Fully Passive Flexible Wireless Neural Recorder for the Acquisition of Neuropotentials from a Rat Model

Abstract

Wireless implantable neural interfaces can record high-resolution neuropotentials without constraining patient movement. Existing wireless systems often require intracranial wires to connect implanted electrodes to an external head stage or/and deploy an application-specific integrated circuit (ASIC), which is battery-powered or externally power-transferred, raising safety concerns such as infection, electronics failure, or heat-induced tissue damage. This work presents a biocompatible, flexible, implantable neural recorder capable of wireless acquisition of neuropotentials without wires, batteries, energy harvesting units, or active electronics. The recorder, fabricated on a thin polyimide substrate, features a small footprint of 9 mm × 8 mm × 0.3 mm and is composed of passive electronic components. The absence of active electronics on the device leads to near zero power consumption, inherently avoiding the catastrophic failure of active electronics. We performed both in vitro validation in a tissue-simulating phantom and in vivo validation in an epileptic rat. The fully passive wireless recorder was implanted under rat scalp to measure neuropotentials from its contact electrodes. The implanted wireless recorder demonstrated its capability to capture low voltage neuropotentials, including somatosensory evoked potentials (SSEPs), and interictal epileptiform discharges (IEDs). Wirelessly recorded SSEP and IED signals were directly compared to those from wired electrodes to demonstrate the efficacy of the wireless data. In addition, a convoluted neural network-based machine learning algorithm successfully achieved IED signal recognition accuracy as high as 100 and 91% in wired and wireless IED data, respectively. These results strongly support the fully passive wireless neural recorder's capability to measure neuropotentials as low as tens of microvolts. With further improvement, the recorder system presented in this work may find wide applications in future brain machine interface systems.