Beta Oscillation Detector Design for Closed-Loop Deep Brain Stimulation of Parkinson’s Disease with Memristive Spiking Neural Networks

Abstract

Deep Brain Stimulation (DBS) is a prominent treatment of Parkinson’s disease (PD) that sends stimulation signals into the brain. The Closed-Loop DBS (CL-DBS) is an adaptive DBS system sending optimized and dynamic stimulation signals in accordance with the PD symptoms. The CL-DBS system is implanted and carried by the patients all the time. Thus, it requires advanced intelligence and energy efficiency to maintain a 24/7 real-time operation. However, the state-of-the-art CL-DBS systems are implemented with the traditional integrated circuits that cannot meet these demands. To tackle this challenge, in this paper, we design a novel energy-efficient beta oscillation detector of the CL-DBS system using Spiking Neural Networks (SNNs) and memristive synapses. The proposed SNN-based beta oscillation detector is trained with PD model data and evaluated using experimental data from the PD rats. The improvement of our SNN-based CL-DBS detector is evaluated with the architecture-level simulator NeuroSIM. The reductions of the proposed system on the chip area, latency, and energy are 67.3%, 41.9%, and 11.7% by using memristive synapses compared to the traditional SRAM (6T).