Spike sorting: Bayesian clustering of non-stationary data.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Spike sorting is crucial for analyzing neural activity, but is labor-intensive due to data non-stationarity.
• Existing methods struggle with the dynamic nature of neural recordings.

Purpose of the Study:

• To develop an automated, Bayesian framework for spike sorting that accounts for non-stationary neural data.
• To improve the efficiency and accuracy of clustering neural spikes.

Main Methods:

• Modeled source neurons as non-stationary Gaussian mixtures within a Bayesian framework.
• Employed a two-stage approach: frame-wise mixture estimation and probabilistic model optimization.
• Utilized a Gaussian Jensen-Shannon divergence for computing transition probabilities.

Main Results:

• The proposed method significantly outperformed existing spike sorting techniques on synthetic non-stationary data.
• High agreement was observed between the automated method and human experts on real neural spike data.
• Achieved comparable performance in both fully unsupervised and semi-supervised modes.

Conclusions:

• The Bayesian framework effectively handles non-stationary neural data for automated spike sorting.
• This approach offers a robust and efficient alternative to manual spike sorting, achieving near-human accuracy.