Spike Neural Network of Motor Cortex Model for Arm Reaching Control.

Area of Science:

• Computational neuroscience
• Neural modeling
• Motor control systems

Background:

• Recurrent neural networks (RNNs) are used for motor cortex modeling.
• Existing RNN models use continuous signals, not biological spike signals.
• Accurate modeling requires simulating spike-based neural activity.

Purpose of the Study:

• To develop a recurrent spike neural network (RSNN) for motor cortex simulation.
• To model neuronal activity during arm-reaching tasks.
• To bridge the gap between continuous and spike-based neural network models.

Main Methods:

• Implemented an RSNN using integrate-and-fire spiking neurons and conductance-based synapses.
• Designed neural interconnections with distinct "fast" and "slow" firing timescales.
• Simulated motor cortical activity during a virtual arm-reaching task.

Main Results:

• The RSNN model demonstrated high agreement with monkey motor cortex data.
• Single-cell and population-level neuronal activity closely matched experimental results.
• A quantitative correlation coefficient of 0.89 was achieved between model and real data.

Conclusions:

• The proposed RSNN effectively simulates motor cortical dynamics.
• The model's success suggests the presence of multiple timescales in motor cortical control.
• This work advances the use of biologically plausible models in neuroscience.