causing fluctuations of interface impedance (at the order of 100–1,000 Ω for dry electrodes). By contrast, the adhesive gel layer in wet gel electrodes reduces electrode movement and fluctuations in interface impedance (in the order of 10–100 Ω)⁴³. However, during body activities, mechanical mismatch between the conductive gel and electrode can lead to electrode delamination and fragmentation as well as to polymer fatigue from the conductive gel, impacting the stability of electrode–gel interface impedance^16,23. The consequential fluctuations in interface impedance instability considerably impact current transmission at the interface, causing distortions to the recorded signals and thereby hindering the acquisition of accurate and reliable electrophysiological and impedance-related signals such as impedance-based skin hydration, electrical impedance tomography and bioimpedance-based blood pressure measurement.

Biopotential motion artefacts

Biopotentials are electrical signals generated by excitable cells during physiological processes⁴⁴ inherent in daily activities. At the cellular level, these signals arise from variations in ionic concentrations across cell membranes (Fig. 2b). Biopotential signals include EOG from the movement of eyeballs and eyelids, EMG from muscular movements, and ECG from heart muscle during heartbeats⁴⁵. These biopotential signals can be captured simultaneously through bioelectronic–tissue interfaces; however, overlapping signals may lead to signal distortion during electrophysiological monitoring²⁰.

EEG is a non-linear and non-stationary signal pattern, characterized by complex components and reflecting important information about brain electrical activity. EEG signals exhibit microvolt-level amplitudes and weak intensities, rendering them susceptible to interference from other physiological signals originating within the human body^46,47. During EEG measurements, eye blinks or movements can produce spike-like signal waveforms, with peak amplitudes reaching up to 800 μV (ref. 48). These waveforms, referred to as EOG signals, produce biopotential motion artefacts for EEG signals owing to their considerably higher magnitude¹⁸. Similarly, higher-amplitude ECG signals can substantially influence EMG signals, in particular, when recording from upper trunk muscles^49–51. This influence results in an augmentation of power content and distortion of the frequency characteristics of EMG signals. Therefore, such biopotential motion artefacts stemming from body activities should be addressed. By contrast, the detection of biochemical and biomechanical signals follows mechanisms that do not require the direct interface of electrodes with biological tissues, rendering them unaffected by biopotential motion artefacts and offering a pathway to uncontaminated and robust data acquisition.

Mechanical motion artefacts

Mechanical motion artefacts are induced by biomechanical activities, causing mechanical disturbances that propagate through organs and tissues, ultimately reaching the soft bioelectronics in contact with the body. This cascade of motion artefacts introduces irrelevant biomechanical signals to the target signals, which often occurs in wearable sensing applications such as biomechanical, biochemical, electrophysiological and temperature sensors. These mechanical disturbances lead to interference and signal distortion, impacting the accuracy and reliability of physiological monitoring.

Mechanical motion artefacts can originate from various sources, including body movements, such as walking, running or jumping, as well as physiological functions such as heartbeat and respiration^52–54 (Fig. 2c). For example, interference from gait motions can introduce signal artefacts with high amplitudes and irregular frequencies, exceeding those of most physiological signals, thereby disturbing physiological monitoring. Consequently, the low signal quality impacts data extraction and analysis, preventing the detection of valuable physiological information, particularly from self-powered soft bioelectronics based on biomechanical-to-electrical conversion⁵⁵. By contrast, interference sources from physiological functions typically exhibit small amplitudes and stable frequencies. The generated signal artefacts often behave at similar levels of frequencies or amplitudes to the target physiological signals, exerting a consistent influence on signal acquisition. Physiological signals and mechanical motion artefact signals usually coexist across different frequency bands, allowing the possibility of separating the contributions of these overlapping functions based on their characteristic frequencies^56,57. Thus, signals with motion artefacts can still provide valuable physiological information through various post-signal processing algorithms.

Interface impedance instability motion artefact management

Interface impedance instability motion artefacts originate from changes in contact status caused by the detachment and displacement of soft bioelectronics, resulting in subsequent fluctuations in interface impedance. The management of interface impedance instability motion artefacts revolves around strategies to ensure a robust connection of soft bioelectronics with irregularly curvilinear and dynamically changing living tissue surfaces. Managing interface impedance instability motion is particularly important for implantable bioelectronics used in electrocorticography, local field potentials and electroneurogram to ensure accurate monitoring of brain and peripheral nerve activity. Maintaining stable contact at the bioelectronic–tissue interface during body movements is key to suppressing interface impedance variations. To enhance conformability, bioelectronics should be stretchable, soft and adhesive to dissipate strain energy and rapidly adapt to dynamic tissue surfaces. This can be achieved by ensuring material–tissue mechanical matching and device structure engineering to enhance structural deformation. Additionally, bioelectronic–tissue adhesion should be ensured to facilitate the seamless integration of bioelectronics with living tissues.

Material–tissue mechanical matching

The mechanical properties of electronic materials, particularly their intrinsic stretchability and softness (modulus matching), are crucial for achieving contact stability with living tissues. Matching the mechanical properties of the bioelectronics with those of the tissue permits conformability under dynamic changes to manage interface impedance instability of the bioelectronic–tissue interface.

Intrinsic stretchability.

The stretchability variation of different living tissues during various body activities reflects their physiological characteristics and functional requirements (Fig. 3a) and needs to be accounted for in the design of soft bioelectronics. Adapting to the surface topography of the tissue is paramount in eliminating the strain mismatch from impedance instability motion artefacts induced by bioelectronic–tissue interfaces⁵⁸. For example, the nervous system, including the brain and spinal cord^59,60, is characterized by low stretchability. The brain and spinal cord can experience tensile strains of 5% and 10–20%, respectively, during normal postural movements. Conversely, muscles, including skeletal and cardiac muscles, exhibit a moderate strain range, typically between 20% and 40%, which corresponds to their muscle activities and cardiac physiological functions⁶¹. Furthermore, joints, such as the knee, fingers and elbows, which perform a range of movements, including bending, stretching and twisting, exhibit the greatest strain range among tissues owing to their need for flexibility and accommodation of large-scale motions⁶².

Fig. 3 |. Interface impedance instability motion artefact management.

a, Stretchability and softness of human tissues. b, Stretchability and softness of soft bioelectronic materials. c, Interface impedance instability motion artefacts can be managed by reducing device thickness. To provide conformal contact with the irregular shape of living tissues during human body activities, the thickness of soft bioelectronics can be reduced to an ultrathin structure to increase the interfacial contact energy (U_interface). d, Device strain can be adjusted by modifying the structure of the device, for example, by including an island bridge design, in which a connecting bridge is used for stretching while not affecting functional rigid islands. To increase the mechanical stability of soft bioelectronics, an elastiff layer can be introduced to divert strain to the deformable, low-modulus interspace. e, Deformation-tolerant designs provide reversible stretchability of soft bioelectronics; a wrinkle structure can be introduced into the device through pre-strain during fabrication; serpentine traces achieve tuneable deformation and tissue-level conformability, depending on the geometry-yielding feature; fractal structures can be applied across the entire device to increase stretchability; kirigami patterns, created from 2D starting materials, achieve device deformation involving structural stretching and bending; knitted textiles, with interlocking yarn loops, can achieve high stretchability and provide breathability to epidermal bioelectronics. f, Bioelectronic–tissue adhesion can be improved to manage interface impedance instability motion artefacts. Physical interactions can be formed at the bioelectronic–tissue interface to ensure that soft bioelectronics conformally attach to dry skin surfaces. Physical interactions, including van der Waals (vdW) forces, hydrogen bonding and electrostatic interactions, result from an electron density transient shift, atoms with differences in electronegativity (H and X, X including O, F, N), and oppositely charged polar group attraction, respectively. g, To provide long-term integration and non-invasive debonding of soft bioelectronics in wet conditions, a robust bioelectronic–tissue interface can be constructed through covalent chemical bonding. h, Capillary forces can achieve conformal wrapping of soft bioelectronics with tissue surfaces in a liquid environment. i, Surface microstructures can be implemented to improve adhesion on wet and rough skin surfaces. Strong physical forces can be introduced through surface microstructures, for example, a hierarchical structure with a large contact area and a micro-sucker with negative pressure. δ⁺, positive charge; δ^–, negative charge; θ, contact angle; EGaIn, eutectic gallium-indium; GO, graphene oxide; NHS, N-hydroxysuccinimide; NWs, nanowires; P3HT-NFs, poly(3-hexylthiophene-2,5-diyl) nanofibrils; PAA, poly(acrylic acid); PAAm, polyacrylamide; PDMS, polydimethylsiloxane; PEDOT:PSS, poly(3,4-ethy lenedioxythiophene):poly(styrenesulfonate); PET, polyethylene terephthalate; PI, polyimide; PU, polyurethane; PVA, polyvinyl alcohol; PVDF, polyvinylidene difluoride; SBS, poly(styrene-butadiene-styrene); SEBS, polystyrene-ethylenebutylene-styrene; r, radius; a, contact circle radius.

To achieve tissue-level stretchability in soft bioelectronics, polymer materials are commonly used owing to their long-chain molecular structure, intermolecular slippages between chains⁶³ and the reversible dissipation of strain energy through dynamic bonds^64–68. Thus, polymer materials can be designed with a high strain range to facilitate the seamless adaptation of bioelectronics to the dynamic surfaces of tissues. For example, elastomers, such as polydimethylsiloxane (PDMS)⁶⁹, silicone^70,71, polystyrene-ethylene-butylene-styrene¹⁶, polyurethane⁷² and PDMS–isophorone bisurea_0.6–methylenebis(phenyl urea)_0.4 (refs. 67,73) can serve as substrates in soft bioelectronics (Table 1 and Fig. 3b). Polymer-based hydrogels, such as polyacrylamide¹⁶ and poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate)⁷⁴ hydrogels, can be modified for conductivity and adhesion to function as substrates, adhesive layers, electrodes or interconnections in soft bioelectronics. Additionally, inorganic materials with certain electrical conductivity, such as graphene⁷⁵, silver nanowires³⁵ and carbon nanotubes¹¹, can be structured into networks on elastomeric surfaces to serve as electrodes, interconnections and functional units. In addition, the physical interactions between the nanosheets of molybdenum disulfide, not relying on bonding, can provide high mechanical stretchability⁶³. Furthermore, conductive fillers (for example, liquid metal, silver nanowires and silver flakes) can be embedded into polymer matrices to form elastic composites that exhibit both electrical conductivity and mechanical stretchability^76–82.

Table 1 |.

The mechanical properties of materials in soft bioelectronics

Material class	Examples	Mechanical properties		Functional components	Fabrication scalability	Biocompatibility	Refs.
		Stretchability (%)	Softness (kPa)
Plastics	PI, PET, PVA, PVDF	<2	1.9–3.4×106	Substrate	High	Moderate	91,92,101,219,220
Elastomers	PDMS, SEBS, PU, silicone, PDMS–IU0.6–MPU0.4	170–1,000	60–3.8×103	Substrate	High	Moderate	16,67,69–73
Hydrogels	PAAm, PEDOT:PSS, GO/PVA/PAA/NHS ester	20–150	2.6–460	Substrate, adhesive layer, electrode, interconnections	Moderate	High	15,16,62,74
Inorganic materials	MoS2, graphene, AgNWs, CNTs	100–200	NA	Electrode, interconnections, functional unit	Moderate	Poor	11,34,63,75,221
Composites	AgNWs/AuNWs/SBS, P3HT-NFs/PDMS, Ag flasks/PAAm/alginate, EGaIn/PU	200–1,170	10–3.28×103	Electrode, interconnections, functional unit	Poor	Moderate	77,79–82

CNTs, carbon nanotubes; EGaIn, eutectic gallium-indium; GO, graphene oxide; IU, isophorone bisurea; MPU, methylenebis(phenyl urea); NA, not applicable; NHS, N-hydroxysuccinimide; NWs, nanowires; P3HT-NFs, poly(3-hexylthiophene-2,5-diyl) nanofibrils; PAA, poly(acrylic acid); PAAm, polyacrylamide; PDMS, polydimethylsiloxane; PEDOT:PSS, poly(3,4-ethylenedioxythiophene): poly(styrenesulfonate); PET, polyethylene terephthalate; PI, polyimide; PU, polyurethane; PVA, polyvinyl alcohol; PVDF, polyvinylidene difluoride; SBS, poly(styrene-butadiene-styrene); SEBS, polystyrene-ethylene-butylene-styrene.

Softness.

Soft tissues are typically characterized by a low Young’s modulus, reflecting their intrinsic elasticity to maintain tissue structure and support various physiological processes. For example, neural and muscle tissues have low moduli of around 1–100 kPa (refs. 83,84), whereas skin has a modulus in the range of 0.85–4.5 MPa (refs. 85,86). Therefore, to establish a compliant and durable connection, bioelectronics should have low modulus levels to ensure uniform strain distribution between soft bioelectronics and the tissue surface and prevent concentrated strains and displacement or detachment of bioelectronics under human body activities, thereby improving their stability and eliminating interface impedance instability motion artefacts. Moreover, tissue-level softness allows conformal contact with living tissues while minimizing mechanical damage of tissue, inflammation and skin irritation during soft bioelectronic integration⁸⁷.

A range of low-modulus materials has been employed in the development of tissue-level soft bioelectronics. For example, hydrogels with tissue-level moduli can be applied in implantable bioelectronics intended to monitor physiological activities, such as neural, cardiac and muscular signals, to allow conformal integration with neural and muscle tissues. Wearable bioelectronics can be based on elastic materials, such as silicone (~60 kPa), PDMS (0.2–3 MPa), polystyrene-ethylene-butylene-styrene (2.83 MPa) and polyurethane (0.5–3.8 MPa), to match the moduli of the skin^88,89 and enable conformal and long-term stable bioelectronic–tissue interfaces for continuous physiological monitoring.

Device structure engineering

The conformability and stretchability of soft bioelectronic systems can be greatly enhanced through device structure engineering, ensuring a dynamic conformal contact between soft bioelectronics and living tissues during human body activities. Device conformability and structural stretchability can be improved through thickness reduction, strain adjustment and deformation-tolerant design.

Thickness reduction.

Reducing the thickness of soft bioelectronics improves adhesion and thus contact at the bioelectronic–tissue interface (Fig. 3c) as the mechanical properties and energy dynamics at the bioelectronic–tissue interface determine the interfacial contact states. To investigate the impact of device thickness reduction at the bioelectronic–tissue interface, an analytical mechanical model can be applied that delineates the morphology of the interface⁹⁰. The total energy related to the interfacial contact (U_interface) can be mathematically expressed as:

$$U_{\text{interface}}=U_{\text{device}\_\text{bending}}+U_{\text{skin}\_\text{elasticity}}+U_{\text{adhesion}}$$ (3)

where U_{device_bending}, U_{skin_elasticity} and U_adhesion represent the bending energy of the device, the skin elastic energy and the adhesion energy of the bioelectronic–tissue interface, respectively. Among these, U_{device_bending} is influenced by device thickness. If the thickness of the device reduces to a point at which the adhesion energy exceeds the sum of bending energy and skin elastic energy, the gap at the bioelectronic–tissue interface is eliminated, thereby enabling conformal contact and allowing the device to adapt to irregular shapes of living tissues. Thus, the flexibility of bioelectronics based on intrinsically stiff materials (for example, polyimide and polyethylene terephthalate) can be increased by designing an ultrathin membrane structure (~10 μm)^91,92 to engineer and maintain a conformal and adaptable interface with dynamically evolving microscopic surface topographies, thereby reducing the impact of motion artefacts on impedance stability.

Strain adjustment.

Strain adjustment refers to the manipulation of strain distribution in soft bioelectronics through device structure engineering, aiming to achieve structural stretchability and bioelectronic–tissue interface stability. Typically, applied strain induces geometric changes and performance variations in the active regions of the device. Device structure engineering can be applied to distribute the strain experienced by active regions to certain deformable regions, resulting in a patterned strain distribution throughout the entire soft bioelectronic system. Implementing strain-adjusting structures, such as an island bridge or elastiff layer, can thus increase structural stretchability, accommodating external strains exceeding 100%.

The island bridge structure seamlessly couples highly stretchable ‘bridges’ with rigid multifunctional ‘island’ devices^93–95 (Fig. 3d). In this configuration, the strain exerted on soft bioelectronics predominantly affects the thin and narrow bridge structures, while not impacting the functional rigid islands. Consequently, if the island bridge structure is subjected to tensile or compressive stress, these deformable connecting bridges are primed for mechanical responses that accommodate complex strain distributions and the active device islands ensure stable performance characterized by high bending stiffness.

Mechanical stability and stretchability can also be improved by including a patterned region of mechanical heterogeneity, referred to as the elastiff layer, to adjust strain distribution^16,96. The patterned elastiff layer provides areas of local stiffness and thus strong adhesion to the substrate, enabling the system to withstand substantial shear stresses during strain-induced deformation. The strain can be adjusted to the lower-modulus interspaces between the active devices, thereby enabling soft bioelectronics to achieve conformable and enduring interfaces while maintaining long-term stability and high stretchability.

Deformation-tolerant design.

Soft bioelectronics that contain rigid and brittle materials, such as metals, silicon and plastics, can be made stretchable through patterning of deformation-tolerant device structures, including wrinkles, serpentines, fractals, kirigami structures and textiles. Wrinkles can be engineered to introduce pre-strain into the device during fabrication^97–100 (Fig. 3e). Upon relaxation of the pre-strain, compressive forces generate complex and wavy patterns of relief through a non-linear buckling process, thereby providing reversible stretchability. Geometric yielding of ultrathin filamentary serpentine traces under stress allows tunability and conformability to living tissues^{70,77,101,102}. Similarly, fractal structures, such as self-similar serpentine geometries, can be included throughout the device to accommodate high levels of stretchability and conformability^103–106. The art of kirigami involves patterns created through cutting and folding, enabling 2D starting materials to undergo 3D structural deformations. This technique achieves high stretchability through structure stretching and bending^107–109. Finally, by leveraging interlocking yarn loops crafted through knitting, a soft and stretchable fabric can be designed that exhibits high shape conformability and stretchability^110,111. In addition, textiles can endow the device with breathability and improve the comfort of wearable bioelectronics^{7,8,112–114}. Such deformation-tolerant designs help mitigate interface impedance instability motion artefacts.

Bioelectronic–tissue adhesion

The constant movements of living organisms result in dynamically evolving microscopic surface topographies that affect the adhesion of soft bioelectronics, which can be addressed by introducing diverse forces and bonds across the device. Soft bioelectronics can adapt and adhere to living tissues through physical interactions, capillary forces, surface microstructures and covalent chemical bonds. Improving adhesion ensures a robust connection with the tissue, thereby eliminating interface impedance instability motion artefacts.

Physical interactions.

Physical interactions, including van der Waals forces, hydrogen bonds and electrostatic interactions, can form at the device–tissue interface through non-covalent bonds^115,116. The nature of this adhesion is determined by the bioelectronic–tissue interfacial energy⁹⁰. If the adhesion energy exceeds the sum of bending energy and elastic energy, physical interactions drive the contact adhesion of ultrathin soft bioelectronics, ensuring their conformal attachment while mitigating impedance instability motion artefacts at the interface.

Van der Waals forces arise from the attraction and repulsion of transient shifts in electron density between molecules (Fig. 3f). Once in contact with living tissues, soft bioelectronics establish stable connections through van der Waals adhesion forces. For example, a molybdenum disulfide-assembled membrane directly adheres to soft living tissues through highly conformal interfaces via van der Waals forces, imparting electronic functions to living organisms⁶³. In ECG monitoring, conventional Ag/AgCl electrodes typically suffer a reduction in signal-to-noise ratio from 44.3 dB to 28.5 dB owing to motion artefacts⁶³. By establishing conformal gate transistors, the membrane maintains stable ECG signals without deviations, achieving signal-to-noise ratios of 49.8 dB and 49.2 dB before and during human motion, respectively. In addition, the adhesive layer can seamlessly integrate with living tissue surfaces through hydrogen bonds, which are primarily formed through interactions with functional groups, such as amino, carboxylic and thiol groups, present on the adhesive surface^117,118. Such interactions facilitate the long-term adherence and functionality of soft bioelectronics.

Living tissue surfaces usually carry either negative or positive charges and polar groups, offering various anchoring sites for adhesive materials equipped with oppositely charged and/or polar groups^119–121. The positive and negative charges induce the generation of localized electric fields, thereby driving the electronic materials with opposite charges to form a robust adhesive interface with living tissues through electrostatic interactions. The introduction of these physical interactions to the bioelectronic–tissue interface holds the potential to bridge bioelectronic–tissue gaps and bestow interface impedance stability¹²².

The adhesive interface, featuring multiple physical interactions, is achieved through material design in the adhesive layer. For example, hydrogels incorporating polydopamine exhibit an adhesive strength to tissues that relies on physical interactions formed between the adhesive and tissue surface^23,123,124 (Fig. 3g). Physical adhesion, including hydrogen bonding between catechol groups and amino groups, as well as π–π stacking interactions with aromatic groups, can drive the stable adhesion of thin bioelectronics to tissues^125–127. Similarly, polymers doped with tannic acid containing gallol groups can provide stable contact between bioelectronics and tissue through a series of physical interactions, including hydrogen bonds, electrostatic interactions and cation–π interactions¹²⁸.

Capillary forces.

Physical interactions play a crucial role in facilitating the adhesion of bioelectronics under dry conditions. However, in a wet environment (for example, implantable devices on moist dynamic organs or wearable devices on perspiring skin), physical interactions are attenuated owing to the charge screening effect from highly polarized water molecules, resulting in a notable deterioration of adhesive properties¹²⁹. By contrast, capillary forces, resulting from surface tension at the solid–liquid–gas interface and the Laplace pressure within the liquid¹³⁰, drive the adhesive behaviour of bioelectronics in wet environments. A model to analyse capillary forces involves the formation of a liquid bridge between two parallel flat plates^131–133 (Fig. 3h). In this model, the parallel flat plates and the liquid bridge represent the living tissue or device and the liquid, respectively. The capillary forces (F_cap) can be quantified as:

$$F_{\text{cap}}=-2\pi \alpha \gamma \text{sin}\theta +{\pi \alpha }^2\Delta p$$ (4)

$$\Delta p=\gamma \left(\frac{1}{r_1}+\frac{1}{r_2}\right)$$ (5)

where a and θ represent the contact circle radius and the contact angle at the bridge–plate intersection, respectively; Δp is the local pressure change at the interface, as determined by the Young–Laplace equation; r is the principal radius of curvature; and γ is the surface tension at the liquid–gas interface. Based on this analysis, capillary forces can drive the conformal wrapping process of a thin film on moist tissue surfaces, achieving a high degree of conformality through physical and stable adhesion of the bioelectronic device to complex curvilinear surfaces¹³⁴.

Surface microstructures.

Attaching wearable bioelectronics to the skin remains challenging owing to the rough morphology of skin, the presence of hair and on-surface body fluids such as sweat, oil and blood¹³⁵. Inspired by adhesive phenomena observed in nature, such as gecko footpads and octopus suckers, surface microstructures have been designed for soft bioelectronics that harness strong physical forces, offering reversible and robust adhesion properties on wet and rough skin surfaces, and thereby eliminating the risk of device detachment.

For example, a micropillar structure, inspired by micro-hierarchical and nano-hierarchical structures in the toe pads of geckos, provides surface roughness for reversible adhesion^136,137. In particular, micropillar structures with a high aspect ratio and wide spatula tips maximize the contact area between bioelectronics and the skin (Fig. 3i), achieving a cyclically high normal adhesive force (~1.3 N cm^–2) through collective van der Waals forces. Consequently, such modified bioelectronics can adhere to human skin, even in dynamic conditions and wet environments¹³⁸. Similarly, inspired by the octopus suction cup, which enables reversible adhesion in dry and wet conditions, micro-sucker structures have been engineered to facilitate stable adhesion through the generation of negative pressure by volume changes in the suction chamber combined with van der Waals forces (in dry contact) or capillary action (in wet contact) at the interface^139,140. Soft bioelectronics based on a micro-sucker structure exhibit a high normal adhesive capability (~1.89 N cm^–2) to the skin, enabling accurate biomonitoring without requiring external pressure^141,142.

Covalent chemical bonds.

Establishing a tissue interface through physical adhesion has several limitations. Physical interactions are unstable in dynamic tissues owing to their weak strength, and typically only provide adhesion for thin devices of less than 5 μm in thickness^134,143,144. Surface microstructures depend on external preloads and pull-off adhesion forces to achieve repeatable adhesion, typically around 35 kPa in the normal direction¹³⁹. Alternatively, sufficient adhesion force and minimal detachment in wet conditions can be achieved by covalent chemical bonding at the bioelectronic–tissue interface^145–147, also enabling on-demand and non-invasive debonding. In particular, low-modulus hydrogels can be applied to form covalent chemical bonds, involving both physical adhesion and chemical bonding. Upon contact with the wet tissue surface, the carboxylic acid groups in the hydrogel can rapidly remove interfacial water while forming physical adhesion through hydrogen bonds and electrostatic interactions with the tissue surface. Subsequently, the functional groups in contact with the tissue surface can further form covalent chemical bonds. For example, N-hydroxysuccinimide ester groups can react with primary amine groups to form amide covalent bonds^14,15,148 or two thiol groups can form disulfide bonds¹⁴⁷. These covalent bonds provide long-term stability for the integration of soft bioelectronics in wet and dynamic physiological environments.

Biopotential motion artefact management

In contrast to the unpredictable and irregular frequency variations associated with interface impedance instability motion artefacts, biopotential motion artefacts typically occur within specific frequency spectra, often distributed in higher frequency ranges. Biopotential signals from physiological activities of nervous and muscular tissues typically have an amplitude ranging from a few millivolts to several hundred microvolts^149,150, including 0.1–100 Hz for ECG signals¹⁵¹, 1–30 Hz for EEG signals^152,153 and 20–500 Hz for EMG signals^154,155 (Fig. 4a).

Fig. 4 |. Biopotential and mechanical motion artefact management.

a, Frequency spectrum of various biopotential and mechanical signals. b, Electrophysiological signal sensing mechanism. c, To address motion artefacts and reduce the computational burden of post-signal processing, a signal compensation circuit can be designed for artefact signal reduction and a soft functional transistor allows filtering of mechanical motion artefacts based on vibration frequencies. d, To provide artefact signal post-processing, filtering algorithms can be used to selectively eliminate unwanted frequency components associated with artefacts. To isolate target signals from artefacts in complex patterns and varying conditions, machine learning models can be trained to extract physiological features from biosignals. e, Biochemical signal sensing mechanism. f, Biomechanical signal sensing mechanism. g, To eliminate motion artefacts from the initial sensing stage, sensors can be designed using dampers with band-pass filtering functions and strain isolators to prevent irrelevant mechanical motions from propagating into the sensing area. ΔB, magnetic flux density changes; ΔC, capacitance change; ΔP, pressure change; ΔQ, transferred charge; ΔR, resistance change; ECG, electrocardiogram; EEG, electroencephalogram; EMG, electromyogram; GOx, glucose oxidase; I_DS, drain-source current; ox, oxidized analyte; red, reduced analyte; V_GS, gate-source voltage.

Biopotential motion artefacts are often observed in the sensing of electrophysiological signals, such as ECG, EMG, EOG or EEG signals, with flexible electrodes^17,18,51 because the bioelectronic–tissue interface, responsible for capturing the target electrophysiological signals, is susceptible to interference by biopotential motion artefacts arising from body activities (Fig. 4b), which may be managed through circuit design and algorithmic intervention.

Circuit design

Biopotential motion artefacts are inevitably collected by large-area, skin-interfaced interconnections and electrodes within the circuit but can be mitigated through tailored circuit design, for example, through signal compensation, which is the process of removing noise from target signals by noise reduction strategies. A typical signal compensation circuit contains two functional modules, the acquisition channel and the compensation channel (Fig. 4c). The acquisition channel can detect both the target signals and irrelevant biopotential signals^55,156,157. The compensation channel is strategically deployed at specific locations on the human body to exclusively capture motion artefact signals. The differential between these two channels results in the reduction of artefact signals, thereby diminishing the interference of motion artefacts with target signals. This strategy can be applied to eliminate EMG biopotential artefact signals during ECG monitoring; by introducing a parallel compensation channel positioned closely to the ECG acquisition channel, similar EMG artefact signals can be detected in both channels^55,156. As grip-induced EMG artefacts increase, the Pearson correlation coefficient of the compensated ECG remains at 0.93, while the Pearson correlation coefficient of the uncompensated ECG drops to 0.77 (ref. 156). Therefore, the compensation circuit design strategy cannot only diminish biopotential motion artefacts but can also reduce the computational burden of the system, thereby simplifying the processing of biosignals before their final interpretation.

Algorithmic intervention

The specific frequency spectra of biopotential motion artefacts render algorithmic intervention an effective post-processing strategy for artefact signal management¹⁵⁸. Algorithmic intervention can be categorized into two types, defined by their fundamental approach and intended purpose: filtering algorithms and machine learning.

Filtering algorithms.

Filtering algorithms play a crucial role in electrophysiological signal processing, systematically enhancing the quality of recorded biosignals by selectively modifying or eliminating unwanted frequency components associated with artefacts¹⁵⁹. Various transformative algorithms, including fast Fourier transform, Laplace transform and wavelet transform, are fundamental for comprehending signal properties and characteristics (such as frequency, time and complexity), thereby allowing efficient and effective filtering^160–162. Different types of filtering algorithm, such as Butterworth, finite impulse response, infinite impulse response and elliptic filters, possess specific characteristics and applications for signal filtering, and filtering tasks typically include low-pass, high-pass, band-pass and band-stop filtering¹⁶³ (Fig. 4d). For example, through well-designed cascade-adaptive finite impulse response filters, ECG, EOG and EMG artefacts in EEG records can be considerably attenuated without compromising the core information embedded in these EEG signals^164,165.

Machine learning.

Machine learning represents a more dynamic and data-driven approach in biosignal post-processing^166–168. Instead of relying on predefined rules, machine learning models are trained on labelled or unlabelled data sets to learn patterns and relationships within data¹⁶⁹. These models use the acquired knowledge to make predictions, classify signals or identify anomalies associated with artefacts. Independent component analysis is a commonly used machine learning method capable of isolating electrophysiological signals, such as EEG, EMG and ECG, even when mixed with biopotential motion artefacts, into their intrinsic and independent source components¹⁸. In addition, neural networks have been explored for motion artefact management; in particular, deep learning-based neural networks are effective in dealing with artefacts that exhibit complex and non-linear patterns (such as some biopotential motion artefacts from EMG and ECG), enabling the algorithm to adapt and generalize across varying conditions^17,170,171. Such approaches improve biosignal quality by learning and extracting relevant features from data, ultimately leading to more precise post-processed biosignals.

Mechanical motion artefact management

Mechanical motion artefacts typically exhibit frequency spectra below 10 Hz and can impact diverse sensing mechanisms; however, they can be addressed by systematic management strategies, including sensor design, functional circuit design and algorithmic intervention.

Artefact impact analysis

Mechanical motion artefacts, stemming from biomechanical activities such as body movements (for example, walking, running, jumping and speaking) and physiological functions (for example, breathing, heartbeat and eye blinks), typically occur below 10 Hz (Fig. 4a) and can influence the measurement performance of soft bioelectronics. The degree and nature of their impact depends on the sensing mechanism, including biochemical, biomechanical and other soft sensors.

Biochemical sensors.

Biochemical sensors, including enzymatic or ion-selective, voltammetric, and bioaffinity sensors, can be integrated into soft bioelectronics to enable continuous monitoring of human health at the biomolecular level (Fig. 4e). However, their operational stability is susceptible to mechanical motions^172–176; for example, mechanical motion-induced device deformation may alter the geometry of the flexible microchannel and affect the stability of microfluidic flow. Furthermore, current signal acquisition techniques (potentiometry, amperometry, voltammetry and impedance spectroscopy) are affected by the dynamic state of biofluids that contain the biomarkers, which may cause signal drift and performance deterioration during body activities^3,176.

Biomechanical sensors.

Biomechanical sensors can monitor respiratory activity, heart rate, pulse wave and muscle contraction by establishing a conformal bioelectronic–tissue interface with living tissues⁷ (Fig. 4f). In particular, self-powered sensors leveraging triboelectric, piezoelectric and magnetoelastic effects^177,178 can convert mechanical energy into electric current or voltage signals, for example, for self-powered electrotherapeutics and energy harvesting from human body motions^7,112,114. Nonetheless, owing to their self-powered sensing mechanisms, these sensors are particularly vulnerable to extraneous mechanical movements, which can introduce electric current or voltage signals that overlap with the target biomechanical signals, resulting in the occurrence of mechanical motion artefacts. For example, mechanical sensors designed to monitor artery pulse waves on the wrist^54,179,180 are affected by hand movements; in arteries, the variation in vessel length during a pressure pulse is negligible whereas the circumferential strain varies significantly, ranging from 2% to 18%¹⁸¹. However, the skin proximal to the wrist joint undergoes 30–40% strain when the joint is moved from full extension to full flexion¹⁸². Therefore, disturbances from hand movements can lead to skin deformation at the location of the sensor, thereby introducing mechanical motion artefacts that overlap with the actual pulse waves. This interference can degrade signal quality or render it unusable.

Mechanical motion artefacts are also frequently observed in electro physiological, temperature and optical sensors. Deformations from biomechanical activities are inevitably transmitted to soft bioelectronics, and these strain-induced disturbances along the interconnection and sensing areas lead to baseline drift and abnormal spikes in the signal. However, the artefact signals related to specific body activities exhibit consistent frequency features, including frequency spectra and propagation directions that are distinct from target signals.

Sensor design

Eliminating motion artefacts at the initial sensing stage through specific sensor design strategies may reduce post-signal processing tasks.

Material-enabled damper.

A material-enabled damper is capable of absorbing specific mechanical vibrations within predefined frequency ranges¹⁸³ using a band-pass-filtering gelatin–chitosan hydrogel inspired by the viscoelastic cuticular pad of spiders. The hydrogel can eliminate the transmission of mechanical artefact signals arising from biomedical activities such as gait motion, heartbeat and respiration, which occur within a frequency range below 30 Hz. Its frequency-dependent phase transition, shifting from a rubbery state for the attenuation of low-frequency noise to a glassy state for the transmission of high-frequency signals (Fig. 4g), allows it to function as a versatile filter, facilitating the acquisition of mechanical motion artefact-free signals, including electrophysiological signals (such as ECG and EEG) and high-frequency biomechanical signals (such as acoustic signals), thereby minimizing post-signal processing.

Strain isolator.

The effects of mechanical motion artefacts can also be mitigated by combining a strain isolator with the sensors to isolate interfering motions. Skin deformations, caused by body movement, can propagate as mechanical signals to the sensing area, causing signal artefacts. The integration of a strain isolator into soft bioelectronics enables the maintenance of conformal contact with the skin while safeguarding against excessive skin strain and vibration during body motion^21,184,185. In this way, irrelevant mechanical motions can be isolated from the sensor, ensuring the accurate measurement of target signals (Fig. 4g). This approach can be applied for physiological monitoring, such as ECG, heart rate and respiratory rate monitoring during a range of human activities^21,184 as well as for wearable biochemical and biomechanical sensors^3,54,186.

In biochemical sensing devices, increasing material stretchability provides a partial solution to mitigate motion artefacts; however, long-term stability in wet environments remains a challenge. A strain isolation strategy can be applied to design motion artefact-free biochemical sensors. For example, a biomarker signal can be vertically transmitted from a deformed region to a different layered strain-isolated region through an interconnection to prevent the sensing of motion artefacts resulting from biomechanical activities¹⁸⁶. Similarly, sensor units can be constructed within unaffected rigid islands, while the strain is adjusted by stretchable bridges¹⁸⁷.

Circuit design

The circuits of soft bioelectronics enable the propagation and preprocessing of signals. By incorporating signal compensation circuits and soft functional transistors in the circuits, mechanical motion artefacts can be diminished, reducing the computational burden of the monitoring system.

Signal compensation circuit.

Similar to the compensation strategy for biopotential motion artefacts, a signal compensation circuit for mechanical motion artefacts reduces artefact signals through the differential between the compensation channel and the acquisition channel. Mechanical motion artefacts generated in local areas of the human body during physical activities have almost identical characteristics. To minimize artefact signals, the compensation channel should be located at a position away from the target signal sensing area but where it can capture the same mechanical motion artefacts. In this configuration, mechanical motion artefacts can be separately reduced through the differential process between the acquisition channel and the compensation channel⁵⁵. For example, during cardiopulmonary monitoring by wearable electronics with accelerometers, the acquisition channel and compensation channel are located at the suprasternal notch and sternal manubrium, respectively. These dual-sensing devices allow measurements of diverse physiological parameters, including heart rate, sounds, respiratory activities and body temperature, while maintaining insensitivity to body motions during routine daily activities⁵⁵. During movements (such as walking, running and jumping), the ratio of the mean difference to the standard deviation of heart rate (beats per minute (bpm)), extracted using a compensation circuit, improves from 2.23/13.92 bpm (normal) to 0.01/2.71 bpm (differential), providing accurate monitoring during daily body activities⁵⁵.

Soft functional transistor.

Soft transistors are central components in soft bioelectronics, serving as switches or amplifiers^{82,91,188,189}. Unlike transistors made from rigid materials (such as metals and silicon semiconductors), soft transistors are typically constructed from intrinsically stretchable materials¹⁹⁰, including polymers, carbon-based compounds or other organic molecules^11,191–193. Soft transistors, such as organic field-effect transistors, can be incorporated into soft bioelectronics to amplify weak biopotential signals and conduct analogue filtering of irrelevant biomechanical signals. Combined with other functional components, such as capacitors and resistors^96,194 (Fig. 4c), these transistors undertake amplifier and filter functions that are vital in boosting signal strength while preserving its inherent characteristics, thereby suppressing signal artefacts at specific frequencies¹⁸⁸. For example, a mechanically flexible amplifier made of organic thin-film transistors and thin-film capacitors can be attached to human skin, amplifying differential input signals while suppressing common-mode noise and mechanical motion artefacts during EEG or ECG monitoring¹⁹⁵. Soft functional transistors also play a key role in the analogue-to-digital conversion of bioelectronics, which refers to the encoding of sensory information at the front end of the sensor. By converting artefact-sensitive analogue signals into discrete, uninfluenced digital signals, the resilience of soft bioelectronics against interference from mechanical motion artefacts can be further enhanced¹⁷.

Algorithmic intervention

The distinct characteristics between mechanical motion artefacts and target signals render algorithmic intervention an effective post-processing strategy for artefact signal management^179,196. Algorithmic intervention can be categorized into two types: filtering algorithms and machine learning (Fig. 4d).

Filtering algorithms.

In contrast to analogue filtering, which requires physical hardware to conduct signal filtering tasks, digital filters, based on various filtering algorithms, have higher adaptability and tunability when dealing with complex and variable mechanical motion artefacts. Assisted by transformative algorithms, the cutoff frequency range can be chosen to achieve precise artefact signal mitigation. For example, in electrophysiological signal monitoring, mechanical motion artefacts, resulting from body movements and physiological functions with frequencies below 30 Hz, can be addressed using Butterworth-based band-pass or high-pass filters¹²⁷. These filters, equipped with cutoff frequencies that cover the target signals, block undesired mechanical motion artefacts¹⁹⁶. Biomechanical signal monitoring can respond to full-body kinematics but is also susceptible to interference from mechanical motion artefacts, which disrupt the target information. A high-pass filter (>10 Hz) can selectively capture biomechanical signals across high-frequency ranges, including acoustic vibrations and cardiac activity, while minimizing interference from respiration and locomotion at low frequencies^197,198. For example, based on the filtered signals, the data classification accuracy reaches 0.9 ± 0.08 for coughing and 0.88 ± 0.1 for speaking with a convolutional neural network model¹⁹⁶.

Machine learning.

Principal component analysis (PCA), which is an unsupervised machine learning algorithm, also enables mechanical motion artefact management. Through a dimensionality reduction calculation, the target signals and artefact signals can be differentiated into distinct principal components. For example, during photoplethysmography for the continuous monitoring of metrics related to vascular resistance, PCA can separate target signals for pulsatile capillary blood volume changes and mechanical motion artefacts into the first and second principal components, respectively^199,200. Employing PCA for feature extraction from blood pressure following pulse reconstruction improves the mean blood pressure from a mean difference and standard deviation of −2.26 and 13.79 mm Hg (in the presence of motion) to −0.19 and 5.65 mm Hg (after artefact elimination), thereby enhancing signal estimation quality²⁰⁰.

More complex data characterized by non-linear and high-dimensional features may be better addressed by neural networks. For example, wearable pressure sensors can achieve continuous and precise blood pressure measurements with the assistance of a neural network trained against mechanical motion artefacts^54,179. In addition, convolutional neural networks have shown high diagnostic performance for shockable and non-shockable arrhythmia classifications with robust tolerance against mechanical motion artefacts from muscular activities²⁰¹. However, if target signals and mechanical motion artefact signals share similar frequency bandwidth, algorithmic intervention strategies may be less effective. Under these circumstances, sensor and circuit design approaches, such as compensation circuits, strain isolators and material-enabled dampers, should be considered.

In addition to biomechanical signals, various other physical parameters, including temperature, vessel volume and tissue impedance, can be detected for physiological activity monitoring. Sensing mechanisms for these physical parameters typically exhibit lower sensitivity to motion artefacts and could thus be leveraged to achieve motion artefact-free detection. For example, physiological monitoring based on thermal signal transmission could substantially reduce the sensitivity to motion artefacts, providing quantitative monitoring of blood flow velocity and direction within the microvascular system^202,203. Additionally, the interaction between light and living tissue could be harnessed to develop photoplethysmography sensors for the monitoring of cardiac output and pulse waveform features¹⁹⁹. Moreover, bioimpedance measurements could exploit the penetration of high-frequency current into tissues for haemodynamic parameter monitoring⁵⁵. These deep tissue monitoring platforms, which harness the interaction between various physical signals and living tissues, offer promising avenues for the advancement of physiological monitoring while mitigating the impact of motion artefacts.

Outlook

Motion artefacts pose a major challenge to the advancement of soft bioelectronics for point-of-care applications but can be managed by the aforementioned strategies (Table 2). Other approaches related to the signal detection stage, including intelligent materials, functional circuits and customized service, as well as to the signal processing stage, including distributed information storage and computing, may further mitigate motion artefacts to improve the accuracy and reliability of bioelectric signal measurements (Box 1).

Table 2 |.

Quantitative analysis of management strategies for motion artefacts

Artefact category	Strategy	Signal category	Frequency (Hz)	Performance	Ref.
Interface impedance instability motion artefacts	Ultrathin, physical interactions and mechanical matching	Electrophysiology: ECG/EEG	0–3	SNR: with strategy, 49.2 dB; without strategy, 28.5 dB	63
	Physical interactions	Electrophysiology: ECG	0–3	SNR: with strategy 19.6 ± 1.8 dB; without strategy, 102.7 ± 9.4 dB	127
	Ultrathin, physical interactions and manufacturing	Electrophysiology: ECG	0–3	SNR: with/without strategy approximately the same, ~45 dB	213
	Ultrathin, capillary forces and mechanical matching	Electrophysiology: neural mapping	0–50	Root mean square amplitude ratio: with strategy, 3.6 ± 1.8; without strategy, 5.2 ± 3.9	134
	Surface microstructures	Electrophysiology: ECG	0–3	Signal amplitude: with strategy in wet condition, 74.4%; without strategy, 0% (detach)	138
Mechanical motion artefacts	Machine learning regression algorithm	Arterial blood pressure: Bio-Z waveform	0–3	With strategy: error: 0.06 ± 2.5 mm Hg (diastolic blood pressure), 0.2 ± 3.6 mm Hg (systolic blood pressure)	56
	Supervised feedforward neural network architecture	Pulse waveform monitoring	0–3	SNR: with strategy, 23.3 dB	179
	PCA	Cardiac monitoring: PPG	0.5–5	Precision–recall curve, with strategy: baseline 0.82, exercise 0.98, cold pressor/breath hold 0.73	199
	PCA	Cardiac monitoring: PPG	0.5–3	MD/SD: without strategy, −2.26/13.79 mmHg; with strategy, −0.19/5.65 mmHg	200
	Digital filtering and CNN	Mechanoacoustic: cardiac and respiration signs	20–55, 0.1–1	Accuracy with strategy: 0.9 ± 0.08 for coughing; 0.88 ± 0.1 for speaking	198
	Compensation circuit design	Cardiopulmonary monitoring: heart rate, acceleration	1–10	MD/SD: without strategy, 2.23/13.92 bpm; with strategy, 0.01/2.71 bpm	55
	Damper materials	Electrophysiological and voice: acoustic signal	80–200	SNR: without strategy, <20 dB; with strategy, 60–120 dB	183
Biopotential motion artefacts	Independent component analysis	Electrophysiology: EEG/EOG	0.5–10	SNR: without strategy, −15 to 0 dB; with strategy, 6.65–7.55 dB	18
	Non-linear filtering	Electrophysiology: EMG/ECG	50–150, 0–3	EMG relative error, with strategy: −0.20%	17
	1D-CNN	Electrophysiology: EEG/EOG, ECG, and EMG	0–256	SNR: without strategy, −5 dB; with strategy: 19.8, 19.3, 20.6 dB	170
	Compensation circuit design	Electrophysiology: ECG	5–50	PCC: without strategy, 0.77; with strategy, 0.93	156
	Low-pass filter algorithm	Electrophysiology: EEG	0.1–35	SNR: without strategy, 10 dB; with strategy, 22.05 dB	164

bpm, beats per minute; CNN, convolutional neural network; ECG, electrocardiogram; EEG, electroencephalogram; EMG, electromyogram; EOG, electrooculogram; MD/SD, ratio of mean difference to the standard deviation; PCA, principal component analysis; PCC, Pearson correlation coefficient; PPG, photoplethysmography; SNR, signal-to-noise ratio

Box 1. Translational considerations.

Scale-up manufacturing of soft bioelectronics

Manufacturing scalability and device reliability are paramount to transferring soft bioelectronics from laboratory prototypes to industrial production^10,208. Efficient workflows, cost reduction through material and process optimization, and meeting market demands are achievable through scaled-up manufacturing^222,223. In addition, device reliability is fundamental for integrating soft bioelectronics into clinical monitoring^83,224,225. Improving reliability across material selection, manufacturing processes and performance validation ensures extended lifespan and consistent performance. Furthermore, soft bioelectronics may be integrated into existing wearable systems; for example, electronic functionalities may be embedded in common textiles to create smart textiles with biomonitoring capabilities^8,114.

Customized functions for personalized needs

Customizing the functionality of soft bioelectronics based on specific needs may contribute to personalized and precise health-care monitoring^226–228. The development of personalized soft bioelectronics can be fostered through collaboration with patient communities and health-care experts, for example, to design interfacing strategies based on auditory, tactile or visual principles^227–231.

Establishing clinical gold standards for evaluation

Establishing clinical gold standards for the evaluation of soft bioelectronics requires the continuous monitoring of conventional clinical indicators to create a dynamic physiological data base^232–234. These data can then be analysed in terms of feature extraction, pattern recognition and machine learning to extract physiological insights²³⁵. Calibration on established clinical assessments ensures data alignment and accuracy. Defining and validating measurement standards for various physiological indicators is fundamental to ensuring stable and repeatable performance.

Identification and prediction of clinical needs

Soft bioelectronics should be developed to address key clinical needs, for example, for newborns or critically ill patients^236,237, who require close monitoring of vital signs. Similarly, acute disease prevention has driven the development of continuous and non-invasive monitoring devices. However, prioritizing encryption and security is crucial for medical monitoring, particularly in the context of cloud services and big data, to ensure patient privacy.

Intelligent materials

Current sensor units in soft bioelectronics typically lack the capacity to selectively detect target biosignals while filtering out irrelevant motion artefact signals. Intelligent materials with selective response capabilities or motion-insensitive features may allow the accurate extraction of intended signals within precise frequency ranges and propagation directions. For example, selective dampers could be integrated that capture vibrations within designated frequency ranges to inhibit the propagation of signal artefacts within the sensor¹⁸³. Materials with anisotropic responses may further enable the regulation of sensitivity in relation to biological information across different propagation directions. Thus, the sensor could maintain high sensitivity in the direction of the target physiological signals whilst not detecting disturbances from artefact signal directions.

Motion-insensitive materials could be integrated into sensors to selectively detect target signals at the front end. For example, by introducing conjugated polymer nanostructures into elastomer-based semiconducting materials, the mobility of soft transistors is maintained stable during stretching¹⁹³. Similarly, composite nanofibers of carbon nanotubes and graphene are insensitive to bending in pressure measurements²⁰⁴. These materials can thus isolate interfering mechanical motion artefacts from target biosignals at the sensing stage. Furthermore, the electrical and mechanical properties of polymer materials could be improved for bioelectronic–tissue interfaces. For example, during dynamic motion, fatigue from polymer deformation may lead to detection errors²⁰⁵. Tough and self-healing hydrogel–elastomer materials, featuring multi-strength hydrogen bonding interactions, in which strongly crosslinked hydrogen bonds provide rigidity and elasticity and weakly crosslinked hydrogen bonds facilitate strain energy dissipation through reversible breaking and reformation^23,73,206, may enable a stable interface with tissue. In addition, modulus and electrical property matching between polymer materials and electrodes may be achieved using soft liquid metal composites and adhesive hydrogels to provide interface stability through chemical bonding between polymer layers and electrodes, thereby improving signal recording performance²³. Moreover, adhesive and conductive hydrogels may simultaneously act as conformable adhesive and biosignal recording electrodes^124,128 in soft bioelectronics.

Functional circuits

Motion artefacts may also be addressed and signal detection optimized by using functional circuit components with tissue-level softness and compensation circuits, reducing the need for post-signal processing. Soft circuit components ensure a uniform distribution of stress within the bioelectronic system, establishing a stable and consistent connection with the dynamic surfaces of living tissues. Moreover, functional circuit components, such as soft transistors, capacitors and resistors, equipped with built-in amplification and signal filtering capabilities, can be integrated into the circuit design^{188–192,207,208} to endow the system with artefact signal filtering capabilities and reduce post-processing of data. Soft functional circuits can also encode detected biosignals from analogue to digital, offering anti-interference capabilities against motion artefacts.

Integrating a compensation channel into the monitoring system allows the distinct detection of biopotential or mechanical signal artefacts. The detected artefacts can then be employed to filter out motion artefact signals in the acquisition channel by a straightforward differential process between the two channels. This approach is particularly effective in eliminating motion artefacts from electrophysiological (EMG, ECG, EOG) and biomechanical (breathing, eye blinking, heartbeat) signals, given their specific frequencies and regular waveforms. Similarly, closed-loop circuit design can provide real-time feedback and adjustments^209,210; owing to their energy efficiency and customization potential, closed-loop designs are particularly suitable for individualized health monitoring and fitness tracking, ensuring both the accuracy and reliability of soft bioelectronics.

The clinical translation of soft bioelectronics will further benefit from miniaturized, implantable designs, which require amplifiers with low power consumption and minimal noise, such as complementary metal oxide semiconductor-based amplifiers^211,212. These amplifiers show beneficial power-to-noise tradeoff in biosignal amplifying, which is a key factor in miniaturizing implantable bioelectronic devices and may also improve the signal processing capabilities of soft bioelectronics.

Customization

The local dynamic surface morphology of biological tissues varies among individuals, which is not accounted for in standardized soft bioelectronic devices and may cause motion artefacts. Therefore, individual surface morphology characteristics may be collected for the customized optimization of soft bioelectronics. Furthermore, different parts of the body exhibit distinct surface features and varying physiological signals. Such differences should be considered in the design of soft bioelectronics^213–218.

Distributed information storage and computing

In addition to hardware optimization, information storage, processing and interpretation can be improved in soft bioelectronics. Distributed information storage and computing may be applied to manage physiological data for personalized health care, integrating a network of nodes. Physiological data of individuals could be transmitted to a dedicated cloud server that functions as a distributed information storage system, overcoming the constraint of limited information storage capacity in soft bioelectronics. These data could then be analysed using distributed information computing to match with data bases, thereby improving the accuracy of data classification and allowing the extraction of physiological parameters from biosignals. The data could also be shared with health-care professionals for medical guidance and advice.

Citation diversity statement

We acknowledge that papers authored by scholars from historically excluded groups are systematically under-cited. Here, we have made every attempt to reference relevant papers in a manner that is equitable in terms of racial, ethnic, gender and geographical representation.

Key points.

• Soft bioelectronic systems for the monitoring of human health rely on a stable and conformal bioelectronic–tissue interface.

• Motion artefacts can occur due to body movements and physiological activities, and can negatively affect signal detection and interpretation in bioelectronic measurements.

• Motion artefact management is crucial to the clinical translation of soft bioelectronics to ensure measurement with high accuracy.

• Materials usage, device design, bioelectronic–tissue adhesion, sensor and circuit designs, and algorithmic intervention are effective motion artefact management strategies.