ERI: Memristor-based Neuromorphic Circuit Design for Closed-Loop Deep Brain Stimulation

Michigan Technological University, Houghton MI

Investigators

Abstract

Parkinson's disease (PD) is a brain disorder affecting millions of patients worldwide each year. PD patients experience motor symptoms including uncontrolled shaking or rigid muscles that may feel tight and difficult to move. Deep brain stimulation (DBS) is an effective therapy used to treat motor symptoms of PD. DBS is a surgical procedure that involves placing a signal generator in the chest to send electrical pulses to a specific region of the brain. These electrical pulses from the signal generator control motor symptoms by affecting the cells in the brain. The DBS system is life-changing for patients, but the continuous and rigid electrical pulses cause unwanted side effects, such as potentially blocking blood flow. Recently, scientists have developed a new DBS system called Closed-Loop DBS (CL-DBS) where the generator can send various electrical pulses back to the brain depending on PD symptoms to avoid side effects. One challenge for a CL-DBS system is it requires a powerful computer to generate the expected and various electrical pulses. However, these powerful computers cannot be placed in the chest of patients due to their large size. To solve this issue, this project aims to design a new type of computer called a neuromorphic chip. This chip mimics human brains by using small spiking signals for calculations to realize high energy efficiency. This new neuromorphic chip will make CL-DBS smarter, smaller, and lighter, greatly benefiting all PD patients. The project provides a unique opportunity for college and high school students from the Copper Country region in the Upper Peninsula of Michigan to participate in interdisciplinary research on neuroscience, Parkinson’s disease, brain rehabilitation, artificial intelligence, and microchip design. The project aims to design a neuromorphic CL-DBS chip consisting of electronic neurons and memristive synapses that adapts to user symptoms while substantially lowering power consumption and device size. The severity of PD symptoms is indicated by beta oscillations from the brain. Two Spiking Neural Networks (SNN) will be placed at the feedforward and feedback branches of the closed-loop system to recognize PD symptoms and generate stimulation signals accordingly. The neurons and synapses in these SNNs will be implemented using complementary metal-oxide semiconductor (CMOS) technology and memristors. The project will focus on two tasks: (1) memristive synapse design; and (2) electronic neuron design. Additionally, peripheral circuitries, including bandpass filters, low-noise amplifiers, and reading/writing circuits, will be developed. The chip will be taped out for further evaluation by comparing it with other CL-DBS systems using different computational hardware, including Graphics Processing Units (GPUs) and Field Programmable Gate Arrays (FPGAs), based on 1) power consumption; 2) response time; 3) chip area. This research on neuromorphic medical chips will significantly benefit the development of energy-efficient and smart implantable medical devices by reducing their size, weight, and energy budget while making them more intelligent and adaptive. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.