Decoding sensorimotor information from superior parietal lobule of macaque via Convolutional Neural Networks

Area of Science:

• Neuroscience
• Computational Neuroscience
• Motor Control

Background:

• The posterior parietal cortex (PPC) is crucial for sensorimotor transformations guiding actions.
• Specific encoding properties within different PPC subregions, like V6A, PEc, and PE, remain unclear.
• Understanding these dynamics is key for developing advanced neuroprosthetic devices.

Purpose of the Study:

• To investigate how reaching movement information is differentially encoded in V6A, PEc, and PE areas of the monkey PPC.
• To explore the application of Convolutional Neural Networks (CNNs) for decoding neuronal activity related to reaching movements.
• To assess the potential of these decoding methods for future neuroprosthetic applications.

Main Methods:

• Trained two macaque monkeys on a delayed reaching task in 3D space.
• Recorded single-cell activity from V6A, PEc, and PE areas.
• Utilized CNNs, optimized with Bayesian Optimization, to decode target positions and reaching trajectories from neuronal firing patterns.

Main Results:

• V6A and PEc areas demonstrated superior spatial position decoding compared to the PE area.
• Decoding accuracy increased significantly after target instruction and plateaued at movement onset.
• Neuronal decoding revealed dynamic encoding of reaching movement phases across interconnected PPC areas.

Conclusions:

• Sensorimotor information for reaching movements is dynamically encoded and distributed across a network of PPC areas.
• CNN-based decoding of neuronal firing rates offers a powerful tool to understand PPC function.
• These findings have implications for designing sophisticated neuroprosthetic devices capable of interpreting user intentions.