A general likelihood framework for characterizing the time course of neural activity

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Accurate estimation of neural firing rates is crucial for understanding brain function.
• Current methods for neural firing rate estimation often lack optimal parameter selection, impacting accuracy.
• Temporal smoothing parameters significantly influence the reliability of firing rate estimates.

Purpose of the Study:

• To introduce a general likelihood-based framework for estimating neural firing rates.
• To develop a method for selecting optimal temporal smoothing parameters that maximize the likelihood of observed data.
• To demonstrate the framework's applicability and improvements in estimation accuracy.

Main Methods:

• Developed a general, algorithm-independent likelihood-based framework.
• Applied the framework to peristimulus time histograms and kernel smoothers.
• Utilized the general point process likelihood as a cost function for parameter optimization.
• Performed simulation studies and applied the method to real experimental spike train data.

Main Results:

• The framework enables principled selection of temporal smoothing parameters.
• Demonstrated substantial improvements in estimation accuracy for basic rate estimation algorithms.
• The resulting kernel smoother accurately determines bandwidth from individual spike trains.
• The likelihood framework effectively predicts missing data by optimizing bandwidth.

Conclusions:

• The general likelihood framework offers a principled approach to neural firing rate estimation.
• This method enhances the accuracy and reliability of common electrophysiological analysis techniques.
• The framework is versatile and can be integrated with more advanced estimation methods.