In quest of the missing neuron: spike sorting based on dominant-sets clustering

Area of Science:

• Computational Neuroscience
• Signal Processing
• Data Analysis

Background:

• Spike sorting algorithms decompose extracellular signals to identify individual neuron activity.
• Determining the precise number of active neurons remains challenging due to factors like sparse firing and background noise.

Purpose of the Study:

• To introduce a graph-theoretical algorithmic procedure for resolving the issue of neuron number determination in spike sorting.
• To enhance the performance of spike sorting methods in complex extracellular recordings.

Main Methods:

• Utilizing dimensionality reduction techniques.
• Implementing a modern, efficient, and progressively executable clustering routine.
• Employing a graph-theoretical algorithmic approach.

Main Results:

• The proposed method demonstrates higher performance standards compared to popular spike sorting techniques.
• Validation using simulated data across various signal-to-noise ratio (SNR) levels confirms the method's efficacy.
• Successfully resolves the issue of determining the number of active neurons.

Conclusions:

• The developed graph-theoretical algorithm offers a superior approach to spike sorting.
• The combination of dimensionality reduction and advanced clustering effectively addresses challenges posed by sparse neural activity and background noise.
• This method provides a reliable solution for accurate neuron number identification in electrophysiological recordings.