Identifying distinct neural features between the initial and corrective phases of precise reaching using AutoLFADS

Area of Science:

• Neuroscience
• Motor Control
• Computational Neuroscience

Background:

• Initial movements often require corrective actions, but the neural basis for these corrections is not fully understood.
• The motor cortex's role in transitioning from initial to corrective movements remains unclear.
• Existing methods struggle to capture the distinct neural encoding of corrective movements.

Approach:

• Recorded neural population activity from non-human primates during precision reaching.
• Applied AutoLFADS, a deep learning model, to analyze single-trial neural dynamics.
• Developed and tested state-dependent decoders for improved movement prediction.

Key Points:

• Neural decoding of reach velocity poorly generalized from initial to corrective submovements.
• Corrective submovements originate from distinct neural states compared to baseline.
• State-dependent decoders significantly improved the prediction of corrective movement velocity.

Conclusions:

• Motor cortex exhibits unique neural dynamics for initial and corrective submovements.
• Neural activity encodes complex combinations of position and velocity specific to submovement types.
• Traditional decoding methods are insufficient for capturing the nuances of online corrective movements.