UC Irvine

The prevalence of neurodegenerative diseases and injuries that result in motor impairment, such as stroke and spinal cord injury, have motivated researchers to explore brain-computer interface (BCI) technologies as a solution to treat damaged neural functions. More specifically, these BCIs promise to either rehabilitate, supplement or bypass affected areas by allowing patients to power motor prostheses or other assistive devices using neural signals. The most recent BCI research has increasingly focused on invasive neural interfacing modalities such as electrocorticography and microelectrode arrays, which confer advantages such as higher spatial resolution and larger frequency bandwidth. They also provide the capability of delivering cortical electrostimulation to elicit artificial sensory feedback, which is critical to motor functions. A BCI implementing closed-loop electrostimulation sensory feedback control is referred to as a “bi-directional” brain-computer interface (BD-BCI). These BD-BCIs improve on uni-directional designs in that they restore function in a more biomimetic fashion, and feedback has the potential to improve BCI performance. Invasive BD-BCIs also possess the potential to be realized as fully implantable devices, which greatly improves the practicality of these devices for chronic use.

A number of design obstacles, however, must be resolved before a fully implantable BD-BCI can be realized. One such problem is that the simultaneous stimulation and recording necessary for motor BD-BCI causes strong electrical artifacts to propagate to the recording site. These artifacts can potentially obscure neural features or saturate analog front-ends. To begin to address these problems, our work first characterizes the propagation of stimulation artifacts using data collected during clinical cortical mapping procedures. We then detail how these artifact characteristics can be exploited by artifact suppression methodologies. Additionally, we also propose dipole cancellation as an analog front-end artifact suppression method designed to safeguard against amplifier saturation. We also propose a digital back-end method that utilizes pre-whitening and null projection to efficiently suppress residual artifacts persisting through front-end artifact suppression methods. Finally, we show a demonstration of an embedded BD-BCI exoskeleton system that abides by fully-implantable constraints. This device allowed a human subject implanted with ECoG electrodes to wirelessly control a robotic leg exoskeleton, as well as delivered electrostimulation-induced leg sensation during leg swing.