Integrating Neuro-Ocular Data for Accurate Hand Movement Decoding in Brain-Controlled Robotic Systems.

Brain-computer interfaces (BCIs) are rapidly advancing, particularly in cognitive areas like the Posterior Parietal Cortex (PPC). This study aims to enhance continuous hand movement decoding by integrating neural and eye movement data for direct robotic prosthetic control. We leveraged single-neuron activity from the PPC area V6A in macaques during two different reaching tasks—a Fixation-to-reach and a Constant-gaze task—combined with eye coordinates data to form a comprehensive neuro-ocular dataset. Convolutional (CNN) and recurrent (RNN) neural networks were employed for decoding hand movements. Our findings reveal that incorporating ocular data significantly improves decoding accuracy, with the RNN-based model outperforming others. This research also explores intuitive, non-invasive systems for robotic arm control using temporary surrogate interfaces instead of decoded velocities. These advancements not only deepen our understanding of neural decoding but also pave the way for sophisticated assistive technologies and enhanced robotic control systems.