An Adaptive and Autonomous System-On-Chip with Data Analysis for μs-Latency Closed-Loop Optogenetics

In this paper, we report a mixed-signal chip in 0.13-μm designed for long term autonomous operation that can trigger low-latency stimulation upon the detection of complex neural firing patterns. This adaptive autonomous neural IC (AANIC) includes: i) circuits for multi-site optogenetics stimulation and multi-unit electrophysiology recording, ii) specialized digital cores to detect, compress, and classify the neural signals in situ and iii) a programmable processor, connected to the specialized cores, running the closed-loop algorithm. The AANIC can process 10 neural channels in parallel, and stimulate through two independent precise current-sources, and is integrated within a tiny wireless optogenetic platform. The AANIC has a latency of 0.6 ms, a power of 165 μW/Ch, an effective number of bits (ENOB) of 8.68 bits (OSR=25), a tunable bandwidth (1-6.5 kHz), and is tested in vivo in freely-moving experiments involving a mouse model of temporal lobe epilepsy expressing ChR2 in its inhibitory neurons.

I. Introduction

OPTICAL brain stimulation synchronized with neural recording, known as closed-loop (CL) optogenetics, makes it possible to control the activity of a given neuronal network based on the neuronal activity of this specific or another network. This technology holds potential for advancing our understanding normal or pathological brain activity using rodent models, such as mice. Several custom integrated circuits (IC) have been developed and used to build lightweight CL electro-optic implants, but they have practical limitations: 1) they cannot detect and trigger stimulation on complex firing patterns [1], [2], [3], [4], and 2) they have processing latencies larger than biological neural circuits delays (which are typically smaller than 5 ms), which does not allow CL scenarios replicating the physiological conditions [1]. Complex FPGA-based alternatives have been developed [5], [6], but such systems cannot be easily implemented in wireless device because of size and power constrains, especially for rodent-based research. In this paper, we describe a system-on-a-chip designed for autonomous operation which can trigger stimulation with low μs-latency upon detection of complex neural firing patterns. Our adaptive autonomous neural IC (AANIC) includes circuits for multichannel recording & optogenetics stimulation, specialized digital cores to detect, compress, and classify the action potentials (AP), and a programmable 16-bit processor that runs a CL algorithm based on the on-chip AP-to-neuron sorting results. We show that AANIC can detect and silence seizures in vivo in a mouse model of temporal lobe epilepsy by CL activation of ChR2expressing inhibitory neurons.