Designing Current Sensing Systems for Multiplexed Neural Interfaces

Abstract

Neural recordings facilitate neurological treatments by enhancing our understanding of the brain’s functions. While significant progress has been made with devices that record from only a handful of locations and neurons, increases in density and coverage reveal that these low-resolution devices overlook a wealth of information. As both electrode density and coverage increase, flexible electronics must be incorporated at each electrode to mitigate wiring constraints. However, present methods introduce significant noise and decrease device lifespan (1, 2). Both are crucial for real-time systems that require high signal to noise ratios (SNR), such as closed-loop recording and stimulation devices, or in-unit seizure monitoring (3, 4). Moving forward, electrode arrays need to have high resolution, large spatial coverage, long lifespans, and SNR. Current sensing technology leverages inherent advantages of unique circuit design to decrease electrical noise, increase electrode density, and increase device lifespan (5). The purpose of this dissertation is to enhance the performance of neural recordings by 1) developing a multiplexed current sensing system to record neural signals and 2) refining this multiplexed current sensing system to record high frequency neural activity and demonstrating device scalability. Highly multiplexed voltage sensing devices have high resolution and large spatial coverage but must choose between lifespan or high accuracy. The voltage bias needed for multiplexed voltage sensing devices greatly decreases device lifespan and can only be mitigated with glass encapsulation, which compromises device flexibility and recording quality. Voltage sensing devices also suffer from aliased (high frequency) noise which is increased by multiplexing at high rates. In the first chapter of this dissertation, I developed a current sensing headstage and electrode, which utilize circuit architectures that can be multiplexed without a bias voltage while also decreasing aliased noise. The custom headstage and electrode had sufficient and uniform electrical performance to record neural signals. I recorded distinguishable single trial neural responses in vivo and showed an increase in neural decoding performance compared to our passive control electrodes. This work demonstrated that multiplexed current sensing solutions have sufficient SNR and sampling rates to record LFP at low multiplexing rates in rodent cortex. Previously, current sensing research focused on local field potentials, but including high frequency neural activity such as high gamma or spiking activity can increase performance of neural decoding substantially (6–8). Increasing electrode sampling rates to record any high frequency signals poses a unique problem with multiplexed recordings. Even slight increases to sampling rates of individual electrodes result in significant system strain relative to the multiplexing rate. In the second chapter of this dissertation, I drastically increased the sampling rate of our current sensing system to record high frequency neural activity and increase device scalability. The first-generation current sensing headstage sampled at 3.2 kS/s, allowing for local field potential neural recordings at low multiplexing rates, but was re-designed to support 37 kS/s. This work demonstrated that highly multiplexed current sensing solutions have sufficient SNR and sampling rates to record high frequency neural activity. This work also showed consistent SNR across multiplexing rates indicating the scalability of multiplexed current sensing solutions. This dissertation shows multiplexed current sensing can record decodable neural signals, enable flexible, active electrodes, and operate at sufficiently high sampling rates to record high frequency neural activity at high multiplexing rates. Current sensing provides an alternative route for neural recordings with its own strengths and weaknesses. However, future work is required to demonstrate in vivo viability before widespread implementation of current sensing in highly multiplexed neural recordings is possible.