Spike sorting using locality preserving projection with gap statistics and landmark-based spectral clustering

Area of Science:

• Neuroscience
• Computational Biology
• Biomedical Engineering

Background:

• Analyzing electrophysiological data is crucial for understanding neural functions.
• Spike sorting, the process of assigning neural spikes to their sources, is a key challenge in data analysis.
• Existing automated spike sorting methods present trade-offs between accuracy and computational cost.

Purpose of the Study:

• To develop a novel, unsupervised spike sorting method that is both highly accurate and computationally efficient.
• To address the limitations of current spike sorting techniques in terms of performance and speed.

Main Methods:

• Feature extraction using Locality Preserving Projection (LPP).
• Clustering using Landmark-based Spectral Clustering (LSC).
• Determining the optimal number of clusters using Gap Statistics (GS) prior to LSC.

Main Results:

• The proposed LPP-LSC method achieves high accuracy due to discriminative LPP spike features.
• LPP-LSC demonstrates significant computational cost reduction compared to existing methods.
• Integration with GS enhances the efficiency and performance of the LSC clustering algorithm.

Conclusions:

• LPP-LSC provides a superior spike sorting solution balancing accuracy and computational efficiency.
• The linear nature of LPP and LSC algorithms minimizes computational burden.
• This method is suitable for real-time spike analysis applications.