A motor BCI reads neural signals from motor cortex — the brain region responsible for planning and executing voluntary movements — and translates them into commands for external devices. Motor BCIs are the most clinically advanced category of brain-computer interface, with multiple devices in human clinical trials for restoring movement and communication to people with paralysis.
Recording Targets
Motor BCIs typically record from:
- Primary motor cortex (M1): Contains neurons that encode movement direction, velocity, and force. The hand-knob area of M1 is the most common recording target for cursor and arm control.
- Dorsal premotor cortex (PMd): Encodes movement planning and preparation before execution. Signals from PMd can predict intended movements before they begin.
- Supplementary motor area (SMA): Involved in movement sequencing and bilateral coordination.
Output Modalities
- Cursor control: 2D cursor movement for computer interaction (the foundational motor BCI application)
- Robotic arm control: 3D or higher-dimensional control of robotic arms for reaching, grasping, and manipulation
- Functional electrical stimulation (FES): Decoded motor commands drive electrical stimulation of the user's own paralyzed muscles, restoring hand grasp and arm movement
- Exoskeleton control: BCI-driven powered exoskeletons for walking and upper limb movement
- Wheelchair control: BCI-directed powered wheelchair navigation
Clinical Systems
The major clinical motor BCI systems include:
- BrainGate: Utah Array in motor cortex, Kalman filter decoder, wired percutaneous connector. The longest-running human motor BCI program.
- Neuralink PRIME: N1 implant with 1,024 channels, wireless data and power, custom decoder. First human implant in 2024.
- Synchron Stentrode: Endovascular BCI recording motor cortex signals through the superior sagittal sinus.
Performance
Motor BCI performance has improved dramatically. Early demonstrations (2004) achieved basic 1D cursor control. Current systems achieve multi-dimensional continuous control with throughputs exceeding 8 bits per second. The ReFIT Kalman filter (Gilja et al., 2015) approximately doubled clinical motor BCI performance, and deep learning decoders continue to push the frontier.