A flexible protruding microelectrode array for neural interfacing in bioelectronic medicine

The study addresses the challenge of selectively recording low-amplitude neural signals from small, delicate autonomic nerves (often dominated by unmyelinated C-fibers with conduction velocities below 1–2 m/s). Conventional treatments for pelvic organ dysfunctions (e.g., constipation, fecal incontinence, urinary dysfunction) do not target the underlying neural pathways. Bioelectronic medicine aims to read and modulate neural activity in closed-loop systems, requiring interfaces with high selectivity and minimal invasiveness. Key constraints include electrode impedance (which impacts thermal noise and SNR), electrode size (trade-off between impedance and selectivity), electrode surface modifications (e.g., TiN, IrO2 to increase electrochemical surface area while maintaining geometric size), spatial resolution, and minimizing electrode-to-source distance. Extraneural devices (e.g., cuffs) are low-invasive but lack selectivity; intraneural devices (e.g., LIFE, TIME, USEA) increase selectivity but often rely on rigid substrates unsuitable for small autonomic nerves. The research question is whether a flexible, polymer-based 3D microelectrode array with protruding, micrometer-scale electrodes and low impedance can provide selective, higher-SNR recordings from small autonomic nerves by reducing the distance to axons while maintaining flexibility and biocompatibility.

The paper reviews neural interface categories and prior art: extraneural cuff electrodes offer longevity and low invasiveness but primarily capture overall nerve activity rather than single units. Intrafascicular devices such as LIFE (longitudinal wire) and TIME (transverse thin-film) offer higher selectivity, with DIME extending coverage via multiple LIFEs but complicating implantation. The USEA provides a 3D array capable of interfascicular access and has been applied from cortex to bladder control; HD-USEA recorded feline pudendal nerve activity but its rigid silicon base and relatively large electrodes limit suitability for chronic recording in submillimeter autonomic nerves. Advances combining 3D arrays with flexible polymers (polyimide, parylene C, PDMS, LCP) show promise. Reducing electrode impedance through larger geometry sacrifices selectivity, whereas increasing electrochemical surface area via coatings (TiN, IrO2) preserves geometric size and enables higher-density arrays. Minimizing electrode-to-source distance improves SNR; thus intraneural approaches are favored for selective autonomic nerve interfacing. The literature motivates a need for flexible, protruding 3D MEAs with modified surfaces to lower impedance while maintaining small feature sizes for autonomic nerve applications.

Design: A flexible polyimide-based 3D MEA targets autonomic nerves and plexuses. The array comprises 36 electrodes in total (including two references and two grounds), with 32 recording electrodes arranged over two geometric sections: a broader area for plexus mapping and a narrow meandering tail for constrained spaces along nerves (e.g., vagus/hypogastric). Handling features (rings), anti-bubble holes (35–70 µm²), a flexible ribbon for mechanical decoupling, and a finger-structured connector facilitate surgical handling and connection. Protruding electrodes reduce distance to neurons. Electrodes: Cylindrical gold pillars of 20 or 50 µm diameter with height ~60 µm are fabricated atop a polyimide thin-film substrate. Sidewalls are insulated with parylene C; electrode tips are selectively exposed and modified. Two variants were produced: Type A (parylene-insulated sides, wet-etched head) and Type B (same plus TiN coating on the head). Initial 2D MEA fabrication: Standard MEMS processing produced a flexible PI substrate with Au conductors and 2D Au electrode sites (details in supplementary Fig. S1). Pillar fabrication (electrodeposition): A Ti (~50 nm)/Au (~100 nm) seed layer was sputtered on the 2D substrate and patterned with photoresist to form a ~70 µm-thick mold. The wafer was hydrophilized with O2 plasma and electrochemically cleaned by cyclic voltammetry (0.5 M H2SO4, 15 cycles, 0.4–1.5 V vs Ag/AgCl, 100 mV/s, 21 °C). Ultrasound treatment (12 kJ) in H2SO4 removed trapped air in microstructures. After DI rinsing, the sample was immersed in a gold electrolyte (pH 9.35; stirred 120 rpm; 33 °C) and again ultrasonicated (15 kJ). Gold pillars were electrodeposited by chronoamperometry at −550 mV vs Ag/AgCl (current ~2 mA) for 24 h, yielding ~60 µm height (rate ~2.5 µm/h). The resist was stripped, and the seed layer was removed by RIE (Ar/CH4). Sidewall insulation and head modification: The substrate received ~4 µm parylene C (with Ar/O2 pretreatment and Silane A-174 adhesion promoter). A photoresist was applied; pillar heads were opened by flood exposure (dose tuned to exposed height; e.g., 85 mJ/cm² to expose ~10 µm on 60 µm pillars) and development. Parylene on exposed heads was etched by O2 RIE (e.g., 25 min at ~12 µm/h). The exposed Au surface of the head was roughened by a 2 min wet etch (TechniEtch AC12). For Type B, Ti (~50 nm)/TiN (~450 nm; total ~500 nm) was sputter-deposited and lifted off in acetone (1 h) to confine TiN to the heads. Residual parylene on planar PI regions was removed by O2 RIE (~18 min) to mitigate curling, leaving ~9 µm substrate thickness. For Type A, only wet-etch roughening was used; head modification height was set by RIE duration. Interconnect/bonding: The MEA was connected to a rigid PCB via a microflex interconnect (MFI) process. A ~90 µm gold ball formed at a capillary was pressed into a via/hole in the MEA pad onto a heated PCB pad under controlled force and ultrasonic energy, creating a microrivet stud that provides mechanical and electrical connection; the wire was then detached. Evaluation: In vitro electrochemical and noise assessments were performed; ex vivo validation used mouse retina preparations to assess spontaneous neuronal spike recording performance. Two electrode diameters (20 and 50 µm) and surface variants (Type A, Type B) were compared for impedance and noise characteristics.

- Fabricated flexible 3D MEAs with 32 protruding recording electrodes (20 or 50 µm diameter, ~60 µm height) on a thin (~9 µm) polyimide substrate, with parylene C sidewall insulation and selectively modified heads. - Type B electrodes (wet-etched heads with TiN coating) achieved low impedance: approximately 7 kΩ at 1 kHz for 50 µm diameter electrodes with ~5000 µm² exposed area, along with low intrinsic noise. - Ex vivo pilot in mouse retina recorded spontaneous neuronal spikes with amplitudes up to 66 µV, demonstrating functional recording capability and improved SNR attributable to reduced electrode-to-source distance and lowered impedance. - Process flow yielded high-aspect-ratio pillars with controllable exposed head height via flood exposure and RIE, enabling tunable electrochemical surface and coating application (TiN) localized to the head.

By combining a flexible polymer substrate with protruding, micrometer-scale electrodes and surface modifications (wet etch and TiN), the device addresses key barriers to recording from small autonomic nerves: it reduces the electrode-to-axon distance to improve SNR, and lowers electrode impedance without enlarging geometric size to preserve selectivity and allow higher spatial density. The low impedance (~7 kΩ at 1 kHz) and low noise achieved with Type B electrodes are consistent with design goals for small-signal detection from C-fiber-dominated nerves. The ex vivo retina recordings (spikes up to 66 µV) validate the concept in delicate neural tissue, suggesting feasibility for selective interfacing with autonomic nerves where cuff electrodes lack selectivity and rigid silicon arrays are too invasive or mismatched mechanically. The approach expands fabrication methods for flexible 3D MEAs and offers a platform for closed-loop bioelectronic medicine targeting pelvic neural circuits.

The work introduces a novel, flexible polyimide-based 3D microelectrode array featuring protruding gold pillars with parylene C sidewall insulation and selectively modified heads (wet etch and optional TiN). The process yields low-impedance, low-noise microelectrodes capable of recording small-amplitude neural activity, validated ex vivo in mouse retina with spikes up to 66 µV. This technology has strong potential for selective recording from small autonomic nerves and for advancing closed-loop bioelectronic medicine. Future work could include in vivo validation on pelvic autonomic nerves, chronic stability and biocompatibility assessments, optimization of pillar geometry and coating parameters for further impedance/noise reduction, and scaling of array density for higher spatial resolution.