OPEN Wireless, miniaturized, semi-implantable electrocorticography microsystem validated *in vivo*

The study addresses the need for higher quality, minimally invasive neural recording systems suited for neuroscience research, diagnostics, and neuroprosthetics. Extracellular neural recording is commonly performed via EEG, ECoG, or intracortical recording. EEG is non-invasive but offers low spatial resolution. Intracortical approaches provide high temporal and spatial resolution but are invasive. Electrocorticography (ECoG), recording from the cortical surface, balances invasiveness and signal quality, offering higher SNR and spatiotemporal resolution than EEG and simpler implantation than intracortical methods. The paper’s purpose is to design and validate a wireless, ultra-compact, semi-implantable µECoG recording system capable of multi-channel acquisition and telemetry, suitable for subdural or epidural placement and validated in vivo. The work aims to provide a stand-alone device that can record spontaneous and evoked cortical activity in rodents and to compare its capabilities with existing compact and integrated ECoG systems.

ECoG systems have been developed in three general ways: (1) distributed systems where electrode arrays connect to benchtop lab equipment; (2) compact or ultra-compact devices assembled from off-the-shelf components used as wearable or semi-implantable recorders; and (3) integrated microsystems where much of the electronics is on a single chip. Based on electrode size and coverage, ECoG falls into macro-ECoG (mm–cm electrodes, wide coverage) and micro-ECoG (sub-mm electrodes for high spatial resolution). Prior works include early self-sufficient and fully wireless intracortical implants, and ECoG systems with various channel counts, sampling rates, and communication methods. Recent ultra-compact ECoG devices have demonstrated feasibility but are often hardwired for power/data or larger in volume. This study positions its system as an ultra-compact ECoG device with fully wireless RF telemetry, small form factor, and suitable packaging for semi-implantable use, comparing favorably in size and autonomy to prior art.

System architecture: The system comprises a semi-implantable recording unit and an external unit connected via a 2.4 GHz RF telemetry link using a Nordic nRF24L01 transceiver (up to 2 Mbps, ~50 m range). The recording unit is housed in a 3D-printed PLA cubic package (2.9 × 2.9 × 2.5 cm; total weight for Version-1, including batteries and rigid MEA: 27 g). Power is supplied by two 3.7 V, 50 mAh Li-ion batteries, enabling about 1 h of continuous operation. Electronics are assembled on a vertically stacked PCB platform (four 2.5 × 2.5 cm boards and one 1.1 × 1.8 cm board). Signal chain: cortical signals captured by the microelectrode array are preconditioned with low-power, low-noise instrumentation amplifiers (Analog Devices AD8222, input-referred noise 8 nV/√Hz at 1 kHz; programmable gain 1–10,000) including DC level shifting and band-pass filtering. Signals are digitized at 1 ksps per channel with 10-bit resolution (Texas Instruments ADS1298). A microcontroller (Atmel ATmega32, 16 MHz) handles data framing and central control, and the RF transceiver handles GFSK modulation with CRC. External unit: the receiver decodes frames and streams data via UART over USB to a host computer. Custom software performs decoding, demultiplexing, real-time visualization, and local storage for offline analysis. Microelectrode arrays: (1) Rigid MEA (Version-1): a 2 × 4 array of gold electrodes on a 3D-printed PLA substrate (6 × 4 × 2 mm). Holes of 350 µm diameter are micro-drilled at 1.5 mm pitch; 300 µm gold wires are inserted and formed into rivet heads accessible on both sides. (2) Flexible MEA (Version-2): a 12 × 6 array of 100 µm-diameter titanium electrodes microfabricated on a 25 µm polyimide substrate. Of the 72 sites, 36 are routed to bonding pads. Interconnects and pads are defined in evaporated titanium; the device is passivated with SU-8, with openings at recording sites and pads. Electrode pitch: vertical 400 µm, horizontal 470 µm, covering 2.55 × 2.4 mm. Packaging and integration: Two package versions were fabricated in PLA. Version-1 integrates the PCB stack and batteries with a concentric smaller cube housing the rigid MEA at the bottom so electrodes contact the cortex through a burr hole; electrode backs are connected via a connector and silver paste to the electronics. Version-2 houses the PCB and batteries with a slot allowing the flexible MEA tail to exit and be bent onto the cortical surface; screw holes allow skull mounting using orthopedic screws; caps secure the top. PCB platform: stacked boards ordered along the signal path: preamplification, A/D conversion, microcontroller-based digital processing, and wireless transmission, plus separate boards for batteries and regulators (3.3 V and 5 V). Digital signal processing and data reduction (MBED): To reduce spatial redundancy among neighboring ECoG channels, a multichannel baseline (common component) is computed as the average across N channels, and channel-specific difference signals are formed by subtracting this baseline from each channel. Since differences have smaller amplitude, they require fewer bits (m) than the original M-bit samples; the baseline is transmitted at M bits. The bit-rate reduction (BRR) equals [(N×M) − (M + N×m)]/(N×M) × 100%. For typical values M = 10 and m = 5 with m/M ≈ 0.5, BRR ≈ 37.5% without loss of information. Temporal compression can be layered atop this spatial reduction. In vitro validation: The Version-1 unit was tested in saline with a prerecorded ECoG signal applied; amplified and filtered outputs were verified on an oscilloscope. Impedance spectroscopy in saline showed rigid electrodes ≈0.9 kΩ at 1 kHz and flexible electrodes ≈6 kΩ at 1 kHz. In vivo experimental protocol: Acute experiments were performed in anesthetized adult male Wistar rats under approved institutional and NIH guidelines. Animals were anesthetized with ketamine (80 mg/kg) and xylazine (12 mg/kg, i.p.), head-fixed in a stereotaxic frame, and body temperature maintained. A 5 × 7 mm skull window over right parietal somatosensory cortex (0.6–7.6 mm posterior to bregma; 1–6 mm lateral) was prepared. MEAs were placed subdurally or epidurally as appropriate. Recordings were performed inside a Faraday cage. For system validation, the rigid MEA version was used to record spontaneous activity from contralateral somatosensory cortex with gain 1000 in 10 s sessions; evoked responses were recorded to electrical stimulation of the hind paw. Flexible and rigid MEAs were also validated independently by connecting them to commercial lab equipment for in vivo signal acquisition.

• The semi-implantable µECoG system functioned as a fully wireless, stand-alone recording unit with 2.4 GHz RF telemetry (up to 2 Mbps), 1 ksps per channel sampling, and 10-bit resolution. Continuous operation time was about 1 hour on two 3.7 V, 50 mAh Li-ion batteries. Package size was 2.9 × 2.9 × 2.5 cm; Version-1 total system weight was 27 g. - Two MEA versions were realized: a rigid 2 × 4 gold-electrode PLA-based array and a flexible 12 × 6 titanium polyimide array with 36 routed channels (100 µm electrodes; 400–470 µm pitch). - In vitro impedance spectroscopy in saline showed electrode impedances of approximately 0.9 kΩ (rigid) and 6 kΩ (flexible) at 1 kHz. - In vivo, both MEAs recorded cortical ECoG signals. The integrated wireless system (rigid MEA version) successfully captured spontaneous somatosensory cortical activity in anesthetized rats; channels 7 and 8 showed poor quality due to lack of adequate electrode-cortex contact. Correlograms indicated small inter-channel correlations, suggesting low redundancy. - Evoked activity to hind paw electrical stimulation was recorded; time-domain traces showed stimulus-locked responses, spectrograms and RMS colormaps indicated changes at stimulus onset, and PSDs differed between spontaneous and evoked conditions across channels. - Applying the MBED lossless spatial redundancy reduction method can reduce bit rate by about 37.5% under typical conditions (M=10, m≈5), enabling bandwidth savings without loss of information. - Comparative specifications (from Table 1): this work achieved 1 kHz sampling, 8/36/72 channels, 10-bit resolution, about 209 mW power consumption, RF link at 2 Mbps, and a 2.9 × 2.9 × 2.5 cm device size, comparing favorably in autonomy and size to other ultra-compact ECoG systems.

The system addresses the need for minimally invasive yet high-quality cortical recordings by combining micro-scale electrode arrays with an ultra-compact, battery-powered, fully wireless recording unit. In vivo results demonstrate the capability to capture both spontaneous and evoked somatosensory cortical activity in anesthetized rats, validating the electrode designs, analog front-end, digitization, RF telemetry, and software pipeline. Low inter-channel correlations in spontaneous recordings support the premise of limited redundancy among channels and motivate the MBED method for lossless spatial data reduction, which can conserve telemetry bandwidth and power. Compared with state-of-the-art ultra-compact ECoG systems, the presented device offers fully wireless operation (data and power untethered), markedly smaller physical volume (~21 cm³ vs. ~180 cm³ in a cited recent device), and a package and MEA integration that facilitate mounting on the head for potential freely moving experiments. The combination of flexible MEA technology and compact packaging provides versatility for subdural or epidural placements with fine spatial resolution, while the rigid MEA version enables simplified implantation through a burr hole.

Two versions of an ultra-compact, semi-implantable µECoG system were designed, fabricated, integrated, and validated in vivo. Version-1 uses an 8-channel rigid PLA-based MEA interfaced through a burr hole; Version-2 employs a flexible 72-site (36 routed) titanium-on-polyimide MEA placed on the cortical surface via a small craniotomy. The wireless recording unit, packaged in a small PLA housing with stacked PCBs and batteries, reliably recorded spontaneous and evoked cortical activities in anesthetized rats, confirming end-to-end functionality. Relative to prior ultra-compact ECoG devices, this system is smaller, fully wireless, and designed for head-mounted use. Future work could extend battery life and power management, increase channel counts and effective data throughput (leveraging MBED and additional temporal compression), validate long-term biocompatibility and chronic implantation, integrate stimulation capability, and demonstrate performance in freely moving animals and larger models.

• In vivo validation was performed acutely in anesthetized rats; chronic performance, long-term stability, and biocompatibility of the semi-implantable package and electrodes were not reported. - Full integrated system testing focused on the rigid MEA version; while the flexible MEA was validated in vivo using commercial equipment, the complete wireless system operation with the flexible MEA was not detailed. - Battery life is approximately 1 hour, which may limit longer recording sessions without recharging or battery replacement. - Some channels exhibited poor signal quality due to suboptimal electrode-cortex contact, indicating sensitivity to implantation contact quality. - The telemetry link’s 2 Mbps data rate was only partially utilized; end-to-end throughput under various compression settings and potential data loss rates were not quantified in detail. - Experiments were conducted in a Faraday cage; performance in less shielded, ambulatory, or freely moving conditions remains to be demonstrated.