Internal Model Estimation for Closed-Loop Brain-Computer Interfaces

The motor system successfully plans and executes sophisticated movements despite sensory feedback delays and effector dynamics that change over time.  Behavioral studies suggest that internal models are central to motor control, but neural correlates thereof have thus far been limited.  In the skeletomotor system, this problem isparticularly challenging due to the large number of neurons involved across multiple brain areas, non-linear limb dynamics, and multiple sensory feedback modalities.  In this talk, I will show how brain-computer interfaces (BCI), developed primarily to assist disabled patients, can be leveraged for basic scientific studies of motor control.  We consider an intracortical cursor-based BCI, which can be viewed as a simplified motor control system.  We found evidence that the subject uses an internal model during closed-loop BCI control and studied the timecourse of internal model adaptation during BCI learning.  We also developed a novel statistical algorithm that canextract an internal model from neural population activity recorded during BCI control.  This work suggests that closed-loop BCI experiments, combined with novel statistical analyses, can provide insight into the neural substrates of feedback motor control and motor learning.