Temporal information encoding in isolated cortical networks

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Sensory stimuli often contain time-dependent features crucial for perception.
• The brain represents stimulus timing using spatial and temporal codes within sensory cortices.
• Reservoir computing, a model of recurrent neural networks, is a potential mechanism for temporal-to-spatial information conversion.

Purpose of the Study:

• To investigate whether isolated cortical networks, acting as reservoirs, can represent temporal information present in sensory inputs.
• To determine the precision and duration of temporal information encoding within these networks.

Main Methods:

• Utilized patterned optogenetic stimulation of dissociated primary rat cortical cultures.
• Delivered input sequences with varying temporal patterns to assess network state representation.
• Analyzed network states to classify temporal features of the input sequences.

Main Results:

• Network states contained information about input sequences for over 1 second with at least 100-ms precision.
• Accurate classification of temporal information relied on a population code involving many neurons.
• The trajectory of the network state was primarily influenced by spatial stimulus features, with temporal features having a subtler impact.
• Spatial information was retained for over 2 seconds, comparable to short-term memory durations in the visual cortex.

Conclusions:

• Isolated cortical networks exhibit properties consistent with reservoir computing.
• These findings suggest that local reservoir computation is a plausible mechanism for temporal-to-spatial code conversion in sensory cortices.
• The study provides experimental evidence for the neural basis of temporal information processing in the brain.