NEuroMOrphic Neural-Response Decoding System for Adaptive and Personalized Neuro-Prosthetics' Control.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Previous work established a Motor Control Decoder (MCD) using electrocorticography (ECoG) for tetraplegic movement intention.
• Patient satisfaction with effector actions is crucial but not guaranteed by intended movement labels.
• A need exists to classify brain signals as satisfactory or unsatisfactory to refine decoding accuracy.

Purpose of the Study:

• To upgrade the neuromorphic MCD with a Neural Response Decoder (NRD) for predicting ECoG signal satisfaction.
• To design an actor-critic framework for reinforcement learning adaptation of MCD using NRD.
• To enhance MCD performance by incorporating satisfaction feedback into the decoding process.

Main Methods:

• Developed an NRD to predict ECoG signal satisfaction.
• Trained the NRD using ECoG signals, MCD predictions, and prescribed intended movements.
• Implemented an actor-critic structure for reinforcement learning adaptation of MCD based on NRD.
• Evaluated the NRD's accuracy and its contribution to MCD performance.

Main Results:

• The trained NRD achieved satisfactory accuracy in predicting ECoG signal satisfaction.
• The NRD integration contributed to improved MCD performance.
• The actor-critic structure demonstrated potential for adaptive MCD optimization.

Conclusions:

• The developed NRD is a promising tool for enhancing brain-computer interface accuracy in tetraplegia.
• Further research is needed for real-time online optimization of MCD using NRD.
• Incorporating direct patient feedback could further improve MCD-NRD system accuracy.