Related Experiment Video
Updated: Jul 27, 2025
07:37
Assessment and Communication for People with Disorders of Consciousness
Published on: August 1, 2017
9.1K
The impact of task context on predicting finger movements in a brain-machine interface
Matthew J Mender1, Samuel R Nason-Tomaszewski1, Hisham Temmar1
1Department of Biomedical Engineering, University of Michigan, Ann Arbor, United States.
Elife
|June 7, 2023
Summary
Brain-machine interfaces (BMIs) show promise for restoring hand function but struggle with task variations. Neural population activity adapts, enabling robust online control despite context changes, suggesting improved BMI robustness.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Rehabilitation Technology
Background:
- Clinical translation of brain-machine interfaces (BMIs) for hand motor function restoration hinges on robustness to task variations.
- Functional electrical stimulation (FES) involves diverse force production within similar movements, posing challenges for decoder generalization.
Purpose of the Study:
- To investigate the impact of task context changes on BMI performance and neural decoding.
- To assess the adaptability of neural population activity to altered physical hand configurations during BMI control.
Main Methods:
- Two rhesus macaques controlled a virtual hand with their physical hand, while task contexts were modified (e.g., adding springs, altering wrist posture).
- Simultaneous recording of intracortical neural activity, finger positions, and electromyography (EMG) was performed.
- Decoder generalization and online BMI control performance were evaluated across different contexts.
Main Results:
- Decoders trained in one context generalized poorly to others, increasing prediction errors, particularly for muscle activations.
- Online BMI control performance remained largely unaffected by changes in decoder training context or physical hand context.
- Neural population activity structure remained similar across contexts, facilitating rapid online adaptation.
- Neural activity shifted trajectories proportionally to required muscle activation, explaining kinematic prediction biases and suggesting a feature for predicting muscle activation magnitudes.
Conclusions:
- Despite poor decoder generalization, BMIs exhibit robust online control due to adaptable neural population activity.
- Neural trajectory shifts offer insights into predicting muscle activation and improving BMI performance in dynamic environments.
- Findings suggest potential for enhancing BMI robustness for clinical applications in motor function restoration.
