A Novel Brain-Computer Interface Architecture: The Brain-Muscle-Hand Interface for replicating the motor pathway

TL;DR

This paper introduces the Brain-Muscle-Hand Interface (BMHI), a novel BCI system that decodes cortical EEG to reconstruct muscle activity, enabling natural control of prosthetics with improved accuracy and stability over traditional BCIs.

Contribution

The study presents the BMHI architecture that mimics the motor pathway, achieving higher accuracy and efficiency in BCI control compared to conventional methods.

Findings

Prediction accuracy of 0.79 achieved.

Training time reduced by approximately eighteenfold.

Enables stable online control of virtual and robotic hands.

Abstract

Myoelectric interfaces enable intuitive and natural control by decoding residual muscle activity, providing an effective pathway for motor restoration in individuals with preserved musculature. However, in patients with severe muscular atrophy or high-level spinal cord injury, the absence of reliable muscle activity renders myoelectric control infeasible. In such cases, motor brain-computer interfaces (BCIs) offer an alternative route. However, conventional brain-computer interface systems rely mainly on noisy cortical signals and classification-based decoding algorithms, which often result in low signal fidelity, limited controllability, and unstable real-time performance. Inspired by the motor pathway--an evolutionarily optimized system that filters, integrates, and transmits motor commands from the brain to the muscles--this study proposes the Brain-Muscle-Hand Interface (BMHI). BMHI decodes cortical EEG signals to reconstruct muscle-level EMG activity, functionally substituting for the muscles and enabling regression-based, continuous, and natural control via a myoelectric interface. To validate this architecture, we performed offline verification, comparative analysis, and online control experiments. Results demonstrate that: (1) the BMHI achieves a prediction accuracy of 0.79; (2) compared with conventional end-to-end brain-hand interfaces, it reduces training time by approximately eighteenfold while improving decoding accuracy; and (3) in online operation, the BMHI enables stable and efficient manipulation of both a virtual hand and a robotic arm. Compared with conventional BCIs, the BMHI, by replicating the motor pathway, enables continuous, stable, and naturally intuitive control.