Improving spike sorting efficiency with separability index and spectral clustering

Area of Science:

• Computational Neuroscience
• Signal Processing

Background:

• Spike sorting is crucial for analyzing neural activity.
• Accurate spike sorting is challenging due to signal variability.

Purpose of the Study:

• To evaluate spectral clustering for spike sorting.
• To introduce a Separability Index for assessing spike sorting difficulty.
• To compare spectral clustering with existing methods.

Main Methods:

• Spectral clustering applied to various feature sets (raw samples, derivatives, PCA).
• Development and validation of a novel Separability Index.
• Comparative analysis against two established spike sorting algorithms.

Main Results:

• Raw samples achieved 73.84% accuracy with spectral clustering.
• The Separability Index effectively predicted sorting difficulty (correlation=0.71).
• Proposed method showed up to 23% higher accuracy than prior techniques.

Conclusions:

• Spectral clustering is a viable method for spike sorting, especially with raw data.
• The Separability Index offers a predictive tool for computational cost and method comparison.
• This approach enhances efficiency and accuracy in neural data analysis.