Leveraging neural dynamics to extend functional lifetime of brain-machine interfaces

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Intracortical brain-machine interfaces (BMIs) decode neural activity to restore motor function.
• Declining neural signal quality over time is a major challenge for long-term BMI performance.
• Existing BMIs suffer from reduced efficacy as electrode signal degradation occurs.

Purpose of the Study:

• To develop a software-based algorithmic technique to extend the functional lifetime of BMIs.
• To improve BMI performance despite a significant loss of recorded neural signals.
• To enhance the clinical viability of neuroprosthetic devices.

Main Methods:

• Developed a novel decoder algorithm that incorporates historical neural population dynamics.
• Implemented the technique entirely in software for seamless integration.
• Validated the approach through closed-loop experiments in non-human primates (rhesus macaques).

Main Results:

• The proposed algorithm significantly outperformed state-of-the-art decoders after a 60% loss of recording electrodes.
• Performance improvements were observed as a factor of 3.2x and 1.7x, recovering 46% and 22% of maximal performance, respectively.
• Neural population dynamics in the motor cortex appear invariant to the number of recorded neurons.

Conclusions:

• The developed algorithmic technique effectively extends functional brain-machine interface lifetime.
• This approach offers a promising solution to the challenge of signal degradation in implanted devices.
• The findings suggest increased clinical feasibility for long-term use of BMIs in individuals with neurological deficits.