Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Robustness of neuroprosthetic decoding algorithms.

Mijail Serruya1, Nicholas Hatsopoulos, Matthew Fellows

  • 1Department of Neuroscience, Brown University, Providence, RI 02912, USA. Mijail_Serruya@brown.edu

Biological Cybernetics
|March 21, 2003
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Sparse component analysis: A method that uncovers separable computations within neural population activity.

Neuron·2026
Same author

Beast3D: Animal behavioral analysis and neural encoding from multi-view video via Gaussian splatting.

ArXiv·2026
Same author

Lightning Pose 3D: an uncertainty-aware framework for data-efficient multi-view animal pose estimation.

bioRxiv : the preprint server for biology·2026
Same author

A Spatially Structured Spiking Network Model of Beta Traveling Waves and Their Attenuation in Motor Cortex.

bioRxiv : the preprint server for biology·2026
Same author

Computational optimization of two-photon holographic stimulation sites<i>in vivo</i>.

Journal of neural engineering·2026
Same author

Animal behavioral analysis and neural encoding with transformer-based self-supervised pretraining.

ArXiv·2026

Predicting hand movements from brain signals requires minimal data. Researchers found that even a few minutes of neural activity from a small number of neurons can build effective decoding models for brain-computer interfaces.

Area of Science:

  • Neuroscience
  • Biomedical Engineering
  • Machine Learning

Background:

  • Predicting movement from neural activity is crucial for developing advanced brain-computer interfaces (BCIs).
  • The amount of neural data required to train effective decoding models remains a key challenge for BCI development.

Purpose of the Study:

  • To assess how the quantity of neural data influences the accuracy of algorithms predicting hand kinematics.
  • To determine the minimum neural data needed to build robust predictive models for motor control.

Main Methods:

  • Recorded single- and multineuron activity from macaque motor cortex during trained tracking tasks.
  • Evaluated maximum-likelihood and linear filter models for predicting discrete and continuous hand movements.
  • Systematically varied the amount of training data and time intervals between model training and testing.

Related Experiment Videos

Main Results:

  • Optimal models for predicting discrete movements required less than 1 minute of data (8-13 neurons).
  • Optimal models for continuous movements required less than 3 minutes of data (8-18 neurons).
  • Model performance showed no significant degradation when tested across different days, indicating robustness.

Conclusions:

  • Robust kinematic prediction models can be rapidly constructed using minimal neural data from a limited number of neurons.
  • These findings support the feasibility of using motor cortex signals for reliable neural prosthetic devices with daily calibration.
  • Efficient decoding models can be built quickly and maintained, paving the way for practical BCI applications.