University of Cambridge > Talks.cam > Computational Neuroscience > A brain-machine interface for locomotion driven by subspace dynamics

  Browse A-Z

Log in

Cambridge users (raven)details Other usersdetails No account?details

Information on

Subscribing to talksdetails Finding a talkdetails Adding a talkdetails Disseminating talksdetails Help and Documentationdetails

A brain-machine interface for locomotion driven by subspace dynamics

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Marcelo Gomes Mattar.

Brain-machine interfaces (BMIs) for reaching have enjoyed continued performance improvements, allowing remarkable 2D cursor control. Yet there remains significant clinical need for locomotor (e.g., wheelchair control) BMIs, which could benefit a larger patient population. Fewer studies have addressed this need, and the best strategy for doing so remains undetermined. Here we demonstrate an approach based upon rhythmic neural activity. We leverage a behavioral task wherein monkeys cycle a hand-held pedal, forward or backward, to advance along a virtual track, pausing on targets for reward. This task does not emulate natural locomotion, but rather provides a view of cortical activity during learned, voluntary, rhythmic movement. Such activity is robust and was recently characterized (Russo et al. 2018), affording opportunities to develop novel decode algorithms and test them in an online setting. We constructed a decoder that decoded virtual self-motion, based on recordings from 192 electrodes implanted in motor cortex. Unlike algorithms for cursor control, we did not directly map neural states to commanded velocity or position. Instead, we leveraged the most robust aspects of response structure: an overall shift in neural state when moving versus stationary, and rotations of the neural state while cycling. We used these features to decode when the subject was moving, and decoded direction based on the finding that neural-state rotations occur in different planes during forward versus backward cycling. Perhaps because the subject need not learn a novel mapping to control the BMI , performance was high even during the first few sessions of brain control. An additional performance gain was obtained by leveraging a neural dimension that reflected cycling direction at movement initiation. Resulting brain-control success rates were very close to those achieved under arm-control. Thus, rhythmic neural activity provides a robust substrate for BMI control, but requires different decode strategies than have been employed previously.

This talk is part of the Computational Neuroscience series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

Other lists

Meeting the Challenge of Healthy Ageing in the 21st Century Chemistry Quantitative History Seminar

Other talks

Are Your Selfies Carbon-Neutral? Human Rights and the Environmental Impact of Digital Technology Neural mechanisms of model-based planning in the rat Searching for the fastest stars in Gaia DR2 The story of a theorem Clueless Voting
 

© 2006-2024 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity