An Energy Efficient non-volatile FPGA Digital Processor for Brain Neuromodulation

Abstract

Brain stimulation technologies have the potential to provide considerable clinical benefits for people with a range of neurological disorders. Recent neuroscience studies have shown that considerable information of brain states is contained in the low frequency local field potential (If-LFP; below 5Hz) recordings with application in real-time closed-loop neurostimulation for treating neurological disorders. Given these signals can be sampled at low sampling rate and hence provide sparse data streams, there is an opportunity to design implantable neuroprosthesis with long battery lifecycles which enables enough processing power to implement long-term, real-time closed loop control algorithms. In this thesis, a closed-loop embedded digital processor has been created for use in rodent neuroscience experiments. The first contribution of this work is to develop a mathematical analytical design approach of feedback controller for suppressing high-amplitude epileptic activity in the neuron mass model to form a better understanding of how to perform a better closed-loop stimulation to control seizures. The second contribution and the third contribution are combined to present an exploratory energy-efficient digital processor architecture built with commercial off-the-shelf non-volatile FPGAs and microcontroller for sparse data processing of brain neuromodulation. A digital hardware design of an exemplar PID control algorithm has been implemented on this proposed digital architecture. A new power computing diagram of this time-driven approach significantly reduced the power consumption which suggests that a digital combined control system of non-volatile FPGAs and microcontroller outweighs a digital control system of microcontroller with microcontroller regarding computing time cost and energy consumption supposing one microcontroller is always required. Taken together, this digital energy-efficient processor architecture gives important insights and viewpoints for the further advancements of neuroprosthesis for brain neurostimulation to achieve lower power consumption for sparse sampling data rate.