Interfaces neuronales CMOS haute résolution pour l'électrophysiologie et l'optogénétique en boucle fermée

Authors: Gagnon-Turcotte, Gabriel
Advisor: Gosselin, Benoit
Abstract: The future of brain research lies in the development of new technologies that will help understand how this complex organ processes, integrates and transfers information. Among these, optogenetics is a recent technology that allows the use of light to selectively activate neurons in the cortex of transgenic animals to observe their effect in a large biological network. This experimental setting is typically based on observing the neuronal activity of transgenic mice, as they express a wide variety of genes and diseases, while being inexpensive. However, most available neural recording or optogenetic devices are not suitable, because they are hard-wired, too heavy and/or too simplistic. Unfortunately, few wireless systems exist, and they are greatly limited by the required bandwidth to transmit neural data, while not providing simultaneous multi-channel neural recording and optogenetic, a must for stimulating and observing several areas of the brain. In current devices, the analysis of the neuronal data is performed ex situ, while the research would greatly benefit from wireless systems that are smart enough to interpret and stimulate the neurons in closed-loop, in situ. The goal of this project is to design analog-digital circuits for acquisition and processing of neural signals, algorithms for analysis and processing of these signals and miniature electrooptical wireless systems for: i) Conducting experiments combining high-resolution multi-channel neuronal recording and high-resolution multi-channel optogenetics with freely-moving animals. ii) Conduct optogenetic experiments synchronized with the neural recording, i.e. in closed loop, with freely-moving animals. iii) Increase the resolution while reducing the size, weight and energy consumption of the wireless optogenetic systems to minimize the impact of research with small animals. This project is in 3 phases, and its main contributions have been reported in ten conferences (ISSCC, ISCAS, EMBC, etc.) and four published journal papers, or submitted, as well as in a patent and two disclosures. The design of a high resolution optogenetic system poses several challenges. In particular, since the neuronal signals have a high frequency content (10 kHz), the number of chanv nels under observation is limited by the bandwidth of the wireless transmitters (2-4 channels in general). Thus, the first phase of the project focused on the development of neural signal compression algorithms and their integration into a high-resolution miniature and lightweight wireless optogenetics system (2.8g), having 32 recording channels and 32 optical stimulation channels. This system detects, compresses and transmits the waveforms of the signals produced by the neurons, i.e. action potentials (AP), in real time, via an embedded low-power field programmable gate array (FPGA). This processor implements an AP detector algorithm based on adaptive thresholding, which allows to compress the signals by transmitting only the detected waveforms. Each AP is further compressed by a Symmlet-2 discrete wavelet transform (DWT) followed dynamic discrimination and requantification of the DWT coefficients, making it possible to achieve high compression ratios with a good reconstruction quality. Results demonstrate that this algorithm is more robust than existing approach, while allowing to reconstruct the compressed signals with better quality (average SNDR of 25 dB 5% for a compression ratio (CR) of 4.2). With detection, CRs greater than 500 are reported during the in vivo validation. The use of commercial components in wireless optogenetic systems increases the size and power consumption, while not being optimized for this application. The second phase of the project consisted in designing a complementary metal oxide semiconductor (CMOS) system-on-chip (SoC) for neural recording and multi-channel optogenetics, which significantly reduces the size and energy consumption compared to commercial alternatives. This is important contribution, since it’s the first chip to integrate both features. This SoC has 10 recording channels and 4 optogenetic stimulation channels. The bioamplifier design includes a programmable bandwidth (0.5 Hz -7 kHz) and a low input-referred noise (IRN of 3.2 μVrms), which allows targeting different biological signals (AP, LFP, etc.). The Delta-Sigma (DS) MASH 1-1-1 low-power analog-to-digital converter (ADC) is designed to work with low OSR (50), as to reduce its power consumption, and has a programmable resolution (ENOB of 9.75 bits with an OSR of 25). This ADC uses a new technique to reduce its circuit size by subtracting the output of each DS branch in the digital domain, rather than in the analog domain, as done conventionally. A recording channel, including the bioamplifier, the DS and the decimation filter, consumes 11.2 μW. Optical stimulation is performed with an on-chip LED driver using a regulated cascode current source with feedback, which accommodates a wide range of LED parameters and battery voltages. The SoC is integrated into a wireless optogenetic platform and validated in vivo.

To date and excluding this project, no wireless system is making closed-loop optogenetics simultaneously to real-time monitoring of neuronal activity. An important contribution of this work is to have developed the first multi-channel optogenetic system that is able to work in closed-loop, and the first to be validated during in vivo experiments involving freely-moving animals. To do so, the third phase of the project aimed to design a digital CMOS chip, called neural decoder integrated circuit (ND-IC). The ND-IC and the SoC developed in Phase 2 are integrated within a wireless optogenetic system. The ND-IC has 3 main cores: 1) the adaptive AP detector core, 2) the compression core with a new sorting tree for discriminating the DWT coefficients, and 3 ) the AP automatic classification core that reuses the data generated by the detection and compression cores to reduce its complexity. A link between a recording channel and a stimulation channel is established according to the association of each AP with a neuron, thanks to the classification, and according to the bursting activity of this neuron. The ND-IC consumes 56.9 μW and occupies 0.08 mm2 per channel. The system weighs 1.05 g, occupies a volume of 1.12 cm3, has an autonomy of 3h, and is validated in vivo.
Document Type: Thèse de doctorat
Issue Date: 2019
Open Access Date: 16 September 2019
Permalink: http://hdl.handle.net/20.500.11794/36493
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
Description SizeFormat 
35323.pdf19.34 MBAdobe PDFThumbnail
View/Open
All documents in CorpusUL are protected by Copyright Act of Canada.