Bringing sensation to prosthetic hands—chronic assessment of implanted thin-film electrodes in humans

The study addresses the need for long-term stable peripheral nerve interfaces (PNIs) that can deliver reliable sensory feedback for bidirectional control of hand prostheses. While PNIs must provide high spatial selectivity, minimal foreign body reaction, and operate within electrochemical safety limits, chronic stability is paramount for clinical translation. Prior solutions (e.g., LIFE, cuff electrodes, FINE, and MEAs) showed trade-offs in selectivity, robustness, or longevity, with documented mechanical and material failures in chronic settings. Polyimide-based transverse intrafascicular multichannel electrodes (TIMEs) demonstrated promising selectivity and subchronic performance, but earlier clinical use revealed adhesion issues in large-area ground contacts. This work investigates whether updated TIME designs—with adhesion-promoting layers and a redesigned split ground—can maintain electrochemical safety, mechanical integrity, and functional sensory performance during up to six months of implantation in human median and ulnar nerves.

The introduction surveys prior PNIs and their chronic performance. LIFE electrodes provided early subchronic success but limited selectivity due to single-contact wires. Thin-film LIFE variants improved channel count but suffered from rising thresholds due to foreign body response. Cuff electrodes offer robustness but limited intrafascicular selectivity; FINE electrodes improve selectivity by reshaping nerves and have been implanted chronically (~3.3 years) but require careful force management. Silicon-based MEAs adapted for peripheral nerves showed subchronic feasibility; however, chronic intracortical data revealed high failure rates driven by mechanical and material issues (e.g., connector failures, corrosion, insulation cracking/delamination). TIMEs, verified in animal models and subchronic human trials, employed SIROF to maintain charges within safe limits but initially exhibited adhesion loss at rectangular ground sites, motivating the use of silicon carbide adhesion layers and a split-ground redesign to mitigate stress concentrations and improve chronic stability.

• Study design and participants: Twelve TIME implants (latest generation TIME-3H_V2 and TIME-4H) were implanted in the median and ulnar nerves of three trans-radial amputees for up to six months (2015–2017; clinical trial NCT02848846). Weekly in vivo impedance measurements and sensory assessments were performed.
• Device design and fabrication: Polyimide-based thin-film electrodes with platinum tracks and SIROF-coated stimulation/ground sites. Silicon carbide was used as an adhesion-promoting layer between PI and platinum. Ground redesign split a previously rectangular ground into 109 redundantly interconnected circular contacts (80 µm exposed diameter) arranged hexagonally. Two active-strip variants were used (TIME-3H_V2 with varying site diameters on one side; TIME-4H with shifted L/R rows to target more fascicles). Implants were assembled via Microflex Interconnection (MFI) to a screen-printed ceramic, a 40 cm helically wound cable, and an Omnetics connector. For patient 3, a protective rubber hose was added at the wire–connector transition for strain relief.
• Thermo-mechanical simulation: COMSOL Multiphysics simulations modeled steam-sterilization-induced stress in rectangular vs split ground designs using PI–platinum sandwiches (standard material properties with specified changes). van Mises stress was evaluated along defined lines near PI edges and mid-contact; integrated stress over 1 mm was compared between designs.
• In vivo electrochemical assessment: Weekly in vivo impedance estimated as voltage at end of cathodic phase divided by current (20 µA, 300 µs, 1 Hz; average of 4 pulses). Functionally working contacts defined as impedance <150 kΩ (stimulator limit). Patient-reported sensation was tracked over time.
• Post-explant electrochemistry: Due to safety cuts during explantation, only one ground contact (TIME-4H | 17-0058, patient 3, ulnar proximal) was intact for EIS before vs after implantation; hydration via cyclic voltammetry performed prior to repeated EIS.
• Optical/material analysis post-explant: Light microscopy with polarization filters, white light interferometry (3D topography), and SEM/FIB imaging of 111/168 stimulation contacts. Five integrity categories defined: 1 (pristine-like), 2 (≤1 µm delam and/or light cracks), 3 (1–6 µm delam and cracks), 4 (partial layer delam), 5 (>6 µm delam, metallization disintegration, or PI compression/destruction). Ground site overviews acquired by SEM.
• Accelerated in vitro stimulation: TIME-like contacts (n=15) stimulated up to 6.5 billion charge-balanced pulses (200 µs/phase, 10 µs delay, 500 Hz, 578 µA; PBS; two-electrode) to probe SIROF durability; voltage recorded and morphology examined.
• Connector analysis: Electrical continuity (resistance) measured between connector and interconnecting ceramic; μ-CT used to visualize wire integrity at the wire–connector transition. Patient 3 devices included strain relief.
• Statistics: Longitudinal, unbalanced in vivo impedance data analyzed with linear mixed effects models (R). Kruskal–Wallis test evaluated impedance differences across integrity categories.

• Stress simulation: Rectangular ground showed extended high-stress zones near PI edges with integrated stress of 3.0×10^11 Nm^−1 (high-stress segment) vs 2.4×10^11 Nm^−1 (low-stress). Split ground maintained similar peak magnitudes but with localized peaks at exposed edges; integrated stress reduced to 1.6×10^11 Nm^−1 over 1 mm, indicating lower spatial extent of high stress.
• In vivo impedance trajectories (linear mixed model): • Patient 1: intercept 94.34 kΩ; slope 0.8 kΩ/week; F=43.1368; p<0.0001; initial ~100 kΩ plateau by week 5; drop after week 22 to 73±18 kΩ when cable manipulated, implicating connector transition issues. • Patient 2: intercept 34.77 kΩ; slope 2.8 kΩ/week; F=57.69014; p<0.0001; stabilized ~100 kΩ by week 11. • Patient 3: intercept 57.95 kΩ; slope 0.8 kΩ/week; F=39.95188; p<0.0001; steady increase from 60±36 kΩ to 67±36 kΩ without mid-trial plateau.
• Functional contacts (impedance <150 kΩ): Patients 1 and 2 declined from ~98% to 73%/68% by week 11, ending at ~50%/54%. Patient 3 declined modestly from 96% (week 1) to 82% at study end.
• Sensory outcomes: Patients 1 and 2 began at ~75% of contacts eliciting sensation until weeks 10–12, decreasing to 34% and 30% respectively by trial end. Patient 3 stabilized at ~82% from week 9 through the end after an initial drop.
• Post-explant EIS (single ground contact): 1 kHz impedance decreased from 508 Ω (pre-implant) to 355 Ω (post-explant); after hydration 354 Ω. Cut-off frequency shifted 5.8→11 Hz post-explant, reverting to 5.4 Hz after hydration, suggesting preserved electrochemical properties and lower access resistance post-cut.
• Optical/material integrity vs impedance: Among 111 analyzed stimulation sites, median impedances across categories 1–5 ranged ~50–67 kΩ (means 46–79 kΩ), all within functional range; no significant impedance differences between categories (Kruskal–Wallis, α=0.05). Most contacts (82%) were category 1–3 (none/low damage). Category 5 often showed signs of external mechanical impact during handling/explant (10/12 contacts), indicating non-intrinsic failure; 10/12 still <150 kΩ, 2 non-measurable.
• Ground contact morphology: Explanted split ground (patient 3, TIME-4H 17-0058) exhibited some cracks and partial delamination in ~50% of sub-contacts but remained electrically functional, supporting redundancy benefits.
• Connector transition failures: Patient 1—only 2/64 channels conductive at connector (179±1 Ω); μ-CT showed 21 connected wires (33%). Patient 2—22/32 conductive (169±40 Ω), μ-CT 26 connected (81%). Patient 3 (with strain relief)—62/64 conductive (171±13 Ω); μ-CT 64/64 connected (100%). Connector strain relief markedly improved reliability.
• Accelerated stimulation: SIROF-coated contacts showed first morphological changes (trident cracks ~20 µm) after ~4.5 billion pulses; up to 6.5 billion pulses tested, indicating high durability and suggesting SIROF stability over years of clinical use under typical duty cycles. No corrosion detected in SIROF layer of explants.
• Overall: TIMEs operated within electrochemical safety limits, delivered consistent amplitude modulation, elicited reliable sensation, and showed no corrosion or detrimental morphological changes attributable to stimulation; observed damages predominantly due to mechanical handling/explantation or connector issues.

The redesigned split ground substantially reduces the spatial extent of intrinsic stress concentrations compared to a continuous rectangular ground, mitigating delamination risks at PI edges. Chronic clinical performance confirmed electrochemical safety and functional sensory feedback across six months. Declines in electrically functional and sensory-eliciting contacts for patients 1 and 2 were traced primarily to mechanical failures at the wire–connector transition due to frequent connect/disconnect cycles; implementing strain relief in patient 3 dramatically improved channel continuity and stabilized both impedance and sensory performance. Post-explant electrochemistry of a ground contact matched pre-implant behavior, while optical analyses showed that most contact degradations were due to external mechanical impacts during explantation rather than in vivo corrosion or stimulation-driven degradation. The redundancy of the split ground maintained function despite localized cracking. Accelerated in vitro testing corroborated SIROF’s long-term electrochemical stability, with only localized cracks appearing after billions of pulses. Collectively, these findings support the feasibility of PI-based thin-film electrodes for chronic sensory neuroprosthetic applications, contingent on robust system-level packaging—especially at connectors—and careful surgical handling.

This work demonstrates that polyimide-based TIMEs with SIROF contacts and silicon-carbide adhesion layers can be implanted chronically in human arm nerves for up to six months, maintaining electrochemical safety and eliciting reliable sensations. The split-ground design reduces intrinsic stress and enhances mechanical resilience through redundancy. System-level improvements, notably strain relief at the wire–connector transition, are critical to sustain channel integrity and clinical performance. No evidence of SIROF corrosion was observed in explants, and accelerated testing supports long-term material stability. Future efforts should focus on fully implantable systems with hermetic packaging, further connector/cable robustness, extended multi-year follow-up, and refined surgical protocols to minimize handling-induced damage, potentially leaving thin-film components in situ at end-of-life to avoid nerve injury.

• Small cohort (three patients) limits generalizability.
• Post-explant electrochemistry was feasible on only one ground contact due to safety-mandated cable cuts, limiting direct pre/post comparisons across contacts.
• Mechanical damage during explantation and handling confounded interpretation of some optical failures (e.g., bending, scratches), making it difficult to ascribe damage to in vivo processes.
• μ-CT resolution (25 µm) constrained precise wire-connector integrity assessments and may undercount partial connections.
• Follow-up limited to six months; long-term performance over several years remains to be demonstrated.
• Impedance-derived functionality threshold (<150 kΩ) is tied to stimulator limits and may not capture all nuances of contact-tissue interface health.