Hidden Markov Models Predict the Future Choice Better Than a PSTH-Based Method.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Traditional analysis of neuronal activity often relies on average firing rates, which may not capture the full complexity of neural processing.
• Hidden Markov Models (HMMs) have emerged as a powerful tool for analyzing dynamic changes in neural ensemble activity over time.
• Understanding collective neural dynamics is essential for decoding information encoded in neural signals during complex cognitive tasks.

Discussion:

• This study compared the efficacy of HMMs against a classification method based on peristimulus time histograms (PSTH) for decoding neural activity.
• The methods were tested using prefrontal cortex neural recordings from monkeys performing a strategy task involving cue-based target selection.
• The accuracy of both methods was evaluated based on their ability to predict the monkeys' choices from single-trial neural activity.

Key Insights:

• Both HMMs and PSTH-based methods could decode monkeys' choices above chance levels using pre-action neural activity.
• HMMs demonstrated significantly higher accuracy than PSTH-based methods, even when HMM performance was modest.
• The accuracy of both decoding methods correlated with the number of spatially selective neurons identified within a session.

Outlook:

• Future research should explore advanced dynamic models beyond average firing rates to fully understand neuronal ensemble function.
• Investigating how collective neural dynamics contribute to decision-making and behavioral strategies remains a key area for future research.
• The findings suggest that incorporating temporal dynamics is fundamental for a comprehensive understanding of neural information processing.