Tracking Neural Modulation Depth by Dual Sequential Monte Carlo Estimation on Point Processes for Brain-Machine...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Neuroscience

Background:

• Classic brain-machine interfaces (BMIs) decode neural signals for prosthetic control but suffer from performance decay due to adaptive changes in brain activity.
• Static neural tuning models used in BMIs have limitations, leading to instability and reduced decoding accuracy over time.
• Adaptive changes, such as altered preferred direction and gain modulation, are observed in neural activity during BMI control.

Purpose of the Study:

• To develop a novel computational approach for stable and accurate decoding of neural signals in brain-machine interfaces.
• To model and decode the gradually changing modulation depth of individual neurons during motor tasks.
• To overcome the limitations of static neural tuning models in BMIs.

Main Methods:

• Proposed a dual sequential Monte Carlo adaptive point process method for real-time neural decoding.
• Utilized multichannel neural spike trains from the primary motor cortex of a monkey performing a target pursuit task.
• Compared the performance of the adaptive model against classic static neural tuning models using simulated and real kinematic data.

Main Results:

• The proposed adaptive point process method successfully tracked neural modulation depth over time.
• Demonstrated a better goodness-of-fit compared to classic static neural tuning models.
• Achieved smaller errors between true kinematics and estimated movements in both simulated and real datasets.

Conclusions:

• The brain may utilize adaptive neural tuning strategies to achieve stable motor output.
• Plastic neural tuning is a fundamental feature of neural systems.
• The developed adaptive algorithm can enhance BMI performance, enabling more complex and controlled movements for users.