Contextualization In the realm of bioelectronics and pervasive computing, efficient processing and real-time interpretation of biological signals is critical. Traditional biosignal sensing methods, which rely on high sampling rates, produce redundant data and consume considerable energy. Addressing these challenges, this paper introduces an innovative self-clocked microelectrode array (MEA) that digitizes biosignals at the pixel level. By encoding significant changes as asynchronous digital address-events, the MEA drastically reduces off-chip data transmission, providing a more efficient solution for biosignal processing.

GAIA MEA System In this paper, we present the Global Asynchronous Intelligent Array (GAIA) system. It features a 64×64 electrode array, an asynchronous 2D-arbiter, and an Address-Event Representation (AER) communication block. Each pixel in this array operates autonomously, monitoring voltage fluctuations from cellular activity and producing a single digital pulse in response to any significant change. These changes are encoded as "up" (positive) and "down" (negative) events, which are then routed off-chip via the asynchronous arbiter.

Architecture and Functionality GAIA’s data transmission approach outputs asynchronous digital events only when detecting a local relative voltage change that surpasses a preset threshold. This method prioritizes the encoding of meaningful biological signals with large transients and discards noise and small fluctuations, thus significantly reducing the overall data output. The system's central 64×64 pixel core, flanked by X and Y address encoders and an AER communication block, forms the backbone of the GAIA system. Each electrode measures 15×15 µm with a pitch of 48 µm.

Signal Processing and Event Generation Within each pixel, signals are first amplified by adjustable gain stages designed to enhance signals in the 1Hz-10kHz range while rejecting large DC components. The amplified signals are monitored for significant changes, which, when detected, trigger the event generation block to emit digital events corresponding to the signal variations. These events are managed by a 2D-arbiter system that encodes the event’s location and polarity, facilitating efficient and collision-free data transmission. GAIA can handle up to 20 million events per second, ensuring a robust and scalable system.

Electrical Characterization The GAIA system's electrical properties were rigorously characterized, demonstrating several key attributes. The system offers programmable gain settings, with in-band amplifications ranging from 37.4 dB to 57 dB. Noise levels were also assessed, showing that the integrated input-referred noise within the 500 Hz - 3 kHz band is 19.04 µV, while across the full 5 Hz - 10 kHz bandwidth, it is 71.05 µV. The power consumption of each pixel is approximately 842.4 nW, with the total chip power consumption, including the 2D arbiter, amounting to 3.58 mW. Latency tests revealed that the system has an average latency of 0.9995 µs for both positive and negative events.

Experimental Validation GAIA was tested with electrogenic cells, demonstrating its ability to detect bioelectric signals effectively. Validation was carried out with a beating cardiomyocyte culture, a model known for generating periodic signals at approximately 1Hz. The cells were cultured on the GAIA sensor, and recordings began on the seventh day after plating, by which time the cells had aggregated and synchronized their beating patterns. The system consistently encoded and transmitted the bioelectric signals, confirming its efficacy.

Spiking Neural Network Interfacing GAIA’s event-based outputs were successfully interfaced with an event-based mixed-signal neuromorphic processor, specifically the DYNAP-SE processor. This integration demonstrated a prototype for real-time bio-signal sensing and processing, leveraging the strengths of both hardware components. In this setup, pixel and neuron computations occur in the analog domain, while spike routing is managed digitally. This combined approach paves the way for advanced applications in edge computing and bio-sensing, highlighting the potential of integrating event-based biosensing with neuromorphic processing.

Conclusions

The GAIA system represents a significant advancement in bio-signal sensing and processing. By encoding signals at the pixel level and transmitting only significant events, GAIA effectively reduces data redundancy and energy consumption. The integration with a neuromorphic processor showcases the potential for efficient, real-time bio-signal processing.

The original article can be found here:

Cartiglia, M. et al. A 4096 Channel Event-Based Multielectrode Array with Asynchronous Outputs Compatible with Neuromorphic Processors. Nat Commun 14, 1344 (2024).