Conformal in-ear bioelectronics for visual and auditory brain-computer interfaces

The study addresses limitations of current BCI hardware—bulky cap-based systems, headband products, and invasive microneedles—that hinder daily use due to inconvenience, restricted application scenarios, and risks of inflammation or irreversible tissue damage. Ear-centered EEG (ear-EEG) offers gel-free signal acquisition enabling discreet, wearable neural monitoring. However, the ear canal’s complex, person-specific geometry and physiological sensitivity make robust, comfortable electrode contact challenging. Existing in-ear solutions often rely on rigid or soft supports (earplugs or 3D-printed attachments) that can cause interface failures, obstruct external sound, or create friction and irritation. The authors propose an in-ear bioelectronic device (SpiralE) that actively conforms to the auditory meatus via electrothermal actuation to achieve stable, comfortable EEG recording for both visual and auditory BCIs, aiming to enable efficient, natural, and practical human–machine interaction in daily life.

Prior ear-EEG systems demonstrated the feasibility of discreet neural monitoring but typically required integration with rigid or soft supports to maintain electrode contact in the curved ear canal. Rigid supports are difficult to fabricate using planar processes, may damage or misalign electrodes on curved surfaces, obstruct external auditory communication, and fail to accommodate circumferential deformation, leading to poor local adhesion. Soft supports improve conformity but can generate strong friction against the sensitive canal during insertion/removal, causing irritation or inflammation. Earlier approaches also faced trade-offs between ensuring tight contact for low-noise recordings and maintaining comfort and safety. These constraints motivated developing adaptive, deformable, and minimally obstructive in-ear bioelectronics capable of reliable EEG acquisition without bulky supports.

Device design and materials: SpiralE is a long, hollow spiral in-ear bioelectronic with in-plane size ~50 mm × 3 mm and circular recording electrodes (~1–1.6 mm diameter). It comprises a multilayer stack: an EEG detection layer (EEGDL) with exposed conductive strips and polyimide insulation; a double-layer thermo-responsive shape memory polymer (SMP) structure (SMP2 with good surface adhesion in rubber state; SMP1 as higher-modulus skeleton); and a stretchable electrothermal actuation layer (EAL) embedded between SMP1 and SMP2 using a flexible conductive mesh for Joule heating. SMPs have melt transition temperatures of ~50 °C (SMP1) and ~37 °C (SMP2). Electrothermal actuation softens and expands the device from a temporarily programmed small spiral to a larger radius spiral, conforming to the ear canal under contact constraints while preserving a hollow lumen for sound passage. Fabrication: Using traditional 2D micro/nanofabrication on PI substrates, conductive layers were patterned for EEGDL and EAL via photolithography and electron-beam evaporation, followed by transfer printing onto SMP substrates (EEGDL→SMP2; EAL→SMP1). The planar stack was reconfigured into a 3D spiral: permanent large spiral shape set on a cylinder mold at 75 °C; temporary small spiral achieved by softening in 50 °C water and wrapping on a smaller cylinder. Integration with electronics used anisotropic conductive film (ACF) to bond to customized PCB; EAL connected to DC power, EEGDL to EEG amplifier via wires. In vitro characterization: Stretchable electrodes (EAL/EEGDL) underwent 400 cycles of 10% tensile strain to assess resistance stability. Electrochemical impedance spectroscopy (EIS) measured electrode impedance in initial, permanently reconfigured, temporary, and recovered spiral states, and across repeated deformation cycles. Sound transmission through the hollow structure was quantified versus earplugs in a water-drop environment. Shape recovery and expansion were observed in tubes (inner diameter ~6.5 mm) and right-angle elbows under Joule heating, with thermal imaging (Fluke Ti400) to monitor temperature; finite element analysis estimated interfacial skin stress under constraint. Safety: Field temperature during actuation limited to ~50 °C; skin surface temperature measured via thermocouple ~40 °C; no redness/swelling after >3 min exposure. Simulated interfacial stresses remained below typical skin pressure thresholds (~20 kPa). Participants and BCI setup: Visual BCI: 9 subjects (7 male, 2 female; mean age 26.9±3.1). Auditory BCI: 5 subjects (4 male, 1 female; mean age 25±1.7). Recordings used a commercial EEG system (Synamps2) with ear channels L1–L5 and R1–R5, common REF/GND. Visual stimuli on 60 Hz LCD; auditory stimuli delivered via two speakers. SpiralE inserted as a small spiral; -10 V DC applied for actuation to conform; impedance checked (<50 kΩ target, with gel applied if needed). Device self-supported after cooling; removed by reheating to lower modulus. Visual paradigms: (i) 10 Hz central-field SSVEP (8 s trials). (ii) Narrow-band 3-target SSVEP (8–12 Hz, 0.5 Hz steps) for SNR/harmonics comparison among in-ear, mastoid, and occipital montages. (iii) 9-target SSVEP BCI with TRCA decoding (cross-validated). (iv) 40-target online SSVEP speller (calibration-free FBCCA; 200 trials; real-time feedback; 6 s inter-stimulus intervals) with phrase spelling demonstrations. Auditory paradigm: Cocktail-party attention decoding (two simultaneous speech streams; subjects covertly attend one). Auditory features computed as onset envelopes from spectrograms (128 sub-bands). Forward modeling estimated temporal response functions (TRFs); backward modeling reconstructed stimuli for correlation-based classification (leave-one-out evaluation). Signal processing: Visual EEG bandpass 0.5–90 Hz; auditory EEG anti-aliasing 2–8 Hz; downsampled to 250 Hz (visual) and 128 Hz (auditory). Automated artifact rejection/interpolation on channels/epochs. Visual decoding used TRCA (offline 9-target) and FBCCA (online 40-target). SNR computed by narrowband ratio to neighboring frequencies. ITR computed using standard formula with number of targets M and accuracy P. Auditory decoding used Pearson correlation between reconstructed and actual stimuli; TRFs examined attention modulation.

- Device conformity and mechanics: Electrothermal actuation adaptively expands the temporary small spiral to a larger spiral, conforming to complex ear-canal curvatures (including right-angle elbows). Modulus decreases during insertion/deformation and extraction reduce friction and discomfort while maintaining sufficient support for stable EEG contact. - Electrode/electronics stability: Stretchable electrodes showed consistent resistance modulation after 400 cycles of 10% strain. EIS across initial, reconfigured, temporary, and recovered states showed minimal variance; electrodes (1.6 mm diameter) had <1 kΩ impedance at 50 Hz in PBS. In vivo ear-channel contact impedances around 50 Hz were 110.6 ± 23.1 kΩ across ear channels. - Acoustic openness: Hollow design preserved external sound transmission, outperforming occlusive earplug designs in decibel measurements, enabling concurrent auditory perception. - Safety: Field temperature during actuation limited to ~50 °C; skin surface measured ~40 °C with thermocouple; no redness/swelling after >3 min contact. Finite element analysis indicated interfacial pressures below 20 kPa threshold. - Visual EEG benchmarks: Clear alpha (~10 Hz) rhythms recorded with in-ear channels. In 10 Hz SSVEP, robust tagging at stimulus frequency and harmonics was observed. - Harmonic characteristics: Across 8–12 Hz narrow-band SSVEPs, in-ear SSVEPs showed stronger 2nd harmonics relative to fundamental compared with occipital montage; fundamental responses localized more to occipital areas, whereas harmonics showed greater inclination toward temporal-parietal regions. - Visual BCI performance: 9-target SSVEP BCI achieved mean decoding accuracy of 95% across 9 subjects (TRCA). For a calibration-free 40-target online speller (200 trials), one subject achieved 75.7% accuracy with ITR peaking at 31.24 bits/min at 5.5 s (29.8 bits/min at 6 s). Overall visual BCI ITR reached 36.86 ± 15.53 bits/min (reported as group-level performance metric), representing state-of-the-art for ear-EEG. - Auditory BCI performance: In a cocktail-party task using only in-ear signals, classification accuracy reached up to 84% using Pearson correlation-based backward modeling. Attention-related TRFs showed clear modulation consistent with top-down control. Combining peripheral and in-ear montages further improved decoding effectiveness. - Reusability and practicality: Device self-supports after cooling, can be reheated for comfortable removal and reuse; 18 devices across 9 subjects used with/without gel; no discomfort reported.

The findings demonstrate that an electrothermally actuated, hollow, spiral in-ear bioelectronic can reliably conform to diverse ear-canal anatomies, providing stable, comfortable contact for high-quality EEG acquisition without obstructing natural hearing. This addresses key barriers to daily-life BCIs by minimizing friction/irritation, avoiding bulky supports, and preserving auditory openness. Visual BCI experiments verify robust alpha and SSVEP detection with competitive accuracies and information transfer rates using only in-ear electrodes, supporting the feasibility of wearable, discreet BCI control. The observed stronger 2nd-harmonic responses in in-ear recordings compared to occipital montages suggest complementary spatial sensitivity to SSVEP harmonics, offering new insights into the spatial propagation of fundamental vs. harmonic components and motivating broader mapping studies. The auditory BCI results in a cocktail-party setting underscore the utility of the hollow design for naturalistic attention decoding, achieving accuracy comparable to literature while using minimally obtrusive in-ear sensors. Overall, SpiralE broadens the spatial monitoring range of ear-centered EEG and provides a practical path toward real-world neural interfaces for communication and cognitive monitoring.

This work introduces SpiralE, a conformal, electrothermally actuated in-ear bioelectronic that achieves stable, comfortable EEG sensing while preserving auditory openness. The device enables high-performance visual BCIs (95% accuracy in 9-target SSVEP; calibration-free 40-target online speller at 75.7% accuracy with high ITR) and effective auditory attention decoding (up to 84% accuracy) in naturalistic scenarios. The distinct harmonic patterns observed in in-ear SSVEPs indicate complementary spatial sensitivity valuable for studying whole-brain SSVEP dynamics. The platform showcases a novel 3D flexible bioelectronics paradigm suitable for discreet, daily-use neural interfaces. Future work should optimize contact force vs. impedance while retaining hollow, small-area contact, explore alternative actuation/driving methods, expand participant cohorts to validate generalizability, and incorporate advanced machine learning to further improve auditory and visual BCI performance.

- Interface impedance: The small contact area and hollow, support-free design can increase interface impedance; future designs should balance external support/contact force with hollow features and comfort. - Algorithmic and dataset limitations: Auditory decoding used simple correlation-based models; more sophisticated machine learning could improve accuracy. Harmonic spatial distribution findings warrant validation in larger, more diverse cohorts. - Sample size and scope: Visual experiments involved 9 subjects and auditory experiments 5 subjects, which may limit generalizability; only specific visual (SSVEP) and auditory attention paradigms were tested.