Quality metrics of spike sorting using neighborhood components analysis

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Microelectrode extracellular recording enables simultaneous recording of multi-neuron activity.
• Quantitative metrics for assessing spike sorting cluster quality are currently lacking.
• Identifying high-quality clusters is crucial for reliable single-unit analysis.

Purpose of the Study:

• To introduce an objective measure for evaluating the quality of spike clusters.
• To assess the proposed method's effectiveness in discriminating well- and poorly-separated clusters.

Main Methods:

• Developed an objective measure based on neighborhood component analysis for spike cluster evaluation.
• Tested the proposed method using simulated and experimental extracellular recordings.
• Compared the new metric against established measures like isolation distance and L ratio.

Main Results:

• The proposed method demonstrated a strong correlation with spike sorting accuracy in both simulations and real data.
• Effectively discriminated between well-separated and poorly-separated neuronal clusters.
• Validated using data from the rodent primary visual cortex.

Conclusions:

• The developed objective measure provides a reliable way to assess spike cluster quality.
• This method can enhance the accuracy of single-unit analysis in neuroscience research.
• Applicable to any study analyzing single-neuron activity from extracellular recordings.