Cortical activity flips among quasi-stationary states

Area of Science:

• Computational Neuroscience
• Systems Neuroscience

Background:

• Analyzing complex neural data requires sophisticated statistical models.
• Understanding dynamic changes in neuronal firing rates is crucial for deciphering brain function.

Purpose of the Study:

• To segment parallel spike train recordings into distinct hidden states using a hidden Markov model.
• To investigate the biological validity and behavioral relevance of these identified neural states.

Main Methods:

• Treated parallel spike trains as a multivariate Poisson process with time-varying firing rates.
• Applied a hidden Markov model to segment neural recordings into statistically discriminated hidden states.
• Examined segmentation consistency with animal behavior, collective activity flips, and neuronal cross-correlations.

Main Results:

• Successfully segmented neural recordings into 6-8 distinct, well-separated activity states with approximately stationary firing rates.
• Observed fast transitions between states, marked by coordinated changes in multiple neurons' firing rates.
• Found that different behavioral modes and stimuli consistently mapped to specific neural states.

Conclusions:

• The hidden Markov model effectively captures distinct neural states reflecting cooperative neuronal action.
• These states are biologically valid, correlate with behavior, and exhibit dynamic changes in neuronal communication.
• The model provides a powerful framework for analyzing complex, multi-neuron activity patterns.