Stabilization of a brain-computer interface via the alignment of low-dimensional spaces of neural activity published in Nature Biomedical Engineering

Our new paper “Stabilization of a brain–computer interface via the alignment of low-dimensional spaces of neural activity” was published today in Nature Biomedical Engineering. In it, we present a stabilized brain-computer interface (BCI) system that addresses the problem of neural recording instabilities. These instabilities, if left unchecked, can quickly lead BCIs to become uncontrollable.

To address this problem, we developed a manifold-based stabilizer that extracts a stable representation of neural activity from neural recordings where instabilities may be present. To do this, our technique leverages the concept of a “neural manifold”: a low-dimensional space that captures specific correlation patterns across neurons. By aligning estimates of the neural manifold made at different points in time, we can stabilize the inputs to the BCI system, thus allowing control to proceed unimpeded. We hope that this work will help to improve the clinical viability of BCIs.

This work was the result of a collaboration between researchers at Carnegie Mellon University and the University of Pittsburgh, including myself, William Bishop, Emily Oby, Elizabeth Tyler-Kabara, MD, PhD, Steven Chase, Aaron Batista, and Byron Yu.